https://infocepo.com/wiki/api.php?action=feedcontributions&feedformat=atom&user=Tcepo
Essential - User contributions [en]
2026-04-05T15:36:35Z
User contributions
MediaWiki 1.39.13
https://infocepo.com/wiki/index.php?title=LAB_project&diff=1943
LAB project
2026-04-04T12:16:17Z
<p>Tcepo: /* Futur */</p>
<hr />
<div>Resilient, low power and encrypted LAB CLOUD project.<br />
<br />
==Diagram==<br />
[[file:Infocepo.drawio.png]]<br />
==Data security==<br />
* Availability is via [[LAB project#HA]] & RSYNC & another location & another internet provider & public DNS<br />
* Integrity is provided by BTRFS and will be provided by MINIO for large static files<br />
* Theft protection is provided by AC and FDE encryption<br />
* Loss protection is performed via AC & RSYNC & BTRFS and locks/unlocks FDE drives<br />
* Scalability will be via the CEPH<br />
<br />
==My LAB tools==<br />
*MEDIAWIKI (sharing editor)<br />
*DEBIAN<br />
*[https://app.diagrams.net app.diagrams.net] (diagram editors)<br />
*x86 CPU (for ARM I will use QEMU emulator)<br />
*OPENCL GPU (AI&SIGNAL processing)<br />
*LXD/SNAP/MICROK8S (virtualization&container)<br />
*LSOF (system&network audit)<br />
*NMAP (Network scanner)<br />
*TCPDUMP (Network monitoring)<br />
<br />
==Tested in this LAB==<br />
*NEXTCLOUD [https://nextcloud.infocepo.com/index.php/s/oKrHsMX9bmxcjMf nextcloud.infocepo.com] (aFM8yKYmDa)<br />
*MEDIAWIKI [https://infocepo.com/wiki infocepo.com]<br />
*KUBERNETES Cluster [https://wiki.infocepo.com/wiki wiki.infocepo.com] (demo available, ask me) <br />
*MARIADB GALERA Cluster<br />
*OPENSTACK <br />
*PROMETHEUS [https://prometheus.infocepo.com prometheus.infocepo.com] (demo available, ask me)<br />
*YACY [https://google.infocepo.com google.infocepo.com] (decentralized search engine) (demo available, ask me)<br />
*GLUSTERFS<br />
*KATA (container runtime like a VM)<br />
*AWS CLI<br />
*ANSIBLE<br />
*GIT<br />
<br />
==HA==<br />
I love COROSYNC/PACEMAKER, but for this LAB I wanted something from scratch:<br />
<syntaxhighlight lang="bash"><br />
#!/usr/bin/env bash<br />
# ynotopec at gmail.com<br />
<br />
set -u<br />
<br />
domainName="$(<domainName)"<br />
portNumber="$(<portNumber)"<br />
publicHost1="$(<publicHost1)"<br />
publicHost2="$(<publicHost2)"<br />
<br />
[ -n "$domainName" ] || exit 1<br />
[ -n "$portNumber" ] || exit 1<br />
[ -n "$publicHost1" ] || exit 1<br />
[ -n "$publicHost2" ] || exit 1<br />
<br />
command -v dig >/dev/null 2>&1 || exit 1<br />
command -v nc >/dev/null 2>&1 || exit 1<br />
command -v rsync >/dev/null 2>&1 || exit 1<br />
command -v ping >/dev/null 2>&1 || exit 1<br />
command -v flock >/dev/null 2>&1 || exit 1<br />
<br />
exec 9>/tmp/"${domainName}".ha.lock<br />
flock -n 9 || exit 0<br />
<br />
mkdir -p /storage/rsync-rollback ha_synced<br />
<br />
resolve_ipv4() {<br />
dig +time=2 +tries=1 +short A "$1" | awk 'NF{print; exit}'<br />
}<br />
<br />
echo "$(date): init"<br />
<br />
# Clean<br />
stop"${domainName}".sh<br />
<br />
# If I was the master sleep a little !<br />
ipPublic="$(dig +time=2 +tries=1 +short myip.opendns.com @resolver1.opendns.com | awk 'NF{print; exit}')"<br />
ipMasterOld="$(resolve_ipv4 "${domainName}")"<br />
[ -n "${ipPublic}" ] && [ "${ipPublic}" = "${ipMasterOld}" ] && sleep 240<br />
<br />
# Wait Internet<br />
while ! ping -w2 -c1 "${publicHost1}" >/dev/null 2>&1; do<br />
sleep 10<br />
done<br />
<br />
# Wait Admin Unlock Backup<br />
waitAdminUnLockBCK.sh<br />
<br />
# If passive merge backup<br />
if nc -zw2 "${domainName}" 443 >/dev/null 2>&1 || { sleep 10; nc -zw4 "${domainName}" 443 >/dev/null 2>&1; }; then<br />
echo "$(date): merge backup from ${domainName}:443"<br />
rsync --max-size=4M --ignore-existing --numeric-ids --modify-window=1 --ignore-errors --block-size=128.00K --inplace --no-whole-file \<br />
-z --compress-level=9 \<br />
-aAXx \<br />
--exclude=.rsync_* \<br />
--rsh="ssh -i ~/.ssh/storage@${domainName}.key -p ${portNumber} -oStrictHostKeyChecking=no" \<br />
"${domainName}:/storage/rsync-rollback/" /storage/rsync-rollback/<br />
fi<br />
<br />
# lock bck source after merge<br />
lockBckSource.sh<br />
<br />
# Wait master down (big loop)<br />
while nc -zw2 "${domainName}" "${portNumber}" >/dev/null 2>&1 || { sleep 10; nc -zw4 "${domainName}" "${portNumber}" >/dev/null 2>&1; }; do<br />
if [ -z "$(find ha_synced -type f -mmin -60 2>/dev/null)" ]; then<br />
echo "$(date): sync from ${domainName}:${portNumber}"<br />
rsync --numeric-ids --delete --force --modify-window=1 --ignore-errors --block-size=128.00K --inplace --no-whole-file \<br />
-z --compress-level=9 \<br />
-aAXx \<br />
--backup-dir="rsync-rollback/$(date '+%Y-%m-%d')" \<br />
--exclude=rsync-rollback \<br />
--exclude=.rsync_* \<br />
--rsh="ssh -i ~/.ssh/storage@${domainName}.key -p ${portNumber} -oStrictHostKeyChecking=no" \<br />
"${domainName}:/storage/" /storage/ && touch ha_synced<br />
fi<br />
sleep 10<br />
done<br />
<br />
echo "$(date): master detected down on ${domainName}:${portNumber}"<br />
<br />
# Maybe it's me down ! If INTERNET down, reboot<br />
if ! ping -w2 -c1 "${publicHost1}" >/dev/null 2>&1; then<br />
echo "$(date): reboot because connectivity confirmation failed"<br />
reboot<br />
fi<br />
<br />
# Become master<br />
<br />
# lock Backup (integrity protection against attacks)<br />
lockBCK.sh<br />
<br />
# Register DNS<br />
ipMasterDown="$(resolve_ipv4 "${domainName}")"<br />
[ -n "${ipPublic}" ] || reboot<br />
updateDns.sh "${ipPublic}"<br />
echo "$(date): dns updated to ${ipPublic}"<br />
<br />
# Random startup time<br />
sleep $((RANDOM % 10))<br />
"${domainName}"Start.sh &<br />
<br />
# Wait DNS propagation<br />
sleep 215<br />
<br />
# Monitor<br />
echo "$(date): up"<br />
[ "${ipPublic}" = "${ipMasterDown}" ] && ipMasterDown=""<br />
while [ "${ipPublic}" = "$(resolve_ipv4 "${domainName}")" ] \<br />
&& { [ -z "${ipMasterDown}" ] || ! nc -zw2 "${ipMasterDown}" 443 >/dev/null 2>&1; } \<br />
&& { nc -zw2 "${ipPublic}" 443 >/dev/null 2>&1 || nc -zw4 "${ipPublic}" 443 >/dev/null 2>&1; } \<br />
&& { [ -n "${ipMasterDown}" ] && nc -zw2 "${ipMasterDown}" "${portNumber}" >/dev/null 2>&1 || ping -w2 -c1 "${publicHost1}" >/dev/null 2>&1 || ping -w2 -c1 "${publicHost2}" >/dev/null 2>&1; }; do<br />
sleep 4<br />
done<br />
echo "$(date): down"<br />
<br />
# Stop<br />
stop"${domainName}".sh<br />
<br />
reboot<br />
</syntaxhighlight><br />
<br />
==Optimizations explained==<br />
To reduce Network data I added options to rsync:<br />
*only send modified data blocks from files<br />
--inplace --block-size=128.00K<br />
*compress transfer<br />
-z --compress-level=9<br />
To increase flash storage life:<br />
--inplace --no-whole-file<br />
<br />
==Cost per month==<br />
<pre><br />
800€*2=1600€ :hypervisors (popular CPU X86 notebook, 8GB+)<br />
-400€*2=-800€ :sale after one year<br />
800€/12m=33€ :price/month<br />
30*2=60€ :Internet (~1PB outbound/month)<br />
(7.5W+7.5W)*24h*30.5d/1000Wh*.15€*2~=3.3€ :electricity<br />
<br />
33+60+3.3~=96€ :cost/month<br />
</pre><br />
<br />
==Futur==<br />
*Migrate rsync to real time replication.<br />
*Add S3 for objects storage and streaming.<br />
*Improve synchronization algorithm ([[bisync.sh]]):<br />
#The NEXTCLOUD algorithm is very good average, but is very poor for dynamic big files like DB or virtual images.<br />
#OSYNC is slow, and I don't know for big files.<br />
#RSYNC is very good average but not dealing with inodes (moving files).<br />
*Alert passive location locked and add de-lock page.<br />
*Double storage to have active/active locations. Storage 1 location A to B. Storage 2 location B to A.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=LAB_project&diff=1942
LAB project
2026-04-04T12:15:18Z
<p>Tcepo: /* HA */</p>
<hr />
<div>Resilient, low power and encrypted LAB CLOUD project.<br />
<br />
==Diagram==<br />
[[file:Infocepo.drawio.png]]<br />
==Data security==<br />
* Availability is via [[LAB project#HA]] & RSYNC & another location & another internet provider & public DNS<br />
* Integrity is provided by BTRFS and will be provided by MINIO for large static files<br />
* Theft protection is provided by AC and FDE encryption<br />
* Loss protection is performed via AC & RSYNC & BTRFS and locks/unlocks FDE drives<br />
* Scalability will be via the CEPH<br />
<br />
==My LAB tools==<br />
*MEDIAWIKI (sharing editor)<br />
*DEBIAN<br />
*[https://app.diagrams.net app.diagrams.net] (diagram editors)<br />
*x86 CPU (for ARM I will use QEMU emulator)<br />
*OPENCL GPU (AI&SIGNAL processing)<br />
*LXD/SNAP/MICROK8S (virtualization&container)<br />
*LSOF (system&network audit)<br />
*NMAP (Network scanner)<br />
*TCPDUMP (Network monitoring)<br />
<br />
==Tested in this LAB==<br />
*NEXTCLOUD [https://nextcloud.infocepo.com/index.php/s/oKrHsMX9bmxcjMf nextcloud.infocepo.com] (aFM8yKYmDa)<br />
*MEDIAWIKI [https://infocepo.com/wiki infocepo.com]<br />
*KUBERNETES Cluster [https://wiki.infocepo.com/wiki wiki.infocepo.com] (demo available, ask me) <br />
*MARIADB GALERA Cluster<br />
*OPENSTACK <br />
*PROMETHEUS [https://prometheus.infocepo.com prometheus.infocepo.com] (demo available, ask me)<br />
*YACY [https://google.infocepo.com google.infocepo.com] (decentralized search engine) (demo available, ask me)<br />
*GLUSTERFS<br />
*KATA (container runtime like a VM)<br />
*AWS CLI<br />
*ANSIBLE<br />
*GIT<br />
<br />
==HA==<br />
I love COROSYNC/PACEMAKER, but for this LAB I wanted something from scratch:<br />
<syntaxhighlight lang="bash"><br />
#!/usr/bin/env bash<br />
# ynotopec at gmail.com<br />
<br />
set -u<br />
<br />
domainName="$(<domainName)"<br />
portNumber="$(<portNumber)"<br />
publicHost1="$(<publicHost1)"<br />
publicHost2="$(<publicHost2)"<br />
<br />
[ -n "$domainName" ] || exit 1<br />
[ -n "$portNumber" ] || exit 1<br />
[ -n "$publicHost1" ] || exit 1<br />
[ -n "$publicHost2" ] || exit 1<br />
<br />
command -v dig >/dev/null 2>&1 || exit 1<br />
command -v nc >/dev/null 2>&1 || exit 1<br />
command -v rsync >/dev/null 2>&1 || exit 1<br />
command -v ping >/dev/null 2>&1 || exit 1<br />
command -v flock >/dev/null 2>&1 || exit 1<br />
<br />
exec 9>/tmp/"${domainName}".ha.lock<br />
flock -n 9 || exit 0<br />
<br />
mkdir -p /storage/rsync-rollback ha_synced<br />
<br />
resolve_ipv4() {<br />
dig +time=2 +tries=1 +short A "$1" | awk 'NF{print; exit}'<br />
}<br />
<br />
echo "$(date): init"<br />
<br />
# Clean<br />
stop"${domainName}".sh<br />
<br />
# If I was the master sleep a little !<br />
ipPublic="$(dig +time=2 +tries=1 +short myip.opendns.com @resolver1.opendns.com | awk 'NF{print; exit}')"<br />
ipMasterOld="$(resolve_ipv4 "${domainName}")"<br />
[ -n "${ipPublic}" ] && [ "${ipPublic}" = "${ipMasterOld}" ] && sleep 240<br />
<br />
# Wait Internet<br />
while ! ping -w2 -c1 "${publicHost1}" >/dev/null 2>&1; do<br />
sleep 10<br />
done<br />
<br />
# Wait Admin Unlock Backup<br />
waitAdminUnLockBCK.sh<br />
<br />
# If passive merge backup<br />
if nc -zw2 "${domainName}" 443 >/dev/null 2>&1 || { sleep 10; nc -zw4 "${domainName}" 443 >/dev/null 2>&1; }; then<br />
echo "$(date): merge backup from ${domainName}:443"<br />
rsync --max-size=4M --ignore-existing --numeric-ids --modify-window=1 --ignore-errors --block-size=128.00K --inplace --no-whole-file \<br />
-z --compress-level=9 \<br />
-aAXx \<br />
--exclude=.rsync_* \<br />
--rsh="ssh -i ~/.ssh/storage@${domainName}.key -p ${portNumber} -oStrictHostKeyChecking=no" \<br />
"${domainName}:/storage/rsync-rollback/" /storage/rsync-rollback/<br />
fi<br />
<br />
# lock bck source after merge<br />
lockBckSource.sh<br />
<br />
# Wait master down (big loop)<br />
while nc -zw2 "${domainName}" "${portNumber}" >/dev/null 2>&1 || { sleep 10; nc -zw4 "${domainName}" "${portNumber}" >/dev/null 2>&1; }; do<br />
if [ -z "$(find ha_synced -type f -mmin -60 2>/dev/null)" ]; then<br />
echo "$(date): sync from ${domainName}:${portNumber}"<br />
rsync --numeric-ids --delete --force --modify-window=1 --ignore-errors --block-size=128.00K --inplace --no-whole-file \<br />
-z --compress-level=9 \<br />
-aAXx \<br />
--backup-dir="rsync-rollback/$(date '+%Y-%m-%d')" \<br />
--exclude=rsync-rollback \<br />
--exclude=.rsync_* \<br />
--rsh="ssh -i ~/.ssh/storage@${domainName}.key -p ${portNumber} -oStrictHostKeyChecking=no" \<br />
"${domainName}:/storage/" /storage/ && touch ha_synced<br />
fi<br />
sleep 10<br />
done<br />
<br />
echo "$(date): master detected down on ${domainName}:${portNumber}"<br />
<br />
# Maybe it's me down ! If INTERNET down, reboot<br />
if ! ping -w2 -c1 "${publicHost1}" >/dev/null 2>&1; then<br />
echo "$(date): reboot because connectivity confirmation failed"<br />
reboot<br />
fi<br />
<br />
# Become master<br />
<br />
# lock Backup (integrity protection against attacks)<br />
lockBCK.sh<br />
<br />
# Register DNS<br />
ipMasterDown="$(resolve_ipv4 "${domainName}")"<br />
[ -n "${ipPublic}" ] || reboot<br />
updateDns.sh "${ipPublic}"<br />
echo "$(date): dns updated to ${ipPublic}"<br />
<br />
# Random startup time<br />
sleep $((RANDOM % 10))<br />
"${domainName}"Start.sh &<br />
<br />
# Wait DNS propagation<br />
sleep 215<br />
<br />
# Monitor<br />
echo "$(date): up"<br />
[ "${ipPublic}" = "${ipMasterDown}" ] && ipMasterDown=""<br />
while [ "${ipPublic}" = "$(resolve_ipv4 "${domainName}")" ] \<br />
&& { [ -z "${ipMasterDown}" ] || ! nc -zw2 "${ipMasterDown}" 443 >/dev/null 2>&1; } \<br />
&& { nc -zw2 "${ipPublic}" 443 >/dev/null 2>&1 || nc -zw4 "${ipPublic}" 443 >/dev/null 2>&1; } \<br />
&& { [ -n "${ipMasterDown}" ] && nc -zw2 "${ipMasterDown}" "${portNumber}" >/dev/null 2>&1 || ping -w2 -c1 "${publicHost1}" >/dev/null 2>&1 || ping -w2 -c1 "${publicHost2}" >/dev/null 2>&1; }; do<br />
sleep 4<br />
done<br />
echo "$(date): down"<br />
<br />
# Stop<br />
stop"${domainName}".sh<br />
<br />
reboot<br />
</syntaxhighlight><br />
<br />
==Optimizations explained==<br />
To reduce Network data I added options to rsync:<br />
*only send modified data blocks from files<br />
--inplace --block-size=128.00K<br />
*compress transfer<br />
-z --compress-level=9<br />
To increase flash storage life:<br />
--inplace --no-whole-file<br />
<br />
==Cost per month==<br />
<pre><br />
800€*2=1600€ :hypervisors (popular CPU X86 notebook, 8GB+)<br />
-400€*2=-800€ :sale after one year<br />
800€/12m=33€ :price/month<br />
30*2=60€ :Internet (~1PB outbound/month)<br />
(7.5W+7.5W)*24h*30.5d/1000Wh*.15€*2~=3.3€ :electricity<br />
<br />
33+60+3.3~=96€ :cost/month<br />
</pre><br />
<br />
==Futur==<br />
*Migrate rsync to real time replication.<br />
*Add MINIO for objects storage and streaming.<br />
*Improve synchronization algorithm ([[bisync.sh]]):<br />
#The NEXTCLOUD algorithm is very good average, but is very poor for dynamic big files like DB or virtual images.<br />
#OSYNC is slow, and I don't know for big files.<br />
#RSYNC is very good average but not dealing with inodes (moving files).<br />
*Alert passive location locked and add de-lock page.<br />
*Double storage to have active/active locations. Storage 1 location A to B. Storage 2 location B to A.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=LAB_project&diff=1941
LAB project
2026-04-04T11:46:32Z
<p>Tcepo: /* HA */</p>
<hr />
<div>Resilient, low power and encrypted LAB CLOUD project.<br />
<br />
==Diagram==<br />
[[file:Infocepo.drawio.png]]<br />
==Data security==<br />
* Availability is via [[LAB project#HA]] & RSYNC & another location & another internet provider & public DNS<br />
* Integrity is provided by BTRFS and will be provided by MINIO for large static files<br />
* Theft protection is provided by AC and FDE encryption<br />
* Loss protection is performed via AC & RSYNC & BTRFS and locks/unlocks FDE drives<br />
* Scalability will be via the CEPH<br />
<br />
==My LAB tools==<br />
*MEDIAWIKI (sharing editor)<br />
*DEBIAN<br />
*[https://app.diagrams.net app.diagrams.net] (diagram editors)<br />
*x86 CPU (for ARM I will use QEMU emulator)<br />
*OPENCL GPU (AI&SIGNAL processing)<br />
*LXD/SNAP/MICROK8S (virtualization&container)<br />
*LSOF (system&network audit)<br />
*NMAP (Network scanner)<br />
*TCPDUMP (Network monitoring)<br />
<br />
==Tested in this LAB==<br />
*NEXTCLOUD [https://nextcloud.infocepo.com/index.php/s/oKrHsMX9bmxcjMf nextcloud.infocepo.com] (aFM8yKYmDa)<br />
*MEDIAWIKI [https://infocepo.com/wiki infocepo.com]<br />
*KUBERNETES Cluster [https://wiki.infocepo.com/wiki wiki.infocepo.com] (demo available, ask me) <br />
*MARIADB GALERA Cluster<br />
*OPENSTACK <br />
*PROMETHEUS [https://prometheus.infocepo.com prometheus.infocepo.com] (demo available, ask me)<br />
*YACY [https://google.infocepo.com google.infocepo.com] (decentralized search engine) (demo available, ask me)<br />
*GLUSTERFS<br />
*KATA (container runtime like a VM)<br />
*AWS CLI<br />
*ANSIBLE<br />
*GIT<br />
<br />
==HA==<br />
I love COROSYNC/PACEMAKER, but for this LAB I wanted something from scratch:<br />
<syntaxhighlight lang="bash"><br />
#!/bin/bash<br />
#ynotopec at gmail.com<br />
<br />
domainName=$(cat domainName)<br />
portNumber=$(cat portNumber)<br />
publicHost1=$(cat publicHost1)<br />
publicHost2=$(cat publicHost2)<br />
<br />
echo "$(date): init"<br />
<br />
#Clean<br />
stop${domainName}.sh<br />
<br />
#If I was the master sleep a little !<br />
ipPublic="$(dig +short myip.opendns.com @resolver1.opendns.com )"<br />
ipMasterOld="$(nslookup ${domainName} |sed -rn 's#^Address: (.*)$#\1#p' )"<br />
[ "${ipPublic}" = "${ipMasterOld}" ] &&sleep 240<br />
<br />
#Wait Internet<br />
while (! ping -w2 -c1 ${publicHost1} >/dev/null 2>&1 );do<br />
sleep 10<br />
done<br />
<br />
#Wait Admin Unlock Backup<br />
waitAdminUnLockBCK.sh<br />
<br />
#If passif merge backup<br />
if (nc -zw2 ${domainName} 443 >/dev/null 2>&1 ||(sleep 10 ;nc -zw4 ${domainName} 443 >/dev/null 2>&1 ) ) ;then<br />
rsync --max-size=4M --ignore-existing --numeric-ids --modify-window=1 --ignore-errors --block-size=128.00K --inplace --no-whole-file \<br />
-z --compress-level=9 \<br />
-aAXxv \<br />
--exclude=.rsync_* \<br />
--rsh="ssh -i ~/.ssh/storage@${domainName}.key -p ${portNumber} -oStrictHostKeyChecking=no" \<br />
${domainName}:/storage/rsync-rollback/ /storage/rsync-rollback/<br />
fi<br />
<br />
#lock bck source after merge<br />
lockBckSource.sh<br />
<br />
#Wait master down (big loop)<br />
while (nc -zw2 ${domainName} ${portNumber} >/dev/null 2>&1 ||(sleep 10 ;nc -zw4 ${domainName} ${portNumber} >/dev/null 2>&1 ) );do<br />
if [ -z "$(find ha_synced -type f -mmin -60 2>/dev/null )" ] ;then<br />
rsync --numeric-ids --delete --force --modify-window=1 --ignore-errors --block-size=128.00K --inplace --no-whole-file \<br />
-z --compress-level=9 \<br />
-aAXxv \<br />
--backup-dir=rsync-rollback/$(date '+%Y-%m-%d') \<br />
--exclude=rsync-rollback \<br />
--exclude=.rsync_* \<br />
--rsh="ssh -i ~/.ssh/storage@${domainName}.key -p ${portNumber} -oStrictHostKeyChecking=no" \<br />
${domainName}:/storage/ /storage/ &&touch ha_synced<br />
fi<br />
sleep 10<br />
done<br />
<br />
#Maybe it's me down ! If INTERNET down, reboot<br />
! ping -w2 -c1 ${publicHost1} >/dev/null 2>&1 &&reboot<br />
<br />
#Become master<br />
<br />
#lock Backup (integrity protection against attacks)<br />
lockBCK.sh<br />
<br />
#Register DNS<br />
ipMasterDown="$(nslookup ${domainName} |sed -rn 's#^Address: (.*)$#\1#p' )"<br />
updateDns.sh "${ipPublic}"<br />
<br />
#Random startup time<br />
sleep $((RANDOM%10))<br />
${domainName}Start.sh &<br />
<br />
#Wait DNS propagation<br />
sleep $((240-25))<br />
<br />
#Monitor<br />
echo "$(date): up"<br />
[ "${ipPublic}" = "${ipMasterDown}" ] &&ipMasterDown=""<br />
while [ "${ipPublic}" = "$(nslookup ${domainName} |sed -rn 's#^Address: (.*)$#\1#p' )" ] \<br />
&&( ! nc -zw2 "${ipMasterDown}" 443 >/dev/null 2>&1 ) \<br />
&&(nc -zw2 "${ipPublic}" 443 >/dev/null 2>&1 ||nc -zw4 "${ipPublic}" 443 >/dev/null 2>&1 ) \<br />
&&(nc -zw2 "${ipMasterDown}" ${portNumber} >/dev/null 2>&1 ||ping -w2 -c1 ${publicHost1} >/dev/null 2>&1 ||ping -w2 -c1 ${publicHost2} >/dev/null 2>&1 ) ;do<br />
sleep 4<br />
done<br />
echo "$(date): down"<br />
<br />
#Stop<br />
stop${domainName}.sh<br />
<br />
reboot<br />
</syntaxhighlight><br />
<br />
==Optimizations explained==<br />
To reduce Network data I added options to rsync:<br />
*only send modified data blocks from files<br />
--inplace --block-size=128.00K<br />
*compress transfer<br />
-z --compress-level=9<br />
To increase flash storage life:<br />
--inplace --no-whole-file<br />
<br />
==Cost per month==<br />
<pre><br />
800€*2=1600€ :hypervisors (popular CPU X86 notebook, 8GB+)<br />
-400€*2=-800€ :sale after one year<br />
800€/12m=33€ :price/month<br />
30*2=60€ :Internet (~1PB outbound/month)<br />
(7.5W+7.5W)*24h*30.5d/1000Wh*.15€*2~=3.3€ :electricity<br />
<br />
33+60+3.3~=96€ :cost/month<br />
</pre><br />
<br />
==Futur==<br />
*Migrate rsync to real time replication.<br />
*Add MINIO for objects storage and streaming.<br />
*Improve synchronization algorithm ([[bisync.sh]]):<br />
#The NEXTCLOUD algorithm is very good average, but is very poor for dynamic big files like DB or virtual images.<br />
#OSYNC is slow, and I don't know for big files.<br />
#RSYNC is very good average but not dealing with inodes (moving files).<br />
*Alert passive location locked and add de-lock page.<br />
*Double storage to have active/active locations. Storage 1 location A to B. Storage 2 location B to A.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1940
Main Page
2026-04-04T10:48:37Z
<p>Tcepo: /* Nouveautés 21/03/2026 */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [[File:Ailab-architecture.png|thumb|'''Infra architecture overview''']]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ailab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://api-audio2txt.ailab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://api-txt2audio.ailab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ailab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ailab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ailab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ailab.infocepo.com/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ailab.infocepo.com/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ailab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ailab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ailab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ailab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ailab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate-rt.ailab.infocepo.com Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ailab.infocepo.com Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''gemma4''' : Gemma 4 models are designed to deliver frontier-level performance at each size. They are well-suited for reasoning, agentic workflows, coding, and multimodal understanding.<br />
<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ailab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ailab.infocepo.com '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ailab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-03-30''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || translategemma, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || glm-4.7-flash – tâches agentiques, raisonnement et explication de code<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ailab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ailab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ailab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ailab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ailab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ailab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ailab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ailab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ailab.infocepo.com translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ailab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ailab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ailab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ailab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ailab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ailab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ailab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ailab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ailab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ailab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ailab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ailab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ailab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ailab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ailab.infocepo.com:wait-2026-06 registry.ailab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ailab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ailab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ailab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ailab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ailab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1939
Main Page
2026-03-30T15:46:21Z
<p>Tcepo: /* Modèles ouverts & endpoints internes */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [[File:Ailab-architecture.png|thumb|'''Infra architecture overview''']]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ailab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://api-audio2txt.ailab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://api-txt2audio.ailab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ailab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ailab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ailab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ailab.infocepo.com/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ailab.infocepo.com/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ailab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ailab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ailab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ailab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ailab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate-rt.ailab.infocepo.com Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ailab.infocepo.com Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ailab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ailab.infocepo.com '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ailab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-03-30''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || translategemma, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || glm-4.7-flash – tâches agentiques, raisonnement et explication de code<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ailab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ailab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ailab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ailab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ailab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ailab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ailab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ailab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ailab.infocepo.com translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ailab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ailab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ailab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ailab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ailab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ailab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ailab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ailab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ailab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ailab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ailab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ailab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ailab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ailab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ailab.infocepo.com:wait-2026-06 registry.ailab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ailab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ailab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ailab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ailab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ailab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1938
Main Page
2026-03-30T13:10:16Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [[File:Ailab-architecture.png|thumb|'''Infra architecture overview''']]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ailab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://api-audio2txt.ailab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://api-txt2audio.ailab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ailab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ailab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ailab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ailab.infocepo.com/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ailab.infocepo.com/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ailab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ailab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ailab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ailab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ailab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate-rt.ailab.infocepo.com Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ailab.infocepo.com Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ailab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ailab.infocepo.com '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ailab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-02-13''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – raisonnement et explication de code<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – auto-complétion rapide<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|-<br />
| '''gpt-oss-20b''' || tâches agentiques<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ailab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ailab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ailab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ailab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ailab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ailab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ailab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ailab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ailab.infocepo.com translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ailab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ailab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ailab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ailab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ailab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ailab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ailab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ailab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ailab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ailab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ailab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ailab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ailab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ailab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ailab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ailab.infocepo.com:wait-2026-06 registry.ailab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ailab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ailab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ailab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ailab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ailab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1937
Main Page
2026-03-30T11:57:01Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [[File:Ailab-architecture.png|thumb|'''Infra architecture overview''']]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://api-audio2txt.ailab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://api-txt2audio.ailab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ai.lab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ai.lab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ai.lab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ai.lab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ai.lab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ai.lab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate.ai.lab.infocepo.com:wait-2026-06 Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ai.lab.infocepo.com:wait-2026-06 Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ai.lab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ai.lab.infocepo.com:wait-2026-06 '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ai.lab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-02-13''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – raisonnement et explication de code<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – auto-complétion rapide<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|-<br />
| '''gpt-oss-20b''' || tâches agentiques<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ai.lab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ai.lab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ai.lab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ai.lab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ai.lab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ai.lab.infocepo.com:wait-2026-06 translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ai.lab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ai.lab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ai.lab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ai.lab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ai.lab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ai.lab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ai.lab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ai.lab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ai.lab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ai.lab.infocepo.com:wait-2026-06 registry.ai.lab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ai.lab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ai.lab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ai.lab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ai.lab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ai.lab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1936
Main Page
2026-03-30T11:56:05Z
<p>Tcepo: /* Vue d’ensemble */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [[File:Ailab-architecture.png|'''Infra architecture overview'''|50%]]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://api-audio2txt.ailab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://api-txt2audio.ailab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ai.lab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ai.lab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ai.lab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ai.lab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ai.lab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ai.lab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate.ai.lab.infocepo.com:wait-2026-06 Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ai.lab.infocepo.com:wait-2026-06 Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ai.lab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ai.lab.infocepo.com:wait-2026-06 '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ai.lab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-02-13''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – raisonnement et explication de code<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – auto-complétion rapide<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|-<br />
| '''gpt-oss-20b''' || tâches agentiques<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ai.lab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ai.lab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ai.lab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ai.lab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ai.lab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ai.lab.infocepo.com:wait-2026-06 translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ai.lab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ai.lab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ai.lab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ai.lab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ai.lab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ai.lab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ai.lab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ai.lab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ai.lab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ai.lab.infocepo.com:wait-2026-06 registry.ai.lab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ai.lab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ai.lab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ai.lab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ai.lab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ai.lab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1935
Main Page
2026-03-30T11:55:45Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [[File:Ailab-architecture.png|'''Infra architecture overview''']]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://api-audio2txt.ailab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://api-txt2audio.ailab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ai.lab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ai.lab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ai.lab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ai.lab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ai.lab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ai.lab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate.ai.lab.infocepo.com:wait-2026-06 Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ai.lab.infocepo.com:wait-2026-06 Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ai.lab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ai.lab.infocepo.com:wait-2026-06 '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ai.lab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-02-13''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – raisonnement et explication de code<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – auto-complétion rapide<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|-<br />
| '''gpt-oss-20b''' || tâches agentiques<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ai.lab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ai.lab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ai.lab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ai.lab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ai.lab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ai.lab.infocepo.com:wait-2026-06 translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ai.lab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ai.lab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ai.lab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ai.lab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ai.lab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ai.lab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ai.lab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ai.lab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ai.lab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ai.lab.infocepo.com:wait-2026-06 registry.ai.lab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ai.lab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ai.lab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ai.lab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ai.lab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ai.lab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=File:Ailab-architecture.png&diff=1934
File:Ailab-architecture.png
2026-03-30T11:54:38Z
<p>Tcepo: </p>
<hr />
<div></div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1932
Main Page
2026-03-28T20:20:52Z
<p>Tcepo: /* Catalogue rapide des services */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [https://github.com/ynotopec/infra-archi-overview/tree/main Infra architecture overview]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://api-audio2txt.ailab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://api-txt2audio.ailab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ai.lab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ai.lab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ai.lab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ai.lab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ai.lab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ai.lab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate.ai.lab.infocepo.com:wait-2026-06 Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ai.lab.infocepo.com:wait-2026-06 Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ai.lab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ai.lab.infocepo.com:wait-2026-06 '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ai.lab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-02-13''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – raisonnement et explication de code<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – auto-complétion rapide<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|-<br />
| '''gpt-oss-20b''' || tâches agentiques<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ai.lab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ai.lab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ai.lab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ai.lab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ai.lab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ai.lab.infocepo.com:wait-2026-06 translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ai.lab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ai.lab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ai.lab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ai.lab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ai.lab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ai.lab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ai.lab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ai.lab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ai.lab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ai.lab.infocepo.com:wait-2026-06 registry.ai.lab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ai.lab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ai.lab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ai.lab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ai.lab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ai.lab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=K8s-users&diff=1931
K8s-users
2026-03-26T20:45:40Z
<p>Tcepo: </p>
<hr />
<div>===AUTOMATED===<br />
* Define :<br />
<syntaxhighlight lang="bash"><br />
export usersList=<br />
</syntaxhighlight><br />
* Execute :<br />
<syntaxhighlight lang="bash"><br />
mkdir -p ~/old &&\<br />
cd ~/old &&\<br />
curl https://infocepo.com/wiki/index.php/Special:Export/K8s-users 2>/dev/null |tac |sed -r '0,/'"#"'24cc42#/d' |tac |sed -r '0,/'"#"'24cc42#/d' |sed 's/'"&"'amp;/\&/g;s/'"&"'gt;/>/g;s/'"&"'lt;/</g' >$$ &&\<br />
bash $$ &&\<br />
cd - >/dev/null<br />
</syntaxhighlight><br />
====code====<br />
<syntaxhighlight lang="bash"><br />
#24cc42#<br />
cd<br />
mkdir -p old<br />
cd old<br />
#bashRc=$(find /etc -type f -name "*bashrc" |grep -vw /etc/skel/ |head -1 )<br />
#sed -i '/kube/d' /home/*/.bashrc ${bashRc}<br />
#echo 'export KUBECONFIG=${HOME}/.kube/config' >>${bashRc}<br />
if [ -z "${usersList}" ] ;then<br />
cat /etc/passwd |grep :/home/ |cut -d':' -f1<br />
else<br />
echo "${usersList}"<br />
fi |while read userLogin ;do<br />
grep 'export KUBECONFIG=${HOME}/.kube/config' /home/${userLogin}/.bashrc >/dev/null 2>&1 ||(echo 'export KUBECONFIG=${HOME}/.kube/config' >>/home/${userLogin}/.bashrc )<br />
nameSpace=${userLogin}<br />
#apiUrl=$(cat ${KUBECONFIG} |sed -rn 's#^[[:space:]]*server:[[:space:]]*([[:graph:]]+)[[:space:]]*$#\1#p' |tail -1 )<br />
apiUrl=$(kubectl config view --raw |sed -rn 's#^[[:space:]]*server:[[:space:]]*([[:graph:]]+)[[:space:]]*$#\1#p' |tail -1 )<br />
userHome=$(cat /etc/passwd |grep ^${userLogin}: |cut -d: -f6 )<br />
#https://kubernetes.io/docs/reference/access-authn-authz/certificate-signing-requests/#normal-user<br />
openssl req -new -nodes -subj "/CN=${userLogin}" \<br />
-keyout ${userLogin}.key -out ${userLogin}.csr<br />
kubectl delete csr ${userLogin} 2>/dev/null<br />
cat <<EOF |kubectl apply -f -<br />
apiVersion: certificates.k8s.io/v1<br />
kind: CertificateSigningRequest<br />
metadata:<br />
name: ${userLogin}<br />
spec:<br />
request: $(cat ${userLogin}.csr |base64 |tr -d '\n' )<br />
signerName: kubernetes.io/kube-apiserver-client<br />
expirationSeconds: 31536000 #86400 # one day<br />
usages:<br />
- client auth<br />
EOF<br />
kubectl certificate approve ${userLogin} &&sleep 2 &&\<br />
kubectl get csr ${userLogin} -o jsonpath='{.status.certificate}'|base64 -d >${userLogin}.crt<br />
kubectl create namespace ${nameSpace} 2>/dev/null<br />
kubectl create rolebinding ${userLogin} --clusterrole=admin --user=${userLogin} --namespace=${nameSpace}<br />
#NAMESPACE NETWORK ISOLATION<br />
cat <<EOF |kubectl -n ${nameSpace} apply -f -<br />
kind: NetworkPolicy<br />
apiVersion: networking.k8s.io/v1<br />
metadata:<br />
name: allow-same-namespace<br />
spec:<br />
podSelector:<br />
matchLabels:<br />
ingress:<br />
- from:<br />
- podSelector: {}<br />
---<br />
apiVersion: networking.k8s.io/v1<br />
kind: NetworkPolicy<br />
metadata:<br />
name: allow-from-ingress<br />
spec:<br />
ingress:<br />
- from:<br />
- namespaceSelector:<br />
matchLabels:<br />
kubernetes.io/metadata.name: ingress<br />
podSelector: {}<br />
policyTypes:<br />
- Ingress<br />
---<br />
apiVersion: networking.k8s.io/v1<br />
kind: NetworkPolicy<br />
metadata:<br />
name: allow-from-observability<br />
spec:<br />
ingress:<br />
- from:<br />
- namespaceSelector:<br />
matchLabels:<br />
kubernetes.io/metadata.name: observability<br />
podSelector: {}<br />
policyTypes:<br />
- Ingress<br />
---<br />
apiVersion: networking.k8s.io/v1<br />
kind: NetworkPolicy<br />
metadata:<br />
name: allow-from-openshift-ingress<br />
spec:<br />
ingress:<br />
- from:<br />
- namespaceSelector:<br />
matchLabels:<br />
policy-group.network.openshift.io/ingress: ""<br />
podSelector: {}<br />
policyTypes:<br />
- Ingress<br />
---<br />
apiVersion: networking.k8s.io/v1<br />
kind: NetworkPolicy<br />
metadata:<br />
name: allow-from-openshift-monitoring<br />
spec:<br />
ingress:<br />
- from:<br />
- namespaceSelector:<br />
matchLabels:<br />
network.openshift.io/policy-group: monitoring<br />
podSelector: {}<br />
policyTypes:<br />
- Ingress<br />
EOF<br />
#KUBECONFIG<br />
apiUri=$(echo $apiUrl |cut -d'/' -f3 |tr '.' '-' )<br />
mkdir -p ${userHome}/.kube<br />
cat <<EOT >${userHome}/.kube/config<br />
apiVersion: v1<br />
clusters:<br />
- cluster:<br />
insecure-skip-tls-verify: true<br />
server: ${apiUrl}<br />
name: ${apiUri}<br />
contexts:<br />
- context:<br />
cluster: ${apiUri}<br />
namespace: ${nameSpace}<br />
user: ${userLogin}/${apiUri}<br />
name: ${nameSpace}/${apiUri}/${userLogin}<br />
current-context: ${nameSpace}/${apiUri}/${userLogin}<br />
kind: Config<br />
preferences: {}<br />
users:<br />
- name: ${userLogin}/${apiUri}<br />
user:<br />
client-certificate-data: $(cat ${userLogin}.crt |base64 |tr -d '\n' )<br />
client-key-data: $(cat ${userLogin}.key |base64 |tr -d '\n' )<br />
EOT<br />
chmod ug+rw,o="" -R ${userHome}/.kube<br />
find ${userHome}/.kube -xdev -type d -exec chmod ug+x {} \;<br />
chown ${userLogin}: -R ${userHome}/.kube<br />
cp -aZ ${userHome}/.kube/config ${userHome}/.kubeconfig<br />
done<br />
#24cc42#<br />
</syntaxhighlight><br />
<br />
==OPENAI summary==<br />
This script creates a new directory "old" in the home directory and changes to that directory. It then downloads the content of a URL (https://infocepo.com/wiki/index.php/Special:Export/K8s-users) and saves it to a file. The file is then executed by the command bash $$ (where $$ is the process id of the current shell). The script then changes back to the original directory and removes any reference to kube in the .bashrc file of all users in the system.<br />
<br />
The script then goes on to define the environment variable KUBECONFIG which is the path to the Kubernetes cluster configuration file. If the usersList environment variable is empty, it retrieves the list of all users in the system, otherwise it uses the value of usersList. For each user, the script creates a new certificate signing request (CSR), approves it, retrieves the certificate and creates a new namespace in the Kubernetes cluster.<br />
<br />
The script then creates several network policies for the namespace that are used to define network isolation for that namespace. Finally, the script creates a new Kubernetes configuration file for the user and saves it in the .kube directory in their home directory.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1930
Main Page
2026-03-26T18:10:22Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [https://github.com/ynotopec/infra-archi-overview/tree/main Infra architecture overview]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://stt.ai.lab.infocepo.com:wait-2026-06/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://tts.ai.lab.infocepo.com:wait-2026-06/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ai.lab.infocepo.com:wait-2026-06 API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ai.lab.infocepo.com:wait-2026-06 ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ai.lab.infocepo.com:wait-2026-06 Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ai.lab.infocepo.com:wait-2026-06/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ai.lab.infocepo.com:wait-2026-06 Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ai.lab.infocepo.com:wait-2026-06/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ai.lab.infocepo.com:wait-2026-06 DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate.ai.lab.infocepo.com:wait-2026-06 Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ai.lab.infocepo.com:wait-2026-06 Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ai.lab.infocepo.com:wait-2026-06 '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ai.lab.infocepo.com:wait-2026-06 '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ai.lab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-02-13''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – raisonnement et explication de code<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – auto-complétion rapide<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|-<br />
| '''gpt-oss-20b''' || tâches agentiques<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Création du token : [https://llm-token.ai.lab.infocepo.com:wait-2026-06 OPENAI_API_KEY]<br />
* Documentation : [https://api.ai.lab.infocepo.com:wait-2026-06 Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ai.lab.infocepo.com:wait-2026-06/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ai.lab.infocepo.com:wait-2026-06/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ai.lab.infocepo.com:wait-2026-06/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ai.lab.infocepo.com:wait-2026-06/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ai.lab.infocepo.com:wait-2026-06/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ai.lab.infocepo.com:wait-2026-06 translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ai.lab.infocepo.com:wait-2026-06/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ai.lab.infocepo.com:wait-2026-06/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ai.lab.infocepo.com:wait-2026-06/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ai.lab.infocepo.com:wait-2026-06/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ai.lab.infocepo.com:wait-2026-06/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ai.lab.infocepo.com:wait-2026-06/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ai.lab.infocepo.com:wait-2026-06/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ai.lab.infocepo.com:wait-2026-06</code><br />
* URL utilisée par MRSIE : <code>https://tei.ai.lab.infocepo.com:wait-2026-06/</code><br />
* Documentation : [https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ai.lab.infocepo.com:wait-2026-06/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ai.lab.infocepo.com:wait-2026-06</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ai.lab.infocepo.com:wait-2026-06<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ai.lab.infocepo.com:wait-2026-06<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ai.lab.infocepo.com:wait-2026-06 registry.ai.lab.infocepo.com:wait-2026-06]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ai.lab.infocepo.com:wait-2026-06/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ai.lab.infocepo.com:wait-2026-06 \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ai.lab.infocepo.com:wait-2026-06</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ai.lab.infocepo.com:wait-2026-06 datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ai.lab.infocepo.com:wait-2026-06/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1929
Main Page
2026-03-26T16:15:06Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki documente l’écosystème '''Cloud, IA, automatisation et lab''' d’Infocepo. <br />
Il s’adresse aux :<br />
<br />
* administrateurs systèmes,<br />
* ingénieurs cloud,<br />
* développeurs,<br />
* étudiants,<br />
* curieux qui veulent apprendre en pratiquant.<br />
<br />
L’objectif est simple : transformer la théorie en '''scripts réutilisables, schémas, architectures, APIs et laboratoires concrets'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
== Portail principal ==<br />
* [https://infocepo.com infocepo.com]<br />
<br />
== Assistant IA ==<br />
* [https://chat.infocepo.com Chat assistant]<br />
<br />
== Liste des pages du wiki ==<br />
* [[Special:AllPages|Toutes les pages]]<br />
<br />
== Vue d’ensemble ==<br />
* [https://github.com/ynotopec/infra-archi-overview/tree/main Infra architecture overview]<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Construire un assistant IA privé<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter un modèle de chat et un modèle de résumé<br />
* Brancher des données internes via '''RAG + embeddings'''<br />
<br />
; 2. Lancer un lab cloud<br />
* Créer un petit cluster Kubernetes, OpenStack ou bare-metal<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA : transcription, résumé, chatbot, OCR…<br />
<br />
; 3. Préparer un audit ou une migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Concevoir l’architecture cible<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Vue d’ensemble du contenu ==<br />
* '''Guides IA & outils''' : assistants, modèles, évaluation, GPU, RAG<br />
* '''Cloud & infrastructure''' : Kubernetes, OpenStack, HA, HPC, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automatisation<br />
* '''Comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Vision =<br />
<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
Le but à long terme est de construire un environnement où :<br />
<br />
* les assistants IA privés accélèrent la production,<br />
* les tâches répétitives sont automatisées,<br />
* les déploiements sont industrialisés,<br />
* l’infrastructure reste '''compréhensible, portable et réutilisable'''.<br />
<br />
[[File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png|thumb|right|Main page summary]]<br />
<br />
----<br />
<br />
= Catalogue rapide des services =<br />
<br />
{| class="wikitable"<br />
|+ Services principaux<br />
! Catégorie !! Service !! Lien !! Rôle<br />
|-<br />
| API || LLM || [https://api.ai.lab.infocepo.com API LLM] || Modèles de chat, code, RAG, OCR<br />
|-<br />
| API || STT || [https://stt.ai.lab.infocepo.com/docs API STT] || Transcription audio<br />
|-<br />
| API || TTS || [https://tts.ai.lab.infocepo.com/docs API TTS] || Synthèse vocale<br />
|-<br />
| API || Realtime AI || [https://github.com/ynotopec/api-realtime-ai api-realtime-ai] || Temps réel WebSocket / WebRTC<br />
|-<br />
| API || Image to Text || [https://api.ai.lab.infocepo.com API LLM] || OCR / VLM via endpoint dédié<br />
|-<br />
| API || Summary || [https://api-summary.ai.lab.infocepo.com/docs API Summary] || Résumé de textes longs<br />
|-<br />
| API || Text Embeddings || [https://text-embeddings.ai.lab.infocepo.com/docs Text Embeddings] || Embeddings pour RAG<br />
|-<br />
| API || ChromaDB || [https://chromadb.ai.lab.infocepo.com ChromaDB] || Base vecteur<br />
|-<br />
| API || Text to Image || [https://api-txt2image.ai.lab.infocepo.com/docs TXT2IMAGE] || Génération d’images<br />
|-<br />
| API || Diarization || [https://api-diarization.ai.lab.infocepo.com/docs Diarization] || Segmentation locuteurs<br />
|-<br />
| Observabilité || Monitoring || [https://grafana.ai.lab.infocepo.com Grafana] || Dashboards techniques<br />
|-<br />
| Observabilité || Status || [https://uptime-kuma.ai.lab.infocepo.com/status/ai Uptime Kuma] || Disponibilité des services<br />
|-<br />
| Observabilité || Web stats || [https://web-stat.c1.ai.lab.infocepo.com Web Stat] || Statistiques web<br />
|-<br />
| Observabilité || LLM stats || [https://api.ai.lab.infocepo.com/ui LLM Stat] || Vue API / usage<br />
|-<br />
| Outils || DataLab || [https://datalab.ai.lab.infocepo.com DataLab] || Environnement de travail hors-production<br />
|-<br />
| Outils || Translation UI || [https://translate.ai.lab.infocepo.com Translation] || Traduction<br />
|-<br />
| Outils || Demos || [https://demos.ai.lab.infocepo.com Demos] || Démonstrateurs<br />
|}<br />
<br />
----<br />
<br />
= Nouveautés =<br />
<br />
== Nouveautés 21/03/2026 ==<br />
* Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.<br />
* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.<br />
* Ajout de [https://localai.ai.lab.infocepo.com '''localai'''] : infrastructure locale unifiée pour STT / TTS / LLM.<br />
* DGX Spark : architecture CPU ARM.<br />
* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.<br />
* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.<br />
* Mise à jour des paramètres '''RAG optimisation'''.<br />
* Ajout de [https://github.com/ynotopec/coder-brain/blob/main/first-architecture.md '''experimental brains'''].<br />
* Ajout de [https://github.com/ynotopec/legal-agent '''legal-agent'''].<br />
* Ajout de [https://github.com/ynotopec/ai-security '''ai-security'''].<br />
* Ajout de [https://langextract.ai.lab.infocepo.com '''langextract'''] : démo extraction d’entités.<br />
* Ajout de [https://sam-audio.c1.ai.lab.infocepo.com '''sam-audio'''] : séparation audio sémantique.<br />
* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.<br />
* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.<br />
* Ajout de '''gpt-oss''' : modèles open-weight conçus pour raisonnement et tâches agentiques.<br />
<br />
----<br />
<br />
= Priorités =<br />
<br />
== Top tasks ==<br />
* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.<br />
* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.<br />
* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.<br />
* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.<br />
* Ajouter [https://github.com/vllm-project/guidellm '''GuideLLM'''] : capacity planning / benchmark réaliste.<br />
* Ajouter [https://github.com/NVIDIA-NeMo/Guardrails '''NeMo Guardrails'''] : garde-fous et politiques.<br />
<br />
== Backlog / veille ==<br />
* OPENRAG > implement / evaluate / add OIDC<br />
* short audio transcription<br />
* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]<br />
* RAG sur PDF avec images<br />
* compatibilité Open WebUI avec Agentic RAG<br />
* scalability<br />
* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]<br />
* [https://github.com/openclaw/openclaw openclaw]<br />
* faster-whisper mutualisé<br />
* API classificateur IA<br />
* API résumé mutualisée<br />
* API KV (LDAP user / group)<br />
* API NER<br />
* parsing structuré docs : granite-docling + meilisearch<br />
* Temporal pour workflows critiques<br />
* [https://github.com/appwrite/appwrite appwrite]<br />
* [https://github.com/vllm-project/semantic-router semantic-router]<br />
* [https://github.com/KeygraphHQ/shannon Shannon]<br />
* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]<br />
* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]<br />
* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]<br />
* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]<br />
* [https://github.com/resemble-ai/chatterbox chatterbox]<br />
* deepset-ai/haystack<br />
* meilisearch<br />
* [https://huggingface.co/ibm-granite/granite-docling-258M granite-docling-258M]<br />
* Airbyte<br />
* [https://github.com/Aider-AI/aider aider]<br />
* [https://github.com/continuedev/continue continue]<br />
* OpenHands<br />
* N8N<br />
* API Compressor<br />
* LightRAG<br />
* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]<br />
* Metabase<br />
* browser-use<br />
* MCP LLM<br />
* Dify<br />
* Rasa<br />
* supabase<br />
* mem0<br />
* DeepResearch<br />
* AppFlowy<br />
* dx8152/Qwen-Edit-2509-Multiple-angles<br />
<br />
----<br />
<br />
= Assistants IA & outils cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chatgpt.com ChatGPT] – Assistant conversationnel public, utile pour exploration, rédaction, expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://ollama.com Ollama] + GPU <br />
: Stack typique pour assistant privé, API OpenAI-compatible et expérimentation locale.<br />
* [https://github.com/ynotopec/summarize Private summary] <br />
: Outil de résumé local, rapide et hors ligne.<br />
<br />
== Développement, modèles & veille ==<br />
<br />
; '''Découverte de modèles'''<br />
* [https://ollama.com/library LLM Trending]<br />
* [https://huggingface.co/models Models Trending]<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation]<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation]<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]<br />
* [https://ann-benchmarks.com Vectors DB Ranking]<br />
* [https://top500.org/lists/green500/ HPC Efficiency]<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending]<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning]<br />
* [https://www.perplexity.ai Perplexity AI]<br />
<br />
== Matériel IA & GPU ==<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100]<br />
* NVIDIA 5080<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]<br />
<br />
----<br />
<br />
= Modèles ouverts & endpoints internes =<br />
<br />
''Dernière mise à jour : 2026-02-13''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – chat généraliste, bon compromis coût / qualité<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – traduction déterministe et reproductible<br />
|-<br />
| '''ai-summary''' || qwen3 – résumé de textes longs<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – raisonnement et explication de code<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – auto-complétion rapide<br />
|-<br />
| '''ai-parse''' || qwen3 – extraction structurée, parsing logs / JSON / tableaux<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG en français<br />
|-<br />
| '''gpt-oss-20b''' || tâches agentiques<br />
|}<br />
<br />
----<br />
<br />
= API Realtime AI (DEV) =<br />
<br />
'''Statut :''' environnement DEV, remplaçante prévue de l’API OpenAI pour les cas temps réel.<br />
<br />
== Configuration ==<br />
{| class="wikitable"<br />
! Variable !! Valeur<br />
|-<br />
| OPENAI_API_BASE || <code>wss://api-realtime-ai.ai.lab.infocepo.com/v1</code><br />
|-<br />
| OPENAI_API_KEY || <code>sk-XXXXX</code><br />
|}<br />
<br />
== Dépôt GitHub ==<br />
* [https://github.com/ynotopec/api-realtime-ai ynotopec/api-realtime-ai]<br />
<br />
== Page de test ==<br />
* <code>external-test/half-duplex.html</code> — annulation d’écho + mode half-duplex.<br />
<br />
== Compatibilité ==<br />
Remplacer l’URL OpenAI par <code>$OPENAI_API_BASE</code> pour tester compatibilité et performances.<br />
<br />
----<br />
<br />
= API LLM (OpenAI compatible) =<br />
<br />
* URL de base : <code>https://api.ai.lab.infocepo.com/v1</code><br />
* Création du token : [https://llm-token.ai.lab.infocepo.com OPENAI_API_KEY]<br />
* Documentation : [https://api.ai.lab.infocepo.com Documentation API]<br />
<br />
== Liste des modèles ==<br />
<pre><br />
curl -X GET \<br />
'https://api.ai.lab.infocepo.com/v1/models' \<br />
-H 'Authorization: Bearer sk-XXXXX' \<br />
-H 'accept: application/json' \<br />
| jq | sed -rn 's#^.*id.*: "(.*)".*$#* \1#p' | sort -u<br />
</pre><br />
<br />
== Modèles mis en avant ==<br />
{| class="wikitable"<br />
! Model !! Commentaire<br />
|-<br />
| '''ai-chat''' || qwen3-coder<br />
|-<br />
| '''ai-translate''' || qwen3-coder<br />
|-<br />
| '''ai-summary''' || qwen3-coder<br />
|-<br />
| '''ai-code-completion''' || qwen3-coder<br />
|-<br />
| '''ai-RAG-FR''' || qwen3-coder<br />
|-<br />
| '''qwen3-coder''' || Function Calling<br />
|-<br />
| '''ai-ocr''' || qwen3-vl<br />
|}<br />
<br />
== Exemple bash ==<br />
<pre><br />
export OPENAI_API_MODEL="ai-chat"<br />
export OPENAI_API_BASE="https://api.ai.lab.infocepo.com/v1"<br />
export OPENAI_API_KEY="sk-XXXXX"<br />
<br />
promptValue="Quel est ton nom ?"<br />
jsonValue='{<br />
"model": "'${OPENAI_API_MODEL}'",<br />
"messages": [{"role": "user", "content": "'${promptValue}'"}],<br />
"temperature": 0<br />
}'<br />
<br />
curl -k ${OPENAI_API_BASE}/chat/completions \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d "${jsonValue}" 2>/dev/null | jq '.choices[0].message.content'<br />
</pre><br />
<br />
== Vue infra LLM ==<br />
[[File:Litellm-proxy-mermaid-diagram-2024-03-24-205202.png|thumb|right]]<br />
<br />
'''DEV (au choix)'''<br />
* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité<br />
* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer<br />
* '''C.''' <code>Ollama</code> direct : POC ultra-léger<br />
<br />
'''DEV – modèle FR / résumé'''<br />
* <code>LiteLLM → Ollama /v1</code><br />
<br />
'''PROD'''<br />
* '''Standard :''' <code>LiteLLM → vLLM</code><br />
* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code><br />
<br />
'''Notes :'''<br />
* '''LiteLLM''' = passerelle unique (clés, quotas, logs)<br />
* '''vLLM''' = performance / stabilité en charge<br />
* '''Ollama''' = simplicité de prototypage<br />
<br />
----<br />
<br />
= API Image to Text =<br />
<br />
* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.<br />
* Modèle recommandé : <code>ai-ocr</code><br />
<br />
== Exemple bash ==<br />
<pre><br />
OPENAI_API_KEY=sk-XXXXX<br />
<br />
base64 -w0 "/path/to/image.png" > img.b64<br />
<br />
jq -n --rawfile img img.b64 \<br />
'{<br />
model: "ai-ocr",<br />
messages: [<br />
{<br />
role: "user",<br />
content: [<br />
{ "type": "text", "text": "Décris cette image." },<br />
{<br />
"type": "image_url",<br />
"image_url": { "url": ("data:image/png;base64," + ($img | rtrimstr("\n"))) }<br />
}<br />
]<br />
}<br />
]<br />
}' > payload.json<br />
<br />
curl https://api.ai.lab.infocepo.com/v1/chat/completions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
--data-binary @payload.json<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import base64<br />
import json<br />
import requests<br />
import os<br />
<br />
API_KEY = os.getenv("OPENAI_API_KEY")<br />
MODEL = "ai-ocr"<br />
IMG_PATH = "/path/to/image.png"<br />
API_URL = "https://api.ai.lab.infocepo.com/v1/chat/completions"<br />
<br />
with open(IMG_PATH, "rb") as f:<br />
img_b64 = base64.b64encode(f.read()).decode("utf-8")<br />
<br />
payload = {<br />
"model": MODEL,<br />
"messages": [<br />
{<br />
"role": "user",<br />
"content": [<br />
{"type": "text", "text": "Décris cette image."},<br />
{<br />
"type": "image_url",<br />
"image_url": {"url": f"data:image/png;base64,{img_b64}"}<br />
}<br />
]<br />
}<br />
]<br />
}<br />
<br />
headers = {<br />
"Authorization": f"Bearer {API_KEY}",<br />
"Content-Type": "application/json"<br />
}<br />
<br />
response = requests.post(API_URL, headers=headers, data=json.dumps(payload))<br />
<br />
if response.ok:<br />
print(json.dumps(response.json(), indent=2, ensure_ascii=False))<br />
else:<br />
print(f"Erreur {response.status_code}: {response.text}")<br />
</pre><br />
<br />
----<br />
<br />
= API STT =<br />
<br />
* URL : <code>https://stt.ai.lab.infocepo.com/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Modèle : <code>whisper-1</code><br />
* Documentation : [https://stt.ai.lab.infocepo.com/docs API STT docs]<br />
<br />
== Exemple Python ==<br />
<pre><br />
import requests<br />
<br />
OPENAI_API_KEY = 'sk-XXXXX'<br />
<br />
url = 'https://stt.ai.lab.infocepo.com/v1/audio/transcriptions'<br />
headers = {<br />
'Authorization': f'Bearer {OPENAI_API_KEY}',<br />
}<br />
files = {<br />
'file': ('file.opus', open('/path/to/file.opus', 'rb')),<br />
'model': (None, 'whisper-1')<br />
}<br />
<br />
response = requests.post(url, headers=headers, files=files)<br />
print(response.json())<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
[ ! -f /tmp/test.ogg ] && wget "https://upload.wikimedia.org/wikipedia/commons/1/17/Fables_de_La_Fontaine_Livre_1_01.ogg" -O /tmp/test.ogg<br />
<br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://stt.ai.lab.infocepo.com/v1/audio/transcriptions \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-F model="whisper-1" \<br />
-F file="@/tmp/test.ogg"<br />
</pre><br />
<br />
== Notes ==<br />
* Plusieurs formats audio sont acceptés.<br />
* Le flux final est normalisé en '''16 kHz mono'''.<br />
* Pour une qualité optimale : privilégier '''OPUS 16 kHz mono'''.<br />
<br />
== UI ==<br />
* [https://translate-rt.ai.lab.infocepo.com translate-rt]<br />
<br />
----<br />
<br />
= API TTS =<br />
<br />
* URL : <code>https://tts.ai.lab.infocepo.com/v1</code><br />
* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code><br />
* Documentation : [https://tts.ai.lab.infocepo.com/docs API TTS docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=sk-XXXXX<br />
<br />
curl https://tts.ai.lab.infocepo.com/v1/audio/speech \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-H "Content-Type: application/json" \<br />
-d '{<br />
"model": "gpt-4o-mini-tts",<br />
"input": "Bonjour, ceci est un test de synthèse vocale.",<br />
"voice": "coral",<br />
"instructions": "Speak in a cheerful and positive tone.",<br />
"response_format": "opus"<br />
}' | ffplay -i -<br />
</pre><br />
<br />
----<br />
<br />
= API Text to Image =<br />
<br />
* URL : <code>https://api-txt2image.ai.lab.infocepo.com/v1</code><br />
* Clé API : <code>OPENAI_API_KEY=EMPTY</code><br />
* Documentation : [https://api-txt2image.ai.lab.infocepo.com/docs API TXT2IMAGE docs]<br />
<br />
== Exemple ==<br />
<pre><br />
export OPENAI_API_KEY=EMPTY<br />
<br />
curl https://api-txt2image.ai.lab.infocepo.com/v1/images/generations \<br />
-H "Content-Type: application/json" \<br />
-H "Authorization: Bearer $OPENAI_API_KEY" \<br />
-d '{<br />
"prompt": "a photo of a happy corgi puppy sitting and facing forward, studio light, longshot",<br />
"n": 1,<br />
"size": "1024x1024"<br />
}'<br />
</pre><br />
<br />
----<br />
<br />
= API Diarization =<br />
<br />
* Documentation : [https://api-diarization.ai.lab.infocepo.com/docs API Diarization docs]<br />
<br />
== Exemple ==<br />
<pre><br />
wget "https://upload.wikimedia.org/wikipedia/commons/6/60/Mike_Peters_on_Politics_and_Emotion_%28Interview_1984%29.mp3" -O /tmp/test.mp3<br />
<br />
curl -X POST "https://api-diarization.ai.lab.infocepo.com/upload-audio/" \<br />
-H "Authorization: Bearer token1" \<br />
-F "file=@/tmp/test.mp3"<br />
</pre><br />
<br />
----<br />
<br />
= API Summary =<br />
<br />
* Documentation : [https://api-summary.ai.lab.infocepo.com/docs API Summary docs]<br />
<br />
== Exemple ==<br />
<pre><br />
text="The tower is 324 metres tall and is one of the most recognizable monuments in the world."<br />
<br />
json_payload=$(jq -nc --arg text "$text" '{"text": $text}')<br />
<br />
curl -X POST https://api-summary.ai.lab.infocepo.com/summary/ \<br />
-H "Content-Type: application/json" \<br />
-d "$json_payload"<br />
</pre><br />
<br />
----<br />
<br />
= API Text Embeddings =<br />
<br />
* URL : <code>https://text-embeddings.ai.lab.infocepo.com</code><br />
* URL utilisée par MRSIE : <code>https://tei.ai.lab.infocepo.com/</code><br />
* Documentation : [https://text-embeddings.ai.lab.infocepo.com/docs Documentation]<br />
<br />
== Exemple ==<br />
<pre><br />
curl -k https://text-embeddings.ai.lab.infocepo.com/embed \<br />
-X POST \<br />
-d '{"inputs":"What is Deep Learning?"}' \<br />
-H 'Content-Type: application/json'<br />
</pre><br />
<br />
----<br />
<br />
= API DB Vectors (ChromaDB) =<br />
<br />
== Production ==<br />
* URL : <code>https://chromadb.ai.lab.infocepo.com</code><br />
* Token : <code>XXXXX</code><br />
<br />
== Lab ==<br />
<pre><br />
export CHROMA_HOST=https://chromadb.c1.ai.lab.infocepo.com<br />
export CHROMA_PORT=443<br />
export CHROMA_TOKEN=XXXX<br />
</pre><br />
<br />
== Exemple curl ==<br />
<pre><br />
curl -v "${CHROMA_HOST}"/api/v1/collections \<br />
-H "Authorization: Bearer ${CHROMA_TOKEN}"<br />
</pre><br />
<br />
== Exemple Python ==<br />
<pre><br />
import chromadb<br />
from chromadb.config import Settings<br />
<br />
def chroma_http(host, port=80, token=None):<br />
return chromadb.HttpClient(<br />
host=host,<br />
port=port,<br />
ssl=host.startswith('https') or port == 443,<br />
settings=(<br />
Settings(<br />
chroma_client_auth_provider='chromadb.auth.token.TokenAuthClientProvider',<br />
chroma_client_auth_credentials=token,<br />
) if token else Settings()<br />
)<br />
)<br />
<br />
client = chroma_http(CHROMA_HOST, CHROMA_PORT, CHROMA_TOKEN)<br />
collections = client.list_collections()<br />
print(collections)<br />
</pre><br />
<br />
== Déployer sa propre instance ==<br />
<pre><br />
export nameSpace=your_namespace<br />
domainRoot=ai.lab.infocepo.com<br />
<br />
helm repo add chroma https://amikos-tech.github.io/chromadb-chart/<br />
helm repo update<br />
<br />
helm upgrade --install chromadb chroma/chromadb -n ${nameSpace} \<br />
--set chromadb.apiVersion="0.4.24" \<br />
--set ingress.enabled=true \<br />
--set ingress.hosts[0].host="${nameSpace}-chromadb.${domainRoot}" \<br />
--set ingress.hosts[0].paths[0].path=/ \<br />
--set ingress.hosts[0].paths[0].pathType=ImplementationSpecific \<br />
--set ingress.annotations."cert-manager\.io/cluster-issuer"=letsencrypt-prod \<br />
--set ingress.tls[0].secretName=${nameSpace}-chromadb.${domainRoot}-tls \<br />
--set ingress.tls[0].hosts[0]="${nameSpace}-chromadb.${domainRoot}"<br />
<br />
kubectl -n ${nameSpace} patch ingress/chromadb --type=json \<br />
-p '[{"op":"add","path":"/metadata/annotations/nginx.ingress.kubernetes.io~1proxy-body-size","value":"0"}]'<br />
</pre><br />
<br />
== Récupérer le token ==<br />
<pre><br />
kubectl --namespace ${nameSpace} get secret chromadb-auth \<br />
-o jsonpath="{.data.token}" | base64 --decode && echo<br />
</pre><br />
<br />
----<br />
<br />
= Registry =<br />
<br />
* URL : [https://registry.ai.lab.infocepo.com registry.ai.lab.infocepo.com]<br />
* Login : <code>user</code><br />
* Password : <code>XXXXX</code><br />
<br />
== Exemple ==<br />
<pre><br />
curl -u "user:XXXXX" https://registry.ai.lab.infocepo.com/v2/_catalog<br />
</pre><br />
<br />
== Exemple K8S ==<br />
<pre><br />
deploymentName=<br />
nameSpace=<br />
<br />
kubectl -n ${nameSpace} create secret docker-registry pull-secret \<br />
--docker-server=registry.ai.lab.infocepo.com \<br />
--docker-username=user \<br />
--docker-password=XXXXX \<br />
--docker-email=contact@example.com<br />
<br />
kubectl -n ${nameSpace} patch deployment ${deploymentName} \<br />
-p '{"spec":{"template":{"spec":{"imagePullSecrets":[{"name":"pull-secret"}]}}}}'<br />
</pre><br />
<br />
----<br />
<br />
= Stockage objet externe (S3) =<br />
<br />
* Endpoint : <code>https://s3.ai.lab.infocepo.com</code><br />
* Access key : <code>XXXX</code><br />
* Secret key : <code>XXXX</code><br />
<br />
Un bucket nommé <code>ORG</code> a été créé pour stocker des documents de démonstration.<br />
<br />
----<br />
<br />
= RAG optimisation =<br />
<br />
* Embeddings : <code>BAAI/bge-m3</code><br />
* <code>chunk_size=1200</code><br />
* <code>chunk_overlap=100</code><br />
* LLM : <code>qwen3</code><br />
* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.<br />
<br />
----<br />
<br />
= Processus usine IA =<br />
<br />
{| class="wikitable" style="width:80%;"<br />
! Étape !! Description !! Outils utilisés !! Responsable(s)<br />
|-<br />
| 1 || Idée || - || Équipe projet<br />
|-<br />
| 2 || Développement || Environnement Onyxia / lab || Équipe projet<br />
|-<br />
| 3 || Déploiement || CI/CD, GitHub, Kubernetes || Équipe DevOps<br />
|-<br />
| 4 || Surveillance || Uptime-Kuma, dashboards || Équipe DevOps<br />
|-<br />
| 5 || Alertes || Mattermost || Équipe DevOps<br />
|-<br />
| 6 || Support infrastructure || - || Équipe SRE<br />
|-<br />
| 7 || Support applicatif || - || Équipe applicative<br />
|}<br />
<br />
----<br />
<br />
= Environnements =<br />
<br />
== Hors production ==<br />
* Utiliser [https://datalab.ai.lab.infocepo.com datalab]<br />
* Support : canal Mattermost Offre IA<br />
* Le pseudo utilisateur doit respecter la convention interne<br />
* Demander si besoin un accès Linux + Kubernetes<br />
<br />
== Production (best-effort) ==<br />
* Publier le code applicatif, les secrets (format SOPS), le Dockerfile et le code infra (Helm ou manifests K8S) sur Git<br />
* Demander un namespace<br />
* Lire la documentation de surveillance associée<br />
<br />
== Limites de l’infrastructure ==<br />
* Les charges GPU sont intentionnellement limitées en journée.<br />
<br />
----<br />
<br />
= Cloud Lab & projets d’audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
Le '''Cloud Lab''' fournit des scénarios reproductibles : audit d’infrastructure, migration cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit ==<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit permettant de :<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer un plan de migration ou de remédiation.<br />
<br />
== Exemple de migration cloud ==<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Diagramme d’architecture || Conception visuelle et documentation || 1.5<br />
|-<br />
| Contrôles de conformité || 2 clouds, 6 hyperviseurs, 6 To RAM || 1.5<br />
|-<br />
| Installation plateforme cloud || Déploiement des environnements cibles || 1.0<br />
|-<br />
| Vérification de stabilité || Premiers tests fonctionnels || 0.5<br />
|-<br />
| Étude d’automatisation || Identification des tâches répétitives || 1.5<br />
|-<br />
| Développement des templates || 6 templates, 8 environnements, 2 clouds / OS || 1.5<br />
|-<br />
| Diagramme de migration || Illustration du processus || 1.0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Stabilisation || Validation de la reproductibilité || 1.5<br />
|-<br />
| Benchmark cloud || Comparaison vs legacy || 1.5<br />
|-<br />
| Réglage des temps d’arrêt || Calcul du downtime || 0.5<br />
|-<br />
| Chargement VM || 82 VMs : OS, code, 2 IP par VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours.homme<br />
|}<br />
<br />
=== Vérifications de stabilité (HA minimale) ===<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services redémarrent automatiquement sur les autres nœuds<br />
|-<br />
| Extinction / redémarrage simultané de tous les nœuds || Les services repartent correctement après reboot<br />
|}<br />
<br />
----<br />
<br />
= Architecture web & bonnes pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principes de conception :<br />
<br />
* privilégier une infrastructure '''simple, modulaire et flexible''',<br />
* rapprocher le contenu du client (GDNS ou équivalent),<br />
* utiliser des load balancers réseau (LVS, IPVS),<br />
* comparer les coûts et éviter le '''vendor lock-in''',<br />
* pour TLS :<br />
** '''HAProxy''' pour les frontends rapides,<br />
** '''Envoy''' pour les cas avancés (mTLS, HTTP/2/3),<br />
* pour le cache :<br />
** '''Varnish''', '''Apache Traffic Server''',<br />
* favoriser les stacks open-source,<br />
* utiliser files, buffers, queues et quotas pour lisser les pics.<br />
<br />
== Références ==<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia infrastructure]<br />
* [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes cloud =<br />
<br />
{| class="wikitable"<br />
! Fonctionnalité !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle routeur''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle firewall''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Stockage''' || Local, cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Cette table sert de point de départ pour choisir la bonne stack selon :<br />
* le niveau de contrôle souhaité,<br />
* le contexte (on-prem, cloud public, HPC…),<br />
* les outils d’automatisation existants.<br />
<br />
----<br />
<br />
= Haute disponibilité, HPC & DevSecOps =<br />
<br />
== Haute disponibilité avec Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Principes :<br />
* clusters multi-nœuds ou multi-sites,<br />
* fencing via IPMI,<br />
* provisioning PXE / NTP / DNS / TFTP,<br />
* pour 2 nœuds : attention au split-brain,<br />
* 3 nœuds ou plus recommandés en production.<br />
<br />
=== Ressources fréquentes ===<br />
* multipath, LUNs, LVM, NFS,<br />
* processus applicatifs,<br />
* IP virtuelles, DNS, listeners réseau.<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* orchestration de jobs (SLURM ou équivalent),<br />
* stockage partagé haute performance,<br />
* intégration possible avec des workloads IA.<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD avec contrôles de sécurité intégrés,<br />
* observabilité dès la conception,<br />
* scans de vulnérabilité,<br />
* gestion des secrets,<br />
* policy-as-code.<br />
<br />
----<br />
<br />
= News & trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News]<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper]<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator]<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM]<br />
<br />
----<br />
<br />
= Formation & apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained]<br />
* Labs, scripts et retours d’expérience concrets dans le projet Cloud Lab<br />
<br />
----<br />
<br />
= Liens cloud & IT utiles =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM]<br />
* [https://access.redhat.com/downloads/content/package-browser Red Hat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM IT]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm Indicateurs salariaux IT]<br />
<br />
----<br />
<br />
= Outils collaboratifs =<br />
<br />
== Dépôts de code ==<br />
* [https://github.com/ynotopec GitHub ynotopec]<br />
<br />
== Base de connaissance ==<br />
* ce wiki<br />
<br />
== Messagerie ==<br />
* contact interne / support selon les projets<br />
<br />
== SSO ==<br />
* [https://auth-lab.ai.lab.infocepo.com/auth Keycloak]<br />
<br />
== MLflow ==<br />
* [[MLFlow|MLFlow]]<br />
<br />
----<br />
<br />
= À propos & contributions =<br />
<br />
Suggestions de corrections, améliorations de schémas, retours d’expérience ou nouveaux labs bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1928
Main Page
2026-03-26T16:09:44Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Cloud, AI and Labs on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Bienvenue sur '''infocepo.com'''.<br />
<br />
Ce wiki centralise des ressources sur :<br />
<br />
* l’'''infrastructure cloud''' et les architectures distribuées,<br />
* l’'''IA appliquée''' (assistants privés, APIs, RAG, GPU),<br />
* les '''labs techniques''' pour apprendre, tester et industrialiser,<br />
* les '''scripts et méthodes''' pour l’audit, la migration et l’automatisation.<br />
<br />
L’objectif est de transformer des idées et des concepts en '''solutions concrètes, reproductibles et utiles'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Accès rapide =<br />
<br />
* [https://chat.infocepo.com '''Assistant IA''']<br />
* [https://infocepo.com '''Portail principal''']<br />
* [[Special:AllPages|'''Toutes les pages''']]<br />
* [https://github.com/ynotopec '''GitHub'''] <br />
* [https://uptime-kuma.ai.lab.infocepo.com/status/ai '''Statut des services''']]<br />
* [https://grafana.ai.lab.infocepo.com '''Monitoring''']]<br />
<br />
----<br />
<br />
= Parcours recommandés =<br />
<br />
== Construire un assistant IA privé ==<br />
* Déployer une stack type '''Open WebUI + Ollama + GPU'''<br />
* Ajouter des modèles de chat, résumé, OCR ou transcription<br />
* Connecter des documents via '''RAG + embeddings'''<br />
<br />
== Lancer un lab cloud ==<br />
* Créer un environnement Kubernetes, OpenStack ou bare-metal<br />
* Déployer avec Helm, Terraform ou Ansible<br />
* Ajouter des services IA et des outils d’observabilité<br />
<br />
== Préparer un audit ou une migration ==<br />
* Inventorier les environnements<br />
* Concevoir l’architecture cible<br />
* Automatiser les opérations avec des scripts reproductibles<br />
<br />
----<br />
<br />
= Sections principales =<br />
<br />
== IA & APIs ==<br />
Services autour des assistants, modèles, OCR, transcription, synthèse vocale, résumé, embeddings et bases vecteur.<br />
<br />
== Cloud & Infrastructure ==<br />
Kubernetes, haute disponibilité, architecture web, HPC, DevSecOps et bonnes pratiques d’exploitation.<br />
<br />
== Labs & Automatisation ==<br />
Scénarios techniques, scripts réutilisables, audits, migrations et démonstrateurs.<br />
<br />
== Comparatifs & Références ==<br />
Tableaux de comparaison, benchmarks, liens utiles et veille technique.<br />
<br />
----<br />
<br />
= Services mis en avant =<br />
<br />
{| class="wikitable"<br />
! Service !! Rôle<br />
|-<br />
| [https://api.ai.lab.infocepo.com '''API LLM'''] || Chat, code, RAG, OCR<br />
|-<br />
| [https://stt.ai.lab.infocepo.com/docs '''API STT'''] || Transcription audio<br />
|-<br />
| [https://tts.ai.lab.infocepo.com/docs '''API TTS'''] || Synthèse vocale<br />
|-<br />
| [https://api-summary.ai.lab.infocepo.com/docs '''API Summary'''] || Résumé de textes<br />
|-<br />
| [https://text-embeddings.ai.lab.infocepo.com/docs '''Text Embeddings'''] || Embeddings pour RAG<br />
|-<br />
| [https://chromadb.ai.lab.infocepo.com '''ChromaDB'''] || Base de données vecteur<br />
|-<br />
| [https://datalab.ai.lab.infocepo.com '''DataLab'''] || Environnement de travail et d’expérimentation<br />
|}<br />
<br />
----<br />
<br />
= Projets et usages =<br />
<br />
Ce wiki sert notamment à :<br />
<br />
* documenter des stacks IA privées,<br />
* publier des schémas d’architecture,<br />
* capitaliser des retours d’expérience,<br />
* préparer des déploiements reproductibles,<br />
* industrialiser des workflows cloud et IA.<br />
<br />
----<br />
<br />
= Aller plus loin =<br />
<br />
* [[ServerDiff.sh|'''Audit avec ServerDiff.sh''']]<br />
* [[MLFlow|'''MLFlow''']]<br />
* [[CI-CD-GITHUB-K8S|'''CI/CD + GitHub + Kubernetes''']]<br />
<br />
Pour plus de détails, parcourir [[Special:AllPages|'''la liste complète des pages''']].<br />
<br />
----<br />
<br />
= Contributions =<br />
<br />
Les suggestions, corrections, améliorations de schémas et nouveaux labs sont les bienvenus.<br />
<br />
Ce wiki a vocation à rester un '''laboratoire vivant''' autour du cloud, de l’IA et de l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png&diff=1927
File:SUMMARY-DIAGRAM-7311e6b1-aede-4989-ade2-a42d1a6e0ff2.png
2026-03-26T15:07:51Z
<p>Tcepo: </p>
<hr />
<div></div>
Tcepo
https://infocepo.com/wiki/index.php?title=Enigme&diff=1926
Enigme
2026-03-21T11:14:01Z
<p>Tcepo: </p>
<hr />
<div>= '''1. Quel est l'âge de Paul ? =<br />
<br />
Paul s'adresse à Pierre :<br />
<br />
"J'ai le double de l'âge que tu avais quand j'avais ton âge.<br />
Quand tu auras mon âge, ensemble nous aurons 9 ans."<br />
<br />
Quel âge a Paul ?<br />
Quel âge a Pierre ? (résultat sans explications)<br />
<br />
* English :<br />
Paul addresses Peter:<br />
<br />
"I am twice as old as you were when I was your age. When you are my age, together we will be 72 years old."<br />
<br />
How old is Paul? How old is Peter?<br />
<br />
<br />
{| class="wikitable" style="text-align:center;"<br />
|+ Valeurs de X, Paul (P) et Pierre (R)<br />
! X !! Paul (P) !! Pierre (R)<br />
|-<br />
| 9 || 4 || 3<br />
|-<br />
| 18 || 8 || 6<br />
|-<br />
| 27 || 12 || 9<br />
|-<br />
| 36 || 16 || 12<br />
|-<br />
| 45 || 20 || 15<br />
|-<br />
| 54 || 24 || 18<br />
|-<br />
| 63 || 28 || 21<br />
|-<br />
| 72 || 32 || 24<br />
|-<br />
| 81 || 36 || 27<br />
|-<br />
| 90 || 40 || 30<br />
|-<br />
| 99 || 44 || 33<br />
|-<br />
| 108 || 48 || 36<br />
|-<br />
| 117 || 52 || 39<br />
|-<br />
| 126 || 56 || 42<br />
|-<br />
| 135 || 60 || 45<br />
|-<br />
| 144 || 64 || 48<br />
|-<br />
| 153 || 68 || 51<br />
|-<br />
| 162 || 72 || 54<br />
|-<br />
| 171 || 76 || 57<br />
|-<br />
| 180 || 80 || 60<br />
|-<br />
| 189 || 84 || 63<br />
|-<br />
| 198 || 88 || 66<br />
|-<br />
| 207 || 92 || 69<br />
|-<br />
| 216 || 96 || 72<br />
|-<br />
| 225 || 100 || 75<br />
|-<br />
| 234 || 104 || 78<br />
|-<br />
| 243 || 108 || 81<br />
|-<br />
| 252 || 112 || 84<br />
|-<br />
| 261 || 116 || 87<br />
|-<br />
| 270 || 120 || 90<br />
|-<br />
| 279 || 124 || 93<br />
|-<br />
| 288 || 128 || 96<br />
|-<br />
| 297 || 132 || 99<br />
|}<br />
<br />
<br />
<br />
'''2. Un lac étonnant!'''<br />
<br />
Un lac double de volume chaque jour.<br />
Au bout de 30 jours il est plein.<br />
<br />
Le lac est donc plein à la moitié au bout de combien de jours ?<br />
<br />
'''3. Les 3 prisionniers<br />
<br />
Trois prisonniers sont ensemble. Chacun portant une pancarte au dos tiré au hasard parmi 3 pancartes blanches et 2 noirs. <br />
<br />
Chaque prisonnier ne peut pas voir sa propre pancarte, mais peut voir celles des autres.<br />
<br />
Dans cette situation, les trois prisonniers reçoivent chacun une pancarte blanche.<br />
<br />
Ceux qui devinent la couleur de leur propre pancarte annoncent à tout le monde qu'ils ont trouvé et sont libérés.<br />
<br />
Comment est-il possible de deviner la couleur de sa propre pancarte dans cette situation ?<br />
<br />
* English :<br />
Three prisoners are together. Each of them wears a sign on their back, randomly drawn from a set of 3 white signs and 2 black signs.<br />
<br />
Each prisoner cannot see the color of their own sign but can see the signs of the others.<br />
<br />
In this situation, all three prisoners receive a white sign.<br />
<br />
The first prisoner who correctly guesses the color of their own sign raises their hand and is set free.<br />
<br />
How is it possible to determine the color of their own sign in this scenario?</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Enigme&diff=1925
Enigme
2026-03-21T11:13:50Z
<p>Tcepo: </p>
<hr />
<div>= '''1. Quel est l'âge de Paul ? =<br />
<br />
Paul s'adresse à Pierre :<br />
<br />
"J'ai le double de l'âge que tu avais quand j'avais ton âge.<br />
Quand tu auras mon âge, ensemble nous aurons 9 ans."<br />
<br />
Quel âge a Paul ?<br />
Quel âge a Pierre ? (résultats sans explications)<br />
<br />
* English :<br />
Paul addresses Peter:<br />
<br />
"I am twice as old as you were when I was your age. When you are my age, together we will be 72 years old."<br />
<br />
How old is Paul? How old is Peter?<br />
<br />
<br />
{| class="wikitable" style="text-align:center;"<br />
|+ Valeurs de X, Paul (P) et Pierre (R)<br />
! X !! Paul (P) !! Pierre (R)<br />
|-<br />
| 9 || 4 || 3<br />
|-<br />
| 18 || 8 || 6<br />
|-<br />
| 27 || 12 || 9<br />
|-<br />
| 36 || 16 || 12<br />
|-<br />
| 45 || 20 || 15<br />
|-<br />
| 54 || 24 || 18<br />
|-<br />
| 63 || 28 || 21<br />
|-<br />
| 72 || 32 || 24<br />
|-<br />
| 81 || 36 || 27<br />
|-<br />
| 90 || 40 || 30<br />
|-<br />
| 99 || 44 || 33<br />
|-<br />
| 108 || 48 || 36<br />
|-<br />
| 117 || 52 || 39<br />
|-<br />
| 126 || 56 || 42<br />
|-<br />
| 135 || 60 || 45<br />
|-<br />
| 144 || 64 || 48<br />
|-<br />
| 153 || 68 || 51<br />
|-<br />
| 162 || 72 || 54<br />
|-<br />
| 171 || 76 || 57<br />
|-<br />
| 180 || 80 || 60<br />
|-<br />
| 189 || 84 || 63<br />
|-<br />
| 198 || 88 || 66<br />
|-<br />
| 207 || 92 || 69<br />
|-<br />
| 216 || 96 || 72<br />
|-<br />
| 225 || 100 || 75<br />
|-<br />
| 234 || 104 || 78<br />
|-<br />
| 243 || 108 || 81<br />
|-<br />
| 252 || 112 || 84<br />
|-<br />
| 261 || 116 || 87<br />
|-<br />
| 270 || 120 || 90<br />
|-<br />
| 279 || 124 || 93<br />
|-<br />
| 288 || 128 || 96<br />
|-<br />
| 297 || 132 || 99<br />
|}<br />
<br />
<br />
<br />
'''2. Un lac étonnant!'''<br />
<br />
Un lac double de volume chaque jour.<br />
Au bout de 30 jours il est plein.<br />
<br />
Le lac est donc plein à la moitié au bout de combien de jours ?<br />
<br />
'''3. Les 3 prisionniers<br />
<br />
Trois prisonniers sont ensemble. Chacun portant une pancarte au dos tiré au hasard parmi 3 pancartes blanches et 2 noirs. <br />
<br />
Chaque prisonnier ne peut pas voir sa propre pancarte, mais peut voir celles des autres.<br />
<br />
Dans cette situation, les trois prisonniers reçoivent chacun une pancarte blanche.<br />
<br />
Ceux qui devinent la couleur de leur propre pancarte annoncent à tout le monde qu'ils ont trouvé et sont libérés.<br />
<br />
Comment est-il possible de deviner la couleur de sa propre pancarte dans cette situation ?<br />
<br />
* English :<br />
Three prisoners are together. Each of them wears a sign on their back, randomly drawn from a set of 3 white signs and 2 black signs.<br />
<br />
Each prisoner cannot see the color of their own sign but can see the signs of the others.<br />
<br />
In this situation, all three prisoners receive a white sign.<br />
<br />
The first prisoner who correctly guesses the color of their own sign raises their hand and is set free.<br />
<br />
How is it possible to determine the color of their own sign in this scenario?</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Enigme&diff=1924
Enigme
2026-03-21T11:13:26Z
<p>Tcepo: </p>
<hr />
<div>= '''1. Quel est l'âge de Paul ? =<br />
<br />
Paul s'adresse à Pierre :<br />
<br />
"J'ai le double de l'âge que tu avais quand j'avais ton âge.<br />
Quand tu auras mon âge, ensemble nous aurons 9 ans."<br />
<br />
Quel âge a Paul ?<br />
Quel âge a Pierre ? (sans explications)<br />
<br />
* English :<br />
Paul addresses Peter:<br />
<br />
"I am twice as old as you were when I was your age. When you are my age, together we will be 72 years old."<br />
<br />
How old is Paul? How old is Peter?<br />
<br />
<br />
{| class="wikitable" style="text-align:center;"<br />
|+ Valeurs de X, Paul (P) et Pierre (R)<br />
! X !! Paul (P) !! Pierre (R)<br />
|-<br />
| 9 || 4 || 3<br />
|-<br />
| 18 || 8 || 6<br />
|-<br />
| 27 || 12 || 9<br />
|-<br />
| 36 || 16 || 12<br />
|-<br />
| 45 || 20 || 15<br />
|-<br />
| 54 || 24 || 18<br />
|-<br />
| 63 || 28 || 21<br />
|-<br />
| 72 || 32 || 24<br />
|-<br />
| 81 || 36 || 27<br />
|-<br />
| 90 || 40 || 30<br />
|-<br />
| 99 || 44 || 33<br />
|-<br />
| 108 || 48 || 36<br />
|-<br />
| 117 || 52 || 39<br />
|-<br />
| 126 || 56 || 42<br />
|-<br />
| 135 || 60 || 45<br />
|-<br />
| 144 || 64 || 48<br />
|-<br />
| 153 || 68 || 51<br />
|-<br />
| 162 || 72 || 54<br />
|-<br />
| 171 || 76 || 57<br />
|-<br />
| 180 || 80 || 60<br />
|-<br />
| 189 || 84 || 63<br />
|-<br />
| 198 || 88 || 66<br />
|-<br />
| 207 || 92 || 69<br />
|-<br />
| 216 || 96 || 72<br />
|-<br />
| 225 || 100 || 75<br />
|-<br />
| 234 || 104 || 78<br />
|-<br />
| 243 || 108 || 81<br />
|-<br />
| 252 || 112 || 84<br />
|-<br />
| 261 || 116 || 87<br />
|-<br />
| 270 || 120 || 90<br />
|-<br />
| 279 || 124 || 93<br />
|-<br />
| 288 || 128 || 96<br />
|-<br />
| 297 || 132 || 99<br />
|}<br />
<br />
<br />
<br />
'''2. Un lac étonnant!'''<br />
<br />
Un lac double de volume chaque jour.<br />
Au bout de 30 jours il est plein.<br />
<br />
Le lac est donc plein à la moitié au bout de combien de jours ?<br />
<br />
'''3. Les 3 prisionniers<br />
<br />
Trois prisonniers sont ensemble. Chacun portant une pancarte au dos tiré au hasard parmi 3 pancartes blanches et 2 noirs. <br />
<br />
Chaque prisonnier ne peut pas voir sa propre pancarte, mais peut voir celles des autres.<br />
<br />
Dans cette situation, les trois prisonniers reçoivent chacun une pancarte blanche.<br />
<br />
Ceux qui devinent la couleur de leur propre pancarte annoncent à tout le monde qu'ils ont trouvé et sont libérés.<br />
<br />
Comment est-il possible de deviner la couleur de sa propre pancarte dans cette situation ?<br />
<br />
* English :<br />
Three prisoners are together. Each of them wears a sign on their back, randomly drawn from a set of 3 white signs and 2 black signs.<br />
<br />
Each prisoner cannot see the color of their own sign but can see the signs of the others.<br />
<br />
In this situation, all three prisoners receive a white sign.<br />
<br />
The first prisoner who correctly guesses the color of their own sign raises their hand and is set free.<br />
<br />
How is it possible to determine the color of their own sign in this scenario?</div>
Tcepo
https://infocepo.com/wiki/index.php?title=CV_PACHECO_BG&diff=1923
CV PACHECO BG
2026-02-17T21:51:29Z
<p>Tcepo: Created page with "= Ръководител на суверенна AI платформа = ''LLM инфраструктура в национален мащаб • Сигурни AI платформи • GPU стратегия • Суверенитет чрез отворен код'' == Профил == Ръководител на суверенна AI платформа, специализиран в проектиране и управление на национални сигур..."</p>
<hr />
<div>= Ръководител на суверенна AI платформа =<br />
''LLM инфраструктура в национален мащаб • Сигурни AI платформи • GPU стратегия • Суверенитет чрез отворен код''<br />
<br />
== Профил ==<br />
Ръководител на суверенна AI платформа, специализиран в проектиране и управление на национални сигурни платформи за генеративен изкуствен интелект.<br />
<br />
Експерт в архитектури с LLM с отворен код, суверенно внедряване на GPU инфраструктура, API съвместими с OpenAI, индустриализирани RAG пайплайни и техническо управление в чувствителна среда.<br />
<br />
Способност за трансформиране на AI експерименти в стабилни, сигурни и мащабируеми платформи, стратегически съобразени с изискванията за технологичен суверенитет.<br />
<br />
== Стратегически обхват ==<br />
<br />
=== Национална AI инфраструктурна архитектура ===<br />
* Проектиране на сигурни вътрешни LLM платформи<br />
* Kubernetes архитектура с множество възли и специализирани GPU<br />
* Внедряване на LLM API, съвместими с OpenAI (стандартизация на интерфейси)<br />
* Интеграция на векторни бази данни и RAG пайплайни<br />
* Оптимизация на латентност / разход / производителност<br />
* Оркестрация на множество модели (open-weight и хибридни)<br />
<br />
=== Стратегия за суверенен отворен код ===<br />
* Оценка на open-weight модели спрямо проприетарни решения<br />
* Подбор на LLM според сигурност, разход и технологична независимост<br />
* Стратегия edge спрямо datacenter<br />
* Стратегически мониторинг на GPU пазара и AI екосистемата<br />
* Анализ на критични зависимости (хардуер / софтуер)<br />
<br />
=== Индустриализация и управление ===<br />
* Стандартизация на AI внедряванията<br />
* Пълна автоматизация (CI/CD за AI инфраструктура)<br />
* Изолация и сегментиране на чувствителни AI натоварвания<br />
* Архивиране, възстановяване и осигуряване на непрекъсваемост<br />
* Структурирана документация и споделяне на добри практики<br />
<br />
== Опит ==<br />
<br />
=== Министерство на вътрешните работи ===<br />
''Ръководител AI инфраструктура и генеративен AI''<br />
<br />
Техническо ръководство на първите вътрешни LLM внедрявания.<br />
<br />
==== Основни постижения ====<br />
* Внедряване на първите суверенни LLM чатботи<br />
* Проектиране и индустриализация на RAG пайплайни<br />
* Създаване на специализиран GPU клъстер за AI<br />
* Автоматизация на LLM API, съвместим с OpenAI<br />
* Бенчмаркинг на производителност и латентност за множество модели<br />
* Тестване на embeddings и оптимизация на chunking<br />
* Внедряване на мултимодални решения (текст, изображения)<br />
* Създаване на защитено техническо Wiki<br />
* Разработване на стратегия за архивиране и възстановяване<br />
* Стратегически мониторинг на open-source модели и GPU пазара<br />
<br />
==== Въздействие ====<br />
* Ускоряване на AI експериментите<br />
* Изграждане на повторно използваема техническа основа<br />
* Намаляване на зависимостта от проприетарни решения<br />
* Полагане на основите на суверенна AI платформа<br />
<br />
=== ADLERE ===<br />
''Linux / Cloud архитект''<br />
<br />
* Архитектура на Linux и Cloud инфраструктури с висока наличност<br />
* Внедряване на защитени Kubernetes клъстери<br />
* Интеграция на AI решения и разпределено съхранение<br />
* Автоматизация и миграция на критични системи<br />
* Осигуряване на мрежова и информационна сигурност<br />
<br />
== Основни компетенции ==<br />
<br />
=== Лидерство в AI платформи ===<br />
* LLMOps / GPUOps<br />
* Мултимоделна архитектура<br />
* Разширен RAG<br />
* Проектиране и стандартизация на API<br />
* Инженеринг на производителност<br />
<br />
=== Инфраструктура ===<br />
* Kubernetes (production среда)<br />
* Оркестрация на NVIDIA GPU<br />
* Ceph / S3 / разпределено съхранение<br />
* Наблюдаемост и мониторинг<br />
<br />
=== Сигурност и суверенитет ===<br />
* Изолация на AI натоварвания<br />
* Криптиране и контрол на достъпа<br />
* Сегментирани среди<br />
* Подход zero-trust<br />
* Архитектура без критични външни зависимости<br />
<br />
== Позициониране ==<br />
<br />
Подходящ профил за:<br />
* Директор на суверенна AI платформа<br />
* Ръководител национална AI инфраструктура<br />
* Lead AI Systems Architect (стратегически публичен сектор)<br />
* Head of AI Platform Engineering</div>
Tcepo
https://infocepo.com/wiki/index.php?title=CV_PACHECO&diff=1922
CV PACHECO
2026-02-17T21:49:44Z
<p>Tcepo: </p>
<hr />
<div><br><br />
[https://nextcloud.infocepo.com/index.php/s/Y4nDaRGzN95GMAn Download & Contact] (Code:oPRgHPoDoH)<br><br />
<br><br />
[https://nextcloud.infocepo.com/index.php/call/b8574xjo VISIO] (Ask me an Access Control Code)<br><br />
<br><br />
<br />
= Sovereign AI Platform Lead =<br />
''National-Scale LLM Infrastructure • Secure AI Platforms • GPU Strategy • Open-Source Sovereignty''<br />
<br />
== Executive Profile ==<br />
Sovereign AI Platform Lead spécialisé dans la conception et le pilotage de plateformes nationales d’IA générative sécurisées.<br />
<br />
Expert en architecture LLM open-source, déploiement GPU souverain, API compatibles OpenAI, pipelines RAG industrialisés et gouvernance technique en environnement sensible.<br />
<br />
Capacité à transformer des expérimentations IA en plateformes robustes, sécurisées, scalables et stratégiquement alignées avec les enjeux de souveraineté technologique.<br />
<br />
== Strategic Scope ==<br />
<br />
=== National AI Infrastructure Architecture ===<br />
* Conception de plateformes LLM internes sécurisées<br />
* Architecture Kubernetes multi-nœuds avec GPU dédiés<br />
* Déploiement d’API LLM compatibles OpenAI (standardisation des interfaces)<br />
* Intégration de bases vectorielles et pipelines RAG<br />
* Optimisation latence / coût / performance<br />
* Orchestration multi-modèles (open-weight & hybrides)<br />
<br />
=== Sovereign Open-Source Strategy ===<br />
* Évaluation modèles open-weight vs propriétaires<br />
* Sélection LLM selon sécurité, coût, indépendance technologique<br />
* Stratégie edge vs datacenter<br />
* Veille stratégique GPU & écosystème IA<br />
* Analyse dépendances critiques (hardware / software)<br />
<br />
=== Industrialization & Governance ===<br />
* Standardisation des déploiements IA<br />
* Automatisation complète (CI/CD infrastructure IA)<br />
* Isolation et cloisonnement workloads sensibles<br />
* Backup, restauration, continuité d’activité<br />
* Documentation structurée & diffusion des bonnes pratiques<br />
<br />
== Experience ==<br />
<br />
=== Ministère de l’Intérieur ===<br />
''AI Infrastructure & Generative AI Lead''<br />
<br />
Pilotage technique des premières implémentations LLM internes.<br />
<br />
==== Réalisations clés ====<br />
* Déploiement des premiers ChatBots LLM souverains<br />
* Conception et industrialisation de pipelines RAG<br />
* Création d’un cluster GPU dédié IA<br />
* Automatisation d’une API LLM compatible OpenAI<br />
* Benchmark performance & latence multi-modèles<br />
* Tests embeddings & optimisation du chunking<br />
* Déploiement multimodal (texte, image)<br />
* Mise en place d’un Wiki technique sécurisé<br />
* Élaboration stratégie de sauvegarde & restauration<br />
* Veille stratégique modèles open-source et marché GPU<br />
<br />
==== Impact ====<br />
* Accélération des expérimentations IA<br />
* Structuration d’une base technique réutilisable<br />
* Réduction dépendance aux solutions propriétaires<br />
* Fondation d’une plateforme IA souveraine<br />
<br />
=== ADLERE ===<br />
''Linux / Cloud Architect''<br />
<br />
* Architecture infrastructures Linux & Cloud haute disponibilité<br />
* Déploiement clusters Kubernetes sécurisés<br />
* Intégration solutions IA & stockage distribué<br />
* Automatisation & migration systèmes critiques<br />
* Sécurisation réseau et données sensibles<br />
<br />
== Core Capabilities ==<br />
<br />
=== AI Platform Leadership ===<br />
* LLMOps / GPUOps<br />
* Architecture multi-modèles<br />
* RAG avancé<br />
* API design & standardisation<br />
* Performance engineering<br />
<br />
=== Infrastructure ===<br />
* Kubernetes production<br />
* Orchestration GPU NVIDIA<br />
* Ceph / S3 / stockage distribué<br />
* Observabilité & monitoring<br />
<br />
=== Security & Sovereignty ===<br />
* Isolation workloads IA<br />
* Chiffrement & contrôle d’accès<br />
* Environnements cloisonnés<br />
* Approche zero-trust<br />
* Architecture sans dépendance critique externe<br />
<br />
== Positioning ==<br />
<br />
Profil adapté à :<br />
* Direction Plateforme IA Souveraine<br />
* Responsable Infrastructure IA nationale<br />
* Lead AI Systems Architect (secteur public stratégique)<br />
* Head of AI Platform Engineering</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1921
Main Page
2026-02-13T00:24:33Z
<p>Tcepo: /* Open models & internal endpoints */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Welcome to the '''infocepo.com''' portal.<br />
<br />
This wiki is intended for system administrators, cloud engineers, developers, students, and enthusiasts who want to:<br />
<br />
* Understand modern architectures (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Deploy private AI assistants and productivity tools<br />
* Build hands-on labs to learn by doing<br />
* Prepare large-scale audits, migrations, and automations<br />
<br />
The goal: turn theory into '''reusable scripts, diagrams, and architectures'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Getting started quickly =<br />
<br />
== Recommended paths ==<br />
<br />
; 1. Build a private AI assistant<br />
* Deploy a typical stack: '''Open WebUI + Ollama + GPU''' (H100 or consumer-grade GPU)<br />
* Add a chat model and a summarization model<br />
* Integrate internal data (RAG, embeddings)<br />
<br />
; 2. Launch a Cloud lab<br />
* Create a small cluster (Kubernetes, OpenStack, or bare-metal)<br />
* Set up a deployment pipeline (Helm, Ansible, Terraform…)<br />
* Add an AI service (transcription, summarization, chatbot…)<br />
<br />
; 3. Prepare an audit / migration<br />
* Inventory servers with '''ServerDiff.sh'''<br />
* Design the target architecture (cloud diagrams)<br />
* Automate the migration with reproducible scripts<br />
<br />
== Content overview ==<br />
<br />
* '''AI guides & tools''' : assistants, models, evaluations, GPUs<br />
* '''Cloud & infrastructure''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Comparison tables''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= future =<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
= AI Assistants & Cloud Tools =<br />
<br />
== AI Assistants ==<br />
<br />
; '''ChatGPT'''<br />
* https://chatgpt.com ChatGPT – Public conversational assistant, suited for exploration, writing, and rapid experimentation.<br />
<br />
; '''Self-hosted AI assistants'''<br />
* https://github.com/open-webui/open-webui Open WebUI + https://www.scaleway.com/en/h100-pcie-try-it-now/ H100 GPU + https://ollama.com Ollama <br />
: Typical stack for private assistants, self-hosted LLMs, and OpenAI-compatible APIs.<br />
* https://github.com/ynotopec/summarize Private summary – Local, fast, offline summarizer for your own data.<br />
<br />
== Development, models & tracking ==<br />
<br />
; '''Discovering and tracking models'''<br />
* https://ollama.com/library LLM Trending – Model library (chat, code, RAG…) for local deployment.<br />
* https://huggingface.co/models Models Trending – Model marketplace, filterable by task, size, and license.<br />
* https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending – Vision-language models (image → text).<br />
* https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation – Image generation model comparisons.<br />
<br />
; '''Evaluation & benchmarks'''<br />
* https://lmarena.ai/leaderboard ChatBot Evaluation – Chatbot rankings (open-source and proprietary models).<br />
* https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard – Benchmark of embedding models for RAG and semantic search.<br />
* https://ann-benchmarks.com Vectors DB Ranking – Vector database comparison (latency, memory, features).<br />
* https://top500.org/lists/green500/ HPC Efficiency – Ranking of the most energy-efficient supercomputers.<br />
<br />
; '''Development & fine-tuning tools'''<br />
* https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending – Major recent open-source projects, sorted by popularity and activity.<br />
* https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning – Advanced framework for LLM fine-tuning (instruction tuning, LoRA, etc.).<br />
* https://www.perplexity.ai Perplexity AI – Advanced research and synthesis oriented as a “research copilot”.<br />
<br />
== AI Hardware & GPUs ==<br />
<br />
; '''GPUs & accelerators'''<br />
* https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100 – Datacenter GPU for Kubernetes clusters and intensive AI workloads.<br />
* NVIDIA 5080 – Consumer GPU for lower-cost private LLM deployments.<br />
* https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator – Hardware accelerator dedicated to LLM inference.<br />
<br />
----<br />
<br />
= Open models & internal endpoints =<br />
<br />
''(Last update: 2026-02-13)''<br />
<br />
The models below correspond to '''logical endpoints''' (for example via a proxy or gateway), selected for specific use cases.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Primary use case<br />
|-<br />
| '''ai-chat''' || Based on '''gpt-oss-20b''' – General-purpose chat, good cost / quality balance.<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, temperature = 0 – Deterministic, reproducible translation (FR, EN, other languages).<br />
|-<br />
| '''ai-summary''' || qwen3 – Model optimized for summarizing long texts (reports, documents, transcriptions).<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – Code reasoning, explanation, and refactoring.<br />
|-<br />
| '''ai-code-completion''' || gpt-oss-20b – Fast code completion, designed for IDE auto-completion.<br />
|-<br />
| '''ai-parse''' || qwen3 – Structured extraction, log / JSON / table parsing.<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG usage in French (business knowledge, internal FAQs).<br />
|-<br />
| '''gpt-oss-20b''' || Agentic tasks.<br />
|}<br />
<br />
Usage idea: each endpoint is associated with one or more labs (chat, summary, parsing, RAG, etc.) in the Cloud Lab section.<br />
<br />
----<br />
<br />
= News & Trends =<br />
<br />
* https://www.youtube.com/@lev-selector/videos Top AI News – Curated AI news videos.<br />
* https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper – Example of real-time transcription with speaker detection.<br />
* https://github.com/openai-translator/openai-translator OpenAI Translator – Modern extension / client for LLM-assisted translation.<br />
* https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM – Conversational search based on LLMs and OpenSearch.<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained – Introduction to Transformers, the core architecture of LLMs.<br />
* Hands-on labs, scripts, and real-world feedback in the [[LAB project|CLOUD LAB]] project below.<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
The '''Cloud Lab''' provides reproducible scenarios: infrastructure audits, cloud migration, automation, high availability.<br />
<br />
== Audit project – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Bash audit script to:<br />
<br />
* detect configuration drift,<br />
* compare multiple environments,<br />
* prepare a migration or remediation plan.<br />
<br />
== Example of Cloud migration ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
Example: migration of virtual environments to a modernized cloud, including audit, architecture design, and automation.<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Infrastructure audit || 82 services, automated audit via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Cloud architecture diagram || Visual design and documentation || 1.5<br />
|-<br />
| Compliance checks || 2 clouds, 6 hypervisors, 6 TB of RAM || 1.5<br />
|-<br />
| Cloud platform installation || Deployment of main target environments || 1.0<br />
|-<br />
| Stability verification || Early functional tests || 0.5<br />
|-<br />
| Automation study || Identification and automation of repetitive tasks || 1.5<br />
|-<br />
| Template development || 6 templates, 8 environments, 2 clouds / OS || 1.5<br />
|-<br />
| Migration diagram || Illustration of the migration process || 1.0<br />
|-<br />
| Migration code writing || 138 lines (see '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Process stabilization || Validation that migration is reproducible || 1.5<br />
|-<br />
| Cloud benchmarking || Performance comparison vs legacy infrastructure || 1.5<br />
|-<br />
| Downtime tuning || Calculation of outage time per migration || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs per VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 person-days<br />
|}<br />
<br />
=== Stability checks (minimal HA) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Expected result<br />
|-<br />
| Shutdown of one node || All services must automatically restart on remaining nodes.<br />
|-<br />
| Simultaneous shutdown / restart of all nodes || All services must recover correctly after reboot.<br />
|}<br />
<br />
----<br />
<br />
= Web Architecture & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principles for designing scalable and portable web architectures:<br />
<br />
* Favor '''simple, modular, and flexible''' infrastructure.<br />
* Follow client location (GDNS or equivalent) to bring content closer.<br />
* Use network load balancers (LVS, IPVS) for scalability.<br />
* Systematically compare costs and beware of '''vendor lock-in'''.<br />
* TLS:<br />
** HAProxy for fast frontends,<br />
** Envoy for compatibility and advanced use cases (mTLS, HTTP/2/3).<br />
* Caching:<br />
** Varnish, Apache Traffic Server for large content volumes.<br />
* Favor open-source stacks and database caches (e.g., Memcached).<br />
* Use message queues, buffers, and quotas to smooth traffic spikes.<br />
* For complete architectures:<br />
** https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture<br />
** https://github.com/systemdesign42/system-design System Design GitHub<br />
<br />
----<br />
<br />
= Comparison of major Cloud platforms =<br />
<br />
{| class="wikitable"<br />
! Feature !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router control''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Network virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
This table serves as a starting point for choosing the right stack based on:<br />
<br />
* Desired level of control (API vs bare-metal),<br />
* Context (on-prem, public cloud, HPC, CRM…),<br />
* Existing automation tooling.<br />
<br />
----<br />
<br />
= Useful Cloud & IT links =<br />
<br />
* https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared – AWS / Azure / GCP service mapping.<br />
* https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map – Global Internet mapping.<br />
* https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape – Overview of cloud-native projects (CNCF).<br />
* https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki – Wikimedia infrastructure, real large-scale example.<br />
* https://openapm.io OpenAPM – SRE Tools – APM / observability tooling.<br />
* https://access.redhat.com/downloads/content/package-browser RedHat Package Browser – Package and version search at Red Hat.<br />
* https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Barometer of IT freelance daily rates.<br />
* https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor) – Salary indicators.<br />
<br />
----<br />
<br />
= Advanced: High Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Basic principles:<br />
<br />
* Multi-node or multi-site clusters for redundancy.<br />
* Use of IPMI for fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* For a 2-node cluster:<br />
– carefully sequence fencing to avoid split-brain,<br />
– 3 or more nodes remain recommended for production.<br />
<br />
=== Common resource patterns ===<br />
<br />
* Multipath storage, LUNs, LVM, NFS.<br />
* User resources and application processes.<br />
* Virtual IPs, DNS records, network listeners.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* Job orchestration (SLURM or equivalent).<br />
* High-performance shared storage (GPFS, Lustre…).<br />
* Possible integration with AI workloads (large-scale training, GPU inference).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD pipelines with built-in security checks (linting, SAST, DAST, SBOM).<br />
* Observability (logs, metrics, traces) integrated from design time.<br />
* Automated vulnerability scanning, secret management, policy-as-code.<br />
<br />
----<br />
<br />
= About & Contributions =<br />
<br />
For more examples, scripts, diagrams, and feedback, see:<br />
<br />
* https://infocepo.com infocepo.com<br />
<br />
Suggestions for corrections, diagram improvements, or new labs are welcome. <br />
This wiki aims to remain a '''living laboratory''' for AI, cloud, and automation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1920
Human needs
2026-01-31T16:19:50Z
<p>Tcepo: </p>
<hr />
<div>= Fundamental Objects of Civilization and Material Dependencies =<br />
<br />
== Introduction ==<br />
This article presents a structured view of the fundamental objects required for a modern civilization, their material dependencies, and the scientific and technological foundations that make them possible.<br />
It connects everyday needs (shelter, hygiene, transport, tools), information systems (processing, storage, transmission), material constraints (abundant vs rare elements), and indispensable scientific discoveries.<br />
<br />
== 1. Fundamental Objects of Civilization (g1) ==<br />
<br />
=== 1.1 Physical and Material Needs ===<br />
* '''[https://en.wikipedia.org/wiki/Shelter Shelter]'''<br />
* '''Structured elementary particles'''<br />
* '''[https://en.wikipedia.org/wiki/Clothes Clothes]'''<br />
* '''[https://en.wikipedia.org/wiki/Toilet Toilet]'''<br />
* '''[https://en.wikipedia.org/wiki/Cleaner Cleaner]'''<br />
* '''[https://en.wikipedia.org/wiki/Transportation Transportation]'''<br />
* '''[https://en.wikipedia.org/wiki/Defense Defense]'''<br />
* '''[https://en.wikipedia.org/wiki/Help Help]'''<br />
* '''[https://en.wikipedia.org/wiki/Cutlery Cutlery]'''<br />
* '''[https://en.wikipedia.org/wiki/Mattress Mattress]'''<br />
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Table (furniture)]'''<br />
* '''[https://en.wikipedia.org/wiki/Chair Chair]'''<br />
<br />
=== 1.2 Leisure and Formation ===<br />
* '''[https://en.wikipedia.org/wiki/Leisure Leisure]'''<br />
* '''[https://en.wikipedia.org/wiki/Training Training]'''<br />
* '''[https://en.wikipedia.org/wiki/Stimulants Stimulants]'''<br />
<br />
== 2. Information Chain (functional objects) ==<br />
<br />
=== 2.1 Information Transformation ===<br />
* '''[https://en.wikipedia.org/wiki/Transducer Information transducer]'''<br />
* '''[https://en.wikipedia.org/wiki/Information_processing Information processing]'''<br />
<br />
=== 2.2 Storage, Search, and Transmission ===<br />
* '''[https://en.wikipedia.org/wiki/Data_storage Data storage]'''<br />
* '''[https://en.wikipedia.org/wiki/Search_engine Information seeker]'''<br />
* '''[https://en.wikipedia.org/wiki/Transmitter Information transmitter]'''<br />
<br />
== 3. Material Dependencies: AU vs RS ==<br />
<br />
=== 3.1 Legend ===<br />
* '''AU''' = Abundant / Universal elements<br />
(CHON + Si, Al, Fe, Ca, Mg, Na, K + S, P, Cl + Cu, Zn)<br />
* '''RS''' = Rare / Strategic elements<br />
(Cr, Ni, Ti, Mn, B, F, Li + Nd/Pr/Dy, Co, W, Ag, Au, Ga, As…)<br />
* Score: 0–5 (5 = critical, 0 = negligible)<br />
<br />
=== 3.2 Dependency Table by Object (g1) ===<br />
{| class="wikitable sortable"<br />
! Item (g1)<br />
! AU (elements + score)<br />
! RS (elements + score)<br />
! Notes<br />
|-<br />
| Shelter<br />
| Si,O,Al,Ca,Fe,Mg,C,H (5)<br />
| Cr,Ni,Ti,B (3)<br />
| Concrete, glass, ceramics, steel; alloys if available<br />
|-<br />
| Structured elementary particles<br />
| C,H,O,N,Si,Fe,Al (5)<br />
| Ti,Cr,Ni,B (2)<br />
| Robust material structures without high technology<br />
|-<br />
| Clothes<br />
| C,H,O,N,S (5)<br />
| F,Ti (2)<br />
| Fibers and treatments; F for membranes<br />
|-<br />
| Toilet<br />
| Si,O,Al,Ca,Fe,C,H (5)<br />
| Cr,Ni,Cu (3)<br />
| Ceramics and plumbing; stainless steel preferred<br />
|-<br />
| Cleaner<br />
| H,O,Na,Cl,C,S (5)<br />
| F,P (2)<br />
| Bases, salts, solvents; P for advanced detergents<br />
|-<br />
| Transportation<br />
| Fe,Al,Si,C,O,Cu (5)<br />
| Ni,Cr,Ti,Mn,Li (4)<br />
| Alloys, wiring; Li for electric systems<br />
|-<br />
| Defense<br />
| Fe,C,Al,Si (4)<br />
| Ti,Cr,Ni,W (4)<br />
| W for high density<br />
|-<br />
| Help<br />
| Fe,C,H,O,N,Ca,Na,Cl (5)<br />
| Cu,Zn,Ag (2)<br />
| Hygiene and basic tools<br />
|-<br />
| Information transducer<br />
| Si,O,Al,Fe,C,Cu (4)<br />
| B,Ga (4)<br />
| Sensors and actuators need dopants<br />
|-<br />
| Information processing<br />
| Si,O,Al,C,Cu (4)<br />
| B,P,As,Ga (5)<br />
| Computing depends on doping and fine processes<br />
|-<br />
| Data storage<br />
| C,H,O (paper), Fe (magnetic), Si,O (glass) (4)<br />
| Co,Ni,Nd (4)<br />
| High-end storage requires Co/Ni and magnets<br />
|-<br />
| Information seeker<br />
| Si,Al,Cu,C (4)<br />
| Li,Ni,Co,Nd (4)<br />
| Compute, energy, interconnects<br />
|-<br />
| Information transmitter<br />
| Cu,Al,Si,O (5)<br />
| Ag,Au (2)<br />
| Copper/aluminum + fiber sufficient<br />
|-<br />
| Stimulants<br />
| C,H,O,N,S,P (5)<br />
| — (0)<br />
| Organic chemistry and nutrition<br />
|-<br />
| Leisure<br />
| C,H,O,N,Si,Fe,Al (4)<br />
| Cu,Li,Nd (2)<br />
| Electronics optional<br />
|-<br />
| Training<br />
| C,H,O,N,Si (4)<br />
| Cu,Li (2)<br />
| Media and tools<br />
|-<br />
| Cutlery<br />
| Fe,C (5)<br />
| Cr,Ni (4)<br />
| Stainless steel upgrade<br />
|-<br />
| Mattress<br />
| C,H,O,N (5)<br />
| Si (1)<br />
| Foams, latex, fibers<br />
|-<br />
| Table (furniture)<br />
| C,H,O or Fe/Al or Si,O (glass) (5)<br />
| Cr,Ni,Ti (1)<br />
| Easily built with abundant elements<br />
|-<br />
| Chair<br />
| C,H,O or Fe/Al (5)<br />
| Ti,Cr (1)<br />
| Abundant elements sufficient<br />
|}<br />
<br />
== 4. Rare / Strategic Elements (RS) and Sourcing ==<br />
<br />
=== 4.1 Main RS Elements and Producing Countries ===<br />
{| class="wikitable sortable"<br />
! RS element<br />
! Main use<br />
! Main sourcing countries<br />
|-<br />
| Cr (Chromium)<br />
| Stainless steel, corrosion resistance<br />
| South Africa; Kazakhstan; India; Turkey<br />
|-<br />
| Ni (Nickel)<br />
| Stainless steel, batteries<br />
| Indonesia; Philippines; Russia; Canada; Australia; New Caledonia; Brazil<br />
|-<br />
| Ti (Titanium)<br />
| Light and strong alloys<br />
| China; Mozambique; South Africa; Australia; Canada; India<br />
|-<br />
| Mn (Manganese)<br />
| Steelmaking, batteries<br />
| South Africa; Gabon; Australia; China; India<br />
|-<br />
| B (Boron)<br />
| Glass and steel modifier<br />
| Turkey; USA; Argentina; Chile; Russia; China<br />
|-<br />
| F (Fluorspar)<br />
| Fluorochemicals, batteries<br />
| China; Mexico; Mongolia; South Africa; Vietnam<br />
|-<br />
| Li (Lithium)<br />
| Batteries, energy storage<br />
| Australia; Chile; Argentina; China; Zimbabwe; Canada; Brazil<br />
|-<br />
| Co (Cobalt)<br />
| Batteries, alloys<br />
| DR Congo; Indonesia; Russia; Australia; Philippines; Cuba<br />
|-<br />
| W (Tungsten)<br />
| Hardness, density<br />
| China; Vietnam; Russia; Rwanda; Spain; Austria; Bolivia; Portugal<br />
|-<br />
| Ga (Gallium)<br />
| RF and power electronics<br />
| China; Japan; South Korea; Russia<br />
|-<br />
| Nd/Pr/Dy<br />
| Permanent magnets<br />
| China; Australia; USA; Myanmar<br />
|}<br />
<br />
== 5. Minimal RS Subsets by Civilization Type ==<br />
* '''Durable everyday civilization''' : Cr, Ni, Ti, Mn, B<br />
* '''Electronics-heavy civilization''' : Ga, Nd/Pr/Dy, Li, Co, Cu<br />
* '''Extreme density and hardness''' : W<br />
<br />
== 6. Examples of Global Information Infrastructure ==<br />
<br />
=== 6.1 Most Visited Websites (functional classification) ===<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messaging<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Adult content<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom] || Videoconferencing<br />
|-<br />
| [https://www.nytimes.com New York Times] || News media<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming platform<br />
|}<br />
<br />
== 7. Open Source Trend (examples) ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] — image generation model<br />
<br />
== 8. Scientific Discoveries Still Indispensable Today (Post-1700) ==<br />
<br />
=== 18th century ===<br />
* 1712–1781 — Steam engine<br />
* 1796 — Vaccination<br />
* Late 18th century — Modern chemistry<br />
<br />
=== 19th century ===<br />
* 1800 — Electric battery<br />
* 1824–1870 — Thermodynamics<br />
* 1831–1860 — Electromagnetism<br />
* 1846–1847 — Modern anesthesia<br />
* 1850–1880 — Germ theory and asepsis<br />
* 1865 — Genetics<br />
* 1869 — Periodic table<br />
* 1895 — X-rays<br />
* 1897 — Electron<br />
<br />
=== 20th century ===<br />
* 1900–1930 — Quantum mechanics<br />
* 1905–1915 — Relativity<br />
* 1909–1913 — Haber–Bosch process<br />
* 1928 — Antibiotics<br />
* 1947 — Transistor<br />
* 1948 — Information theory<br />
* 1953 — DNA structure<br />
* 1958–1959 — Integrated circuit<br />
* 1960 — Laser<br />
* 1960s–1980s — Internet<br />
* 1970s — Modern cryptography<br />
* 1983 — PCR<br />
<br />
=== 21st century ===<br />
* 2012 — CRISPR<br />
* 2010s — Deep learning<br />
<br />
== 9. Notes and Limits ==<br />
* Refining is often more critical than extraction<br />
* Some RS elements can be substituted at performance cost<br />
* Recycling partially reduces RS dependency but does not eliminate it</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1919
Human needs
2026-01-31T16:16:11Z
<p>Tcepo: Undo revision 1918 by Tcepo (talk)</p>
<hr />
<div>* '''[https://en.wikipedia.org/wiki/Shelter Shelter]''' : g1[[https://en.wikipedia.org/wiki/Particle Particle], [https://en.wikipedia.org/wiki/Motion Motion], [https://en.wikipedia.org/wiki/Fundamental_interaction Fundamental interaction], [https://en.wikipedia.org/wiki/Science Science]]<br />
* '''Structured elementary particles''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Clothes Clothes]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Toilet Toilet]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cleaner Cleaner]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transportation Transportation]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Defense Defense]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Help Help]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transducer Information transducer]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Information_processing Information processing]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Data_storage Information storer]''' (like disks, like books) : g1<br />
* '''[https://en.wikipedia.org/wiki/Search_engine Information seeker]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transmitter Information transmitter]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Stimulants Stimulants]''' (like structured elementary particles, like love, like entertainment...) : g1<br />
* '''[https://en.wikipedia.org/wiki/Particle Particle]'''<br />
* '''[https://en.wikipedia.org/wiki/Leisure Leisure]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Training Training]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cutlery Cutlery]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Mattress Mattress]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Table (furniture)]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Chair Chair]''' : g1<br />
<br><br />
Legend<br />
<br />
AU = Abundant / Universal elements<br />
(CHON + Si, Al, Fe, Ca, Mg, Na, K + S, P, Cl + Cu, Zn)<br />
<br />
RS = Rare / Strategic elements<br />
(Cr, Ni, Ti, Mn, B, F, Li + Nd/Pr/Dy, Co, W, Ag, Au, Ga, As…)<br />
<br />
Score: 0–5 (5 = critical, 0 = negligible)<br />
<br><br />
{| class="wikitable sortable"<br />
! Item (g1)<br />
! AU (elements + score)<br />
! RS (elements + score)<br />
! Notes (one line)<br />
|-<br />
| Shelter<br />
| Si,O,Al,Ca,Fe,Mg,C,H (5)<br />
| Cr,Ni,Ti,B (3)<br />
| Concrete/glass/ceramics + steel; alloys if available<br />
|-<br />
| Structured elementary particles (material structures)<br />
| C,H,O,N,Si,Fe,Al (5)<br />
| Ti,Cr,Ni,B (2)<br />
| Robust structures without high tech<br />
|-<br />
| Clothes<br />
| C,H,O,N,S (5)<br />
| F,Ti (2)<br />
| Fibers + treatments; F for high-performance membranes<br />
|-<br />
| Toilet<br />
| Si,O,Al,Ca,Fe,C,H (5)<br />
| Cr,Ni,Cu (3)<br />
| Ceramics + plumbing; stainless steel preferred<br />
|-<br />
| Cleaner<br />
| H,O,Na,Cl,C,S (5)<br />
| F,P (2)<br />
| Bases/salts/solvents; P for advanced detergents<br />
|-<br />
| Transportation<br />
| Fe,Al,Si,C,O,Cu (5)<br />
| Ni,Cr,Ti,Mn,Li (4)<br />
| Alloys + wiring; Li for electric systems<br />
|-<br />
| Defense<br />
| Fe,C,Al,Si (4)<br />
| Ti,Cr,Ni,W (4)<br />
| W useful for high density (rare)<br />
|-<br />
| Help (tools/basic care)<br />
| Fe,C,H,O,N,Ca,Na,Cl (5)<br />
| Cu,Zn,Ag (2)<br />
| Hygiene + tools; Ag optional antimicrobial<br />
|-<br />
| Information transducer<br />
| Si,O,Al,Fe,C,Cu (4)<br />
| B,Ga (4)<br />
| Sensors/actuators need dopants/semiconductors<br />
|-<br />
| Information processing<br />
| Si,O,Al,C,Cu (4)<br />
| B,P,As,Ga (5)<br />
| Modern computing = doping + fine processes<br />
|-<br />
| Data storage (disks/books)<br />
| C,H,O (paper), Fe (magnetic), Si,O (glass) (4)<br />
| Co,Ni,Nd (4)<br />
| High-end magnetic storage needs Co/Ni + Nd magnets<br />
|-<br />
| Information seeker (search engine)<br />
| Si,Al,Cu,C (4)<br />
| Li,Ni,Co,Nd (4)<br />
| Compute + energy + interconnects<br />
|-<br />
| Information transmitter<br />
| Cu,Al,Si,O (5)<br />
| Ag,Au (2)<br />
| Cu/Al + SiO2 fiber sufficient; nobles for reliability<br />
|-<br />
| Stimulants (broad sense)<br />
| C,H,O,N,S,P (5)<br />
| — (0)<br />
| Organic chemistry + nutrition; rares add little<br />
|-<br />
| Leisure<br />
| C,H,O,N,Si,Fe,Al (4)<br />
| Cu,Li,Nd (2)<br />
| Simple materials; electronics optional<br />
|-<br />
| Training<br />
| C,H,O,N,Si (4)<br />
| Cu,Li (2)<br />
| Media + tools; electronics optional<br />
|-<br />
| Cutlery<br />
| Fe,C (5)<br />
| Cr,Ni (4)<br />
| Stainless steel is a major practical upgrade<br />
|-<br />
| Mattress<br />
| C,H,O,N (5)<br />
| Si (1)<br />
| Foams/latex/fibers; silicones optional<br />
|-<br />
| Table (furniture)<br />
| C,H,O (wood/polymers) or Fe/Al or Si,O (glass) (5)<br />
| Cr,Ni,Ti (1)<br />
| Easily built with abundant elements<br />
|-<br />
| Chair<br />
| C,H,O or Fe/Al (5)<br />
| Ti,Cr (1)<br />
| Same: abundants are sufficient<br />
|}<br />
<br />
<br />
=== Rare/Strategic elements (RS) – where to source (countries) ===<br />
<br />
;Legend<br />
* AU = Abundant/Universal (easy)<br />
* RS = Rare/Strategic (hard / concentrated supply)<br />
<br />
{| class="wikitable sortable"<br />
! RS element<br />
! Main use (why you care)<br />
! Main sourcing countries (today)<br />
|-<br />
| Cr (Chromium)<br />
| Stainless / corrosion resistance<br />
| South Africa (dominant); Kazakhstan, India, Turkey (notable)<br />
|-<br />
| Ni (Nickel)<br />
| Stainless + batteries<br />
| Indonesia (dominant); Philippines, Russia, Canada, Australia, New Caledonia, Brazil<br />
|-<br />
| Ti (Titanium minerals: ilmenite/rutile)<br />
| Light strong alloys, durable structures<br />
| China; Mozambique; South Africa; Australia; Canada; India<br />
|-<br />
| Mn (Manganese)<br />
| Steelmaking + some battery chemistries<br />
| South Africa (largest); Gabon; Australia; China; India<br />
|-<br />
| B (Boron)<br />
| Glass/ceramics; steel/glass modifiers<br />
| Turkey (very large); USA; Argentina; Chile; Russia; China<br />
|-<br />
| F (Fluorspar → fluorochemicals)<br />
| HF/fluoropolymers; batteries; refrigerants<br />
| China (largest); Mexico; Mongolia; South Africa; Vietnam<br />
|-<br />
| Li (Lithium)<br />
| Batteries / energy storage<br />
| Australia; Chile; Argentina; China; Zimbabwe; Canada; Brazil<br />
|-<br />
| Co (Cobalt)<br />
| Batteries / high-performance alloys<br />
| DR Congo (mining dominant); Indonesia (growing); Russia, Australia, Philippines, Cuba (notable); refining largely China<br />
|-<br />
| W (Tungsten)<br />
| High density + hardness (tools/defense)<br />
| China (dominant); Vietnam; Russia; Rwanda; Spain; Austria; Bolivia; Portugal<br />
|-<br />
| Ga (Gallium)<br />
| RF/power electronics (hard bottleneck)<br />
| China (overwhelming); Japan, South Korea, Russia (secondary)<br />
|-<br />
| Nd/Pr/Dy (Rare earths for magnets)<br />
| Motors/actuators/sensors; “high torque density”<br />
| China (largest); Australia; USA; Myanmar (notable); processing often China-centric<br />
|}<br />
<br />
=== Minimal RS subsets (pick what you build) ===<br />
<br />
;RS for “durable everyday civilization” (shelter + toilet + tools + transport)<br />
* Cr + Ni (stainless)<br />
* Ti (light/strong)<br />
* Mn (steel)<br />
* B (glass/steel modifier)<br />
<br />
;RS for “electronics-heavy” (processing + transducer + transmitter + seeker)<br />
* Ga (critical)<br />
* Nd/Pr/Dy (magnets)<br />
* Li (batteries)<br />
* Co (some chemistries)<br />
* (plus Cu from AU for wiring)<br />
<br />
;RS for “extreme density/hardness”<br />
* W<br />
<br />
<br />
==TOP WEBSITES (CATEGORY AWARE)==<br />
# https://en.wikipedia.org/wiki/List_of_most-visited_websites<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing service<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messenger<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Online gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming service<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Pornography<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming service<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting service<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom Communications] || Computers Electronics and Technology - Other<br />
|-<br />
| [https://www.nytimes.com The New York Times] || News & Media Publishers<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming and game development platform<br />
|}<br />
<br />
== Open Source trend ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] : Image generative<br />
<br />
<br />
== Scientific discoveries (after 1700) still indispensable today ==<br />
Chronological list of scientific discoveries and breakthroughs made after 1700 that are still directly used in modern civilization (energy, health, industry, digital technologies, agriculture, communications).<br />
<br />
=== 18th century ===<br />
* '''1712–1781''' – '''Steam engine''' : foundation of thermal engines, turbines, and energy engineering.<br />
* '''1796''' – '''Vaccination''' : large-scale prevention of infectious diseases.<br />
* '''Late 18th century''' – '''Modern chemistry''' (conservation of matter, chemical nomenclature) : chemical industry, materials, analytical science.<br />
<br />
=== 19th century ===<br />
* '''1800''' – '''Electric battery''' : batteries, electrochemistry, energy storage.<br />
* '''1824–1870''' – '''Thermodynamics''' : engines, power plants, refrigeration, energy efficiency.<br />
* '''1831–1860''' – '''Electromagnetism''' : electricity generation and distribution, telecommunications.<br />
* '''1846–1847''' – '''Modern anesthesia''' : safe and scalable surgery.<br />
* '''1850–1880''' – '''Germ theory + asepsis''' : hygiene, drinking water, modern hospitals.<br />
* '''1865''' – '''Genetics''' : heredity, selection, biotechnology.<br />
* '''1869''' – '''Periodic table of elements''' : chemistry, materials, pharmaceuticals.<br />
* '''1895''' – '''X-rays''' : non-invasive medical imaging.<br />
* '''1897''' – '''Discovery of the electron''' : foundation of electronics.<br />
<br />
=== First half of the 20th century ===<br />
* '''1900–1930''' – '''Quantum mechanics''' : semiconductors, lasers, sensors, modern chemistry.<br />
* '''1905–1915''' – '''Relativity''' : precise time synchronization (GPS, networks).<br />
* '''1909–1913''' – '''Haber–Bosch process''' : nitrogen fertilizers, global intensive agriculture.<br />
* '''1928''' – '''Antibiotics''' : treatment of bacterial infections.<br />
* '''1947''' – '''Transistor''' : basis of modern electronics and computing.<br />
* '''1948''' – '''Information theory''' : compression, error correction, digital communications.<br />
<br />
=== Second half of the 20th century ===<br />
* '''1953''' – '''Structure of DNA''' : molecular biology, modern medicine.<br />
* '''1958–1959''' – '''Integrated circuit''' : miniaturization, mass computing.<br />
* '''1960''' – '''Laser''' : fiber optics, industry, medicine, metrology.<br />
* '''1960s–1980s''' – '''Packet-switched networks → Internet''' : global communication.<br />
* '''1970s''' – '''Modern cryptography''' : secure digital communications.<br />
* '''1983''' – '''PCR (DNA amplification)''' : diagnostics, biotechnology, forensics.<br />
<br />
=== 21st century ===<br />
* '''2012''' – '''CRISPR''' : targeted gene editing.<br />
* '''2010s''' – '''Deep learning''' : vision, translation, optimization, digital assistants.<br />
<br />
=== Note ===<br />
This list includes only discoveries that are '''still indispensable today''' — removing any of them would cause immediate collapse of major sectors (energy, health, food, digital systems, communications, industry).<br />
<br />
<br />
ynotopec at gmail.com (mail me)</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1918
Human needs
2026-01-31T16:14:42Z
<p>Tcepo: </p>
<hr />
<div>= Fundamental Objects of Civilization and Material Dependencies =<br />
<br />
== Introduction ==<br />
This article presents a structured view of the fundamental objects required for a modern civilization, their material dependencies, and the scientific and technological foundations that make them possible. <br />
It connects everyday needs (shelter, hygiene, transport, tools), information systems (processing, storage, transmission), material constraints (abundant vs rare elements), and indispensable scientific discoveries.<br />
<br />
---<br />
<br />
== 1. Fundamental Objects of Civilization (g1) ==<br />
<br />
=== 1.1 Physical and Material Needs ===<br />
* '''[[https://en.wikipedia.org/wiki/Shelter|Shelter]]'''<br />
* '''Structured elementary particles'''<br />
* '''[[https://en.wikipedia.org/wiki/Clothes|Clothes]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Toilet|Toilet]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Cleaner|Cleaner]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Transportation|Transportation]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Defense|Defense]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Help|Help]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Cutlery|Cutlery]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Mattress|Mattress]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Table_(furniture)|Table (furniture)]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Chair|Chair]]'''<br />
<br />
=== 1.2 Leisure and Formation ===<br />
* '''[[https://en.wikipedia.org/wiki/Leisure|Leisure]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Training|Training]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Stimulants|Stimulants]]'''<br />
<br />
---<br />
<br />
== 2. Information Chain (functional objects) ==<br />
<br />
=== 2.1 Information Transformation ===<br />
* '''[[https://en.wikipedia.org/wiki/Transducer|Information transducer]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Information_processing|Information processing]]'''<br />
<br />
=== 2.2 Storage, Search, and Transmission ===<br />
* '''[[https://en.wikipedia.org/wiki/Data_storage|Data storage]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Search_engine|Information seeker]]'''<br />
* '''[[https://en.wikipedia.org/wiki/Transmitter|Information transmitter]]'''<br />
<br />
---<br />
<br />
== 3. Material Dependencies: AU vs RS ==<br />
<br />
=== 3.1 Legend ===<br />
* '''AU''' = Abundant / Universal elements <br />
(CHON + Si, Al, Fe, Ca, Mg, Na, K + S, P, Cl + Cu, Zn)<br />
* '''RS''' = Rare / Strategic elements <br />
(Cr, Ni, Ti, Mn, B, F, Li + Nd/Pr/Dy, Co, W, Ag, Au, Ga, As…)<br />
* Score: 0–5 (5 = critical, 0 = negligible)<br />
<br />
---<br />
<br />
=== 3.2 Dependency Table by Object (g1) ===<br />
<br />
{| class="wikitable sortable"<br />
! Item (g1)<br />
! AU (elements + score)<br />
! RS (elements + score)<br />
! Notes<br />
|-<br />
| Shelter<br />
| Si,O,Al,Ca,Fe,Mg,C,H (5)<br />
| Cr,Ni,Ti,B (3)<br />
| Concrete, glass, ceramics, steel; alloys if available<br />
|-<br />
| Structured elementary particles<br />
| C,H,O,N,Si,Fe,Al (5)<br />
| Ti,Cr,Ni,B (2)<br />
| Robust material structures without high technology<br />
|-<br />
| Clothes<br />
| C,H,O,N,S (5)<br />
| F,Ti (2)<br />
| Fibers and treatments; F for membranes<br />
|-<br />
| Toilet<br />
| Si,O,Al,Ca,Fe,C,H (5)<br />
| Cr,Ni,Cu (3)<br />
| Ceramics and plumbing; stainless steel preferred<br />
|-<br />
| Cleaner<br />
| H,O,Na,Cl,C,S (5)<br />
| F,P (2)<br />
| Bases, salts, solvents; P for advanced detergents<br />
|-<br />
| Transportation<br />
| Fe,Al,Si,C,O,Cu (5)<br />
| Ni,Cr,Ti,Mn,Li (4)<br />
| Alloys, wiring; Li for electric systems<br />
|-<br />
| Defense<br />
| Fe,C,Al,Si (4)<br />
| Ti,Cr,Ni,W (4)<br />
| W for high density<br />
|-<br />
| Help<br />
| Fe,C,H,O,N,Ca,Na,Cl (5)<br />
| Cu,Zn,Ag (2)<br />
| Hygiene and basic tools<br />
|-<br />
| Information transducer<br />
| Si,O,Al,Fe,C,Cu (4)<br />
| B,Ga (4)<br />
| Sensors and actuators need dopants<br />
|-<br />
| Information processing<br />
| Si,O,Al,C,Cu (4)<br />
| B,P,As,Ga (5)<br />
| Computing depends on doping and fine processes<br />
|-<br />
| Data storage<br />
| C,H,O (paper), Fe (magnetic), Si,O (glass) (4)<br />
| Co,Ni,Nd (4)<br />
| High-end storage requires Co/Ni and magnets<br />
|-<br />
| Information seeker<br />
| Si,Al,Cu,C (4)<br />
| Li,Ni,Co,Nd (4)<br />
| Compute, energy, interconnects<br />
|-<br />
| Information transmitter<br />
| Cu,Al,Si,O (5)<br />
| Ag,Au (2)<br />
| Copper/aluminum + fiber sufficient<br />
|-<br />
| Stimulants<br />
| C,H,O,N,S,P (5)<br />
| — (0)<br />
| Organic chemistry and nutrition<br />
|-<br />
| Leisure<br />
| C,H,O,N,Si,Fe,Al (4)<br />
| Cu,Li,Nd (2)<br />
| Electronics optional<br />
|-<br />
| Training<br />
| C,H,O,N,Si (4)<br />
| Cu,Li (2)<br />
| Media and tools<br />
|-<br />
| Cutlery<br />
| Fe,C (5)<br />
| Cr,Ni (4)<br />
| Stainless steel upgrade<br />
|-<br />
| Mattress<br />
| C,H,O,N (5)<br />
| Si (1)<br />
| Foams, latex, fibers<br />
|-<br />
| Table (furniture)<br />
| C,H,O or Fe/Al or Si,O (glass) (5)<br />
| Cr,Ni,Ti (1)<br />
| Easily built with abundant elements<br />
|-<br />
| Chair<br />
| C,H,O or Fe/Al (5)<br />
| Ti,Cr (1)<br />
| Abundant elements sufficient<br />
|}<br />
<br />
---<br />
<br />
== 4. Rare / Strategic Elements (RS) and Sourcing ==<br />
<br />
=== 4.1 Main RS Elements and Producing Countries ===<br />
<br />
{| class="wikitable sortable"<br />
! RS element<br />
! Main use<br />
! Main sourcing countries<br />
|-<br />
| Cr (Chromium)<br />
| Stainless steel, corrosion resistance<br />
| South Africa; Kazakhstan; India; Turkey<br />
|-<br />
| Ni (Nickel)<br />
| Stainless steel, batteries<br />
| Indonesia; Philippines; Russia; Canada; Australia; New Caledonia; Brazil<br />
|-<br />
| Ti (Titanium)<br />
| Light and strong alloys<br />
| China; Mozambique; South Africa; Australia; Canada; India<br />
|-<br />
| Mn (Manganese)<br />
| Steelmaking, batteries<br />
| South Africa; Gabon; Australia; China; India<br />
|-<br />
| B (Boron)<br />
| Glass and steel modifier<br />
| Turkey; USA; Argentina; Chile; Russia; China<br />
|-<br />
| F (Fluorspar)<br />
| Fluorochemicals, batteries<br />
| China; Mexico; Mongolia; South Africa; Vietnam<br />
|-<br />
| Li (Lithium)<br />
| Batteries, energy storage<br />
| Australia; Chile; Argentina; China; Zimbabwe; Canada; Brazil<br />
|-<br />
| Co (Cobalt)<br />
| Batteries, alloys<br />
| DR Congo; Indonesia; Russia; Australia; Philippines; Cuba<br />
|-<br />
| W (Tungsten)<br />
| Hardness, density<br />
| China; Vietnam; Russia; Rwanda; Spain; Austria; Bolivia; Portugal<br />
|-<br />
| Ga (Gallium)<br />
| RF and power electronics<br />
| China; Japan; South Korea; Russia<br />
|-<br />
| Nd/Pr/Dy<br />
| Permanent magnets<br />
| China; Australia; USA; Myanmar<br />
|}<br />
<br />
---<br />
<br />
== 5. Minimal RS Subsets by Civilization Type ==<br />
<br />
* '''Durable everyday civilization''' <br />
Cr, Ni, Ti, Mn, B<br />
<br />
* '''Electronics-heavy civilization''' <br />
Ga, Nd/Pr/Dy, Li, Co, Cu<br />
<br />
* '''Extreme density and hardness''' <br />
W<br />
<br />
---<br />
<br />
== 6. Examples of Global Information Infrastructure ==<br />
<br />
=== 6.1 Most Visited Websites (functional classification) ===<br />
<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messaging<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Adult content<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom] || Videoconferencing<br />
|-<br />
| [https://www.nytimes.com New York Times] || News media<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming platform<br />
|}<br />
<br />
---<br />
<br />
== 7. Open Source Trend (examples) ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] — image generation model<br />
<br />
---<br />
<br />
== 8. Scientific Discoveries Still Indispensable Today (Post-1700) ==<br />
<br />
=== 18th century ===<br />
* 1712–1781 — Steam engine<br />
* 1796 — Vaccination<br />
* Late 18th century — Modern chemistry<br />
<br />
=== 19th century ===<br />
* 1800 — Electric battery<br />
* 1824–1870 — Thermodynamics<br />
* 1831–1860 — Electromagnetism<br />
* 1846–1847 — Modern anesthesia<br />
* 1850–1880 — Germ theory and asepsis<br />
* 1865 — Genetics<br />
* 1869 — Periodic table<br />
* 1895 — X-rays<br />
* 1897 — Electron<br />
<br />
=== 20th century ===<br />
* 1900–1930 — Quantum mechanics<br />
* 1905–1915 — Relativity<br />
* 1909–1913 — Haber–Bosch process<br />
* 1928 — Antibiotics<br />
* 1947 — Transistor<br />
* 1948 — Information theory<br />
* 1953 — DNA structure<br />
* 1958–1959 — Integrated circuit<br />
* 1960 — Laser<br />
* 1960s–1980s — Internet<br />
* 1970s — Modern cryptography<br />
* 1983 — PCR<br />
<br />
=== 21st century ===<br />
* 2012 — CRISPR<br />
* 2010s — Deep learning<br />
<br />
---<br />
<br />
== 9. Notes and Limits ==<br />
* Refining is often more critical than extraction<br />
* Some RS elements can be substituted at performance cost<br />
* Recycling partially reduces RS dependency but does not eliminate it</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1917
Human needs
2026-01-31T15:58:48Z
<p>Tcepo: </p>
<hr />
<div>* '''[https://en.wikipedia.org/wiki/Shelter Shelter]''' : g1[[https://en.wikipedia.org/wiki/Particle Particle], [https://en.wikipedia.org/wiki/Motion Motion], [https://en.wikipedia.org/wiki/Fundamental_interaction Fundamental interaction], [https://en.wikipedia.org/wiki/Science Science]]<br />
* '''Structured elementary particles''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Clothes Clothes]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Toilet Toilet]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cleaner Cleaner]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transportation Transportation]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Defense Defense]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Help Help]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transducer Information transducer]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Information_processing Information processing]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Data_storage Information storer]''' (like disks, like books) : g1<br />
* '''[https://en.wikipedia.org/wiki/Search_engine Information seeker]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transmitter Information transmitter]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Stimulants Stimulants]''' (like structured elementary particles, like love, like entertainment...) : g1<br />
* '''[https://en.wikipedia.org/wiki/Particle Particle]'''<br />
* '''[https://en.wikipedia.org/wiki/Leisure Leisure]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Training Training]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cutlery Cutlery]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Mattress Mattress]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Table (furniture)]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Chair Chair]''' : g1<br />
<br><br />
Legend<br />
<br />
AU = Abundant / Universal elements<br />
(CHON + Si, Al, Fe, Ca, Mg, Na, K + S, P, Cl + Cu, Zn)<br />
<br />
RS = Rare / Strategic elements<br />
(Cr, Ni, Ti, Mn, B, F, Li + Nd/Pr/Dy, Co, W, Ag, Au, Ga, As…)<br />
<br />
Score: 0–5 (5 = critical, 0 = negligible)<br />
<br><br />
{| class="wikitable sortable"<br />
! Item (g1)<br />
! AU (elements + score)<br />
! RS (elements + score)<br />
! Notes (one line)<br />
|-<br />
| Shelter<br />
| Si,O,Al,Ca,Fe,Mg,C,H (5)<br />
| Cr,Ni,Ti,B (3)<br />
| Concrete/glass/ceramics + steel; alloys if available<br />
|-<br />
| Structured elementary particles (material structures)<br />
| C,H,O,N,Si,Fe,Al (5)<br />
| Ti,Cr,Ni,B (2)<br />
| Robust structures without high tech<br />
|-<br />
| Clothes<br />
| C,H,O,N,S (5)<br />
| F,Ti (2)<br />
| Fibers + treatments; F for high-performance membranes<br />
|-<br />
| Toilet<br />
| Si,O,Al,Ca,Fe,C,H (5)<br />
| Cr,Ni,Cu (3)<br />
| Ceramics + plumbing; stainless steel preferred<br />
|-<br />
| Cleaner<br />
| H,O,Na,Cl,C,S (5)<br />
| F,P (2)<br />
| Bases/salts/solvents; P for advanced detergents<br />
|-<br />
| Transportation<br />
| Fe,Al,Si,C,O,Cu (5)<br />
| Ni,Cr,Ti,Mn,Li (4)<br />
| Alloys + wiring; Li for electric systems<br />
|-<br />
| Defense<br />
| Fe,C,Al,Si (4)<br />
| Ti,Cr,Ni,W (4)<br />
| W useful for high density (rare)<br />
|-<br />
| Help (tools/basic care)<br />
| Fe,C,H,O,N,Ca,Na,Cl (5)<br />
| Cu,Zn,Ag (2)<br />
| Hygiene + tools; Ag optional antimicrobial<br />
|-<br />
| Information transducer<br />
| Si,O,Al,Fe,C,Cu (4)<br />
| B,Ga (4)<br />
| Sensors/actuators need dopants/semiconductors<br />
|-<br />
| Information processing<br />
| Si,O,Al,C,Cu (4)<br />
| B,P,As,Ga (5)<br />
| Modern computing = doping + fine processes<br />
|-<br />
| Data storage (disks/books)<br />
| C,H,O (paper), Fe (magnetic), Si,O (glass) (4)<br />
| Co,Ni,Nd (4)<br />
| High-end magnetic storage needs Co/Ni + Nd magnets<br />
|-<br />
| Information seeker (search engine)<br />
| Si,Al,Cu,C (4)<br />
| Li,Ni,Co,Nd (4)<br />
| Compute + energy + interconnects<br />
|-<br />
| Information transmitter<br />
| Cu,Al,Si,O (5)<br />
| Ag,Au (2)<br />
| Cu/Al + SiO2 fiber sufficient; nobles for reliability<br />
|-<br />
| Stimulants (broad sense)<br />
| C,H,O,N,S,P (5)<br />
| — (0)<br />
| Organic chemistry + nutrition; rares add little<br />
|-<br />
| Leisure<br />
| C,H,O,N,Si,Fe,Al (4)<br />
| Cu,Li,Nd (2)<br />
| Simple materials; electronics optional<br />
|-<br />
| Training<br />
| C,H,O,N,Si (4)<br />
| Cu,Li (2)<br />
| Media + tools; electronics optional<br />
|-<br />
| Cutlery<br />
| Fe,C (5)<br />
| Cr,Ni (4)<br />
| Stainless steel is a major practical upgrade<br />
|-<br />
| Mattress<br />
| C,H,O,N (5)<br />
| Si (1)<br />
| Foams/latex/fibers; silicones optional<br />
|-<br />
| Table (furniture)<br />
| C,H,O (wood/polymers) or Fe/Al or Si,O (glass) (5)<br />
| Cr,Ni,Ti (1)<br />
| Easily built with abundant elements<br />
|-<br />
| Chair<br />
| C,H,O or Fe/Al (5)<br />
| Ti,Cr (1)<br />
| Same: abundants are sufficient<br />
|}<br />
<br />
<br />
=== Rare/Strategic elements (RS) – where to source (countries) ===<br />
<br />
;Legend<br />
* AU = Abundant/Universal (easy)<br />
* RS = Rare/Strategic (hard / concentrated supply)<br />
<br />
{| class="wikitable sortable"<br />
! RS element<br />
! Main use (why you care)<br />
! Main sourcing countries (today)<br />
|-<br />
| Cr (Chromium)<br />
| Stainless / corrosion resistance<br />
| South Africa (dominant); Kazakhstan, India, Turkey (notable)<br />
|-<br />
| Ni (Nickel)<br />
| Stainless + batteries<br />
| Indonesia (dominant); Philippines, Russia, Canada, Australia, New Caledonia, Brazil<br />
|-<br />
| Ti (Titanium minerals: ilmenite/rutile)<br />
| Light strong alloys, durable structures<br />
| China; Mozambique; South Africa; Australia; Canada; India<br />
|-<br />
| Mn (Manganese)<br />
| Steelmaking + some battery chemistries<br />
| South Africa (largest); Gabon; Australia; China; India<br />
|-<br />
| B (Boron)<br />
| Glass/ceramics; steel/glass modifiers<br />
| Turkey (very large); USA; Argentina; Chile; Russia; China<br />
|-<br />
| F (Fluorspar → fluorochemicals)<br />
| HF/fluoropolymers; batteries; refrigerants<br />
| China (largest); Mexico; Mongolia; South Africa; Vietnam<br />
|-<br />
| Li (Lithium)<br />
| Batteries / energy storage<br />
| Australia; Chile; Argentina; China; Zimbabwe; Canada; Brazil<br />
|-<br />
| Co (Cobalt)<br />
| Batteries / high-performance alloys<br />
| DR Congo (mining dominant); Indonesia (growing); Russia, Australia, Philippines, Cuba (notable); refining largely China<br />
|-<br />
| W (Tungsten)<br />
| High density + hardness (tools/defense)<br />
| China (dominant); Vietnam; Russia; Rwanda; Spain; Austria; Bolivia; Portugal<br />
|-<br />
| Ga (Gallium)<br />
| RF/power electronics (hard bottleneck)<br />
| China (overwhelming); Japan, South Korea, Russia (secondary)<br />
|-<br />
| Nd/Pr/Dy (Rare earths for magnets)<br />
| Motors/actuators/sensors; “high torque density”<br />
| China (largest); Australia; USA; Myanmar (notable); processing often China-centric<br />
|}<br />
<br />
=== Minimal RS subsets (pick what you build) ===<br />
<br />
;RS for “durable everyday civilization” (shelter + toilet + tools + transport)<br />
* Cr + Ni (stainless)<br />
* Ti (light/strong)<br />
* Mn (steel)<br />
* B (glass/steel modifier)<br />
<br />
;RS for “electronics-heavy” (processing + transducer + transmitter + seeker)<br />
* Ga (critical)<br />
* Nd/Pr/Dy (magnets)<br />
* Li (batteries)<br />
* Co (some chemistries)<br />
* (plus Cu from AU for wiring)<br />
<br />
;RS for “extreme density/hardness”<br />
* W<br />
<br />
<br />
==TOP WEBSITES (CATEGORY AWARE)==<br />
# https://en.wikipedia.org/wiki/List_of_most-visited_websites<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing service<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messenger<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Online gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming service<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Pornography<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming service<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting service<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom Communications] || Computers Electronics and Technology - Other<br />
|-<br />
| [https://www.nytimes.com The New York Times] || News & Media Publishers<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming and game development platform<br />
|}<br />
<br />
== Open Source trend ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] : Image generative<br />
<br />
<br />
== Scientific discoveries (after 1700) still indispensable today ==<br />
Chronological list of scientific discoveries and breakthroughs made after 1700 that are still directly used in modern civilization (energy, health, industry, digital technologies, agriculture, communications).<br />
<br />
=== 18th century ===<br />
* '''1712–1781''' – '''Steam engine''' : foundation of thermal engines, turbines, and energy engineering.<br />
* '''1796''' – '''Vaccination''' : large-scale prevention of infectious diseases.<br />
* '''Late 18th century''' – '''Modern chemistry''' (conservation of matter, chemical nomenclature) : chemical industry, materials, analytical science.<br />
<br />
=== 19th century ===<br />
* '''1800''' – '''Electric battery''' : batteries, electrochemistry, energy storage.<br />
* '''1824–1870''' – '''Thermodynamics''' : engines, power plants, refrigeration, energy efficiency.<br />
* '''1831–1860''' – '''Electromagnetism''' : electricity generation and distribution, telecommunications.<br />
* '''1846–1847''' – '''Modern anesthesia''' : safe and scalable surgery.<br />
* '''1850–1880''' – '''Germ theory + asepsis''' : hygiene, drinking water, modern hospitals.<br />
* '''1865''' – '''Genetics''' : heredity, selection, biotechnology.<br />
* '''1869''' – '''Periodic table of elements''' : chemistry, materials, pharmaceuticals.<br />
* '''1895''' – '''X-rays''' : non-invasive medical imaging.<br />
* '''1897''' – '''Discovery of the electron''' : foundation of electronics.<br />
<br />
=== First half of the 20th century ===<br />
* '''1900–1930''' – '''Quantum mechanics''' : semiconductors, lasers, sensors, modern chemistry.<br />
* '''1905–1915''' – '''Relativity''' : precise time synchronization (GPS, networks).<br />
* '''1909–1913''' – '''Haber–Bosch process''' : nitrogen fertilizers, global intensive agriculture.<br />
* '''1928''' – '''Antibiotics''' : treatment of bacterial infections.<br />
* '''1947''' – '''Transistor''' : basis of modern electronics and computing.<br />
* '''1948''' – '''Information theory''' : compression, error correction, digital communications.<br />
<br />
=== Second half of the 20th century ===<br />
* '''1953''' – '''Structure of DNA''' : molecular biology, modern medicine.<br />
* '''1958–1959''' – '''Integrated circuit''' : miniaturization, mass computing.<br />
* '''1960''' – '''Laser''' : fiber optics, industry, medicine, metrology.<br />
* '''1960s–1980s''' – '''Packet-switched networks → Internet''' : global communication.<br />
* '''1970s''' – '''Modern cryptography''' : secure digital communications.<br />
* '''1983''' – '''PCR (DNA amplification)''' : diagnostics, biotechnology, forensics.<br />
<br />
=== 21st century ===<br />
* '''2012''' – '''CRISPR''' : targeted gene editing.<br />
* '''2010s''' – '''Deep learning''' : vision, translation, optimization, digital assistants.<br />
<br />
=== Note ===<br />
This list includes only discoveries that are '''still indispensable today''' — removing any of them would cause immediate collapse of major sectors (energy, health, food, digital systems, communications, industry).<br />
<br />
<br />
ynotopec at gmail.com (mail me)</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Enigme&diff=1916
Enigme
2026-01-22T10:10:33Z
<p>Tcepo: /* 1. Quel est l'âge de Paul ? */</p>
<hr />
<div>= '''1. Quel est l'âge de Paul ? =<br />
<br />
Paul s'adresse à Pierre :<br />
<br />
"J'ai le double de l'âge que tu avais quand j'avais ton âge.<br />
Quand tu auras mon âge, ensemble nous aurons 72 ans."<br />
<br />
Quel âge a Paul ?<br />
Quel âge a Pierre ?<br />
<br />
* English :<br />
Paul addresses Peter:<br />
<br />
"I am twice as old as you were when I was your age. When you are my age, together we will be 72 years old."<br />
<br />
How old is Paul? How old is Peter?<br />
<br />
<br />
{| class="wikitable" style="text-align:center;"<br />
|+ Valeurs de X, Paul (P) et Pierre (R)<br />
! X !! Paul (P) !! Pierre (R)<br />
|-<br />
| 9 || 4 || 3<br />
|-<br />
| 18 || 8 || 6<br />
|-<br />
| 27 || 12 || 9<br />
|-<br />
| 36 || 16 || 12<br />
|-<br />
| 45 || 20 || 15<br />
|-<br />
| 54 || 24 || 18<br />
|-<br />
| 63 || 28 || 21<br />
|-<br />
| 72 || 32 || 24<br />
|-<br />
| 81 || 36 || 27<br />
|-<br />
| 90 || 40 || 30<br />
|-<br />
| 99 || 44 || 33<br />
|-<br />
| 108 || 48 || 36<br />
|-<br />
| 117 || 52 || 39<br />
|-<br />
| 126 || 56 || 42<br />
|-<br />
| 135 || 60 || 45<br />
|-<br />
| 144 || 64 || 48<br />
|-<br />
| 153 || 68 || 51<br />
|-<br />
| 162 || 72 || 54<br />
|-<br />
| 171 || 76 || 57<br />
|-<br />
| 180 || 80 || 60<br />
|-<br />
| 189 || 84 || 63<br />
|-<br />
| 198 || 88 || 66<br />
|-<br />
| 207 || 92 || 69<br />
|-<br />
| 216 || 96 || 72<br />
|-<br />
| 225 || 100 || 75<br />
|-<br />
| 234 || 104 || 78<br />
|-<br />
| 243 || 108 || 81<br />
|-<br />
| 252 || 112 || 84<br />
|-<br />
| 261 || 116 || 87<br />
|-<br />
| 270 || 120 || 90<br />
|-<br />
| 279 || 124 || 93<br />
|-<br />
| 288 || 128 || 96<br />
|-<br />
| 297 || 132 || 99<br />
|}<br />
<br />
<br />
<br />
'''2. Un lac étonnant!'''<br />
<br />
Un lac double de volume chaque jour.<br />
Au bout de 30 jours il est plein.<br />
<br />
Le lac est donc plein à la moitié au bout de combien de jours ?<br />
<br />
'''3. Les 3 prisionniers<br />
<br />
Trois prisonniers sont ensemble. Chacun portant une pancarte au dos tiré au hasard parmi 3 pancartes blanches et 2 noirs. <br />
<br />
Chaque prisonnier ne peut pas voir sa propre pancarte, mais peut voir celles des autres.<br />
<br />
Dans cette situation, les trois prisonniers reçoivent chacun une pancarte blanche.<br />
<br />
Ceux qui devinent la couleur de leur propre pancarte annoncent à tout le monde qu'ils ont trouvé et sont libérés.<br />
<br />
Comment est-il possible de deviner la couleur de sa propre pancarte dans cette situation ?<br />
<br />
* English :<br />
Three prisoners are together. Each of them wears a sign on their back, randomly drawn from a set of 3 white signs and 2 black signs.<br />
<br />
Each prisoner cannot see the color of their own sign but can see the signs of the others.<br />
<br />
In this situation, all three prisoners receive a white sign.<br />
<br />
The first prisoner who correctly guesses the color of their own sign raises their hand and is set free.<br />
<br />
How is it possible to determine the color of their own sign in this scenario?</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Enigme&diff=1915
Enigme
2026-01-22T10:04:06Z
<p>Tcepo: </p>
<hr />
<div>= '''1. Quel est l'âge de Paul ? =<br />
<br />
Paul s'adresse à Pierre :<br />
<br />
"J'ai le double de l'âge que tu avais quand j'avais ton âge.<br />
Quand tu auras mon âge, ensemble nous aurons 72 ans."<br />
<br />
Quel âge a Paul ?<br />
Quel âge a Pierre ?<br />
<br />
* English :<br />
Paul addresses Peter:<br />
<br />
"I am twice as old as you were when I was your age. When you are my age, together we will be 72 years old."<br />
<br />
How old is Paul? How old is Peter?<br />
<br />
<br />
== Solution ==<br />
<br />
Let <br />
<br />
* <math>P</math> = Paul’s current age <br />
* <math>Q</math> = Pierre’s current age <br />
* <math>X</math> = the total age “at the moment when Pierre has Paul’s age”<br />
<br />
---<br />
<br />
=== 1. From Paul’s first sentence ===<br />
<br />
''“I have double the age that you had when I was your age.”''<br />
<br />
When Paul was the same age as Pierre '''now''', Pierre’s age was:<br />
<br />
<math><br />
\text{Pierre’s age at that time} = Q - (P - Q) = 2Q - P<br />
</math><br />
<br />
Paul says his current age is twice that:<br />
<br />
<math><br />
P = 2(2Q - P)<br />
</math><br />
<br />
<math><br />
P = 4Q - 2P \Longrightarrow 3P = 4Q<br />
</math><br />
<br />
Hence:<br />
<br />
<math><br />
\boxed{P = \frac{4}{3}Q}<br />
</math><br />
<br />
So <math>Q</math> must be a multiple of 3.<br />
<br />
---<br />
<br />
=== 2. From Paul’s second sentence ===<br />
<br />
''“When you have my age, together we will have <math>X</math> years.”''<br />
<br />
Pierre will reach Paul’s current age in <math>P - Q</math> years. <br />
At that moment:<br />
<br />
* Pierre’s age = <math>P</math> <br />
* Paul’s age = <math>P + (P - Q) = 2P - Q</math><br />
<br />
So:<br />
<br />
<math><br />
X = P + (2P - Q) = 3P - Q<br />
</math><br />
<br />
Insert <math>P = \frac{4}{3}Q</math>:<br />
<br />
<math><br />
X = 3\left(\frac{4}{3}Q\right) - Q = 4Q - Q = 3Q<br />
</math><br />
<br />
Thus:<br />
<br />
<math><br />
\boxed{X = 3Q}<br />
</math><br />
<br />
And because <math>Q = 3k</math> (from the first part), we get:<br />
<br />
<math><br />
\boxed{Q = 3k,\; P = 4k,\; X = 9k} \quad (k \in \mathbb{Z}^+)<br />
</math><br />
<br />
---<br />
<br />
=== 3. Integers between 1 and 100 ===<br />
<br />
We need <math>X \le 100</math>:<br />
<br />
<math><br />
9k \le 100 \Longrightarrow k \le \left\lfloor \frac{100}{9} \right\rfloor = 11<br />
</math><br />
<br />
So <math>k = 1,2,\dots,11</math>.<br />
<br />
{| class="wikitable"<br />
! k !! Q (Pierre) !! P (Paul) !! X = 3Q<br />
|-<br />
| 1 || 3 || 4 || 9<br />
|-<br />
| 2 || 6 || 8 || 18<br />
|-<br />
| 3 || 9 || 12 || 27<br />
|-<br />
| 4 || 12 || 16 || 36<br />
|-<br />
| 5 || 15 || 20 || 45<br />
|-<br />
| 6 || 18 || 24 || 54<br />
|-<br />
| 7 || 21 || 28 || 63<br />
|-<br />
| 8 || 24 || 32 || 72<br />
|-<br />
| 9 || 27 || 36 || 81<br />
|-<br />
| 10 || 30 || 40 || 90<br />
|-<br />
| 11 || 33 || 44 || 99<br />
|}<br />
<br />
All ages are integers, Paul is older than Pierre (since <math>P = 4k > 3k = Q</math>), and the total <math>X</math> lies in the required range 1 to 100.<br />
<br />
---<br />
<br />
== Answer ==<br />
<br />
The possible values of <math>X</math> (1 ≤ X ≤ 100) that give integer ages for Paul and Pierre are:<br />
<br />
<math><br />
\boxed{X \in \{9, 18, 27, 36, 45, 54, 63, 72, 81, 90, 99\}}<br />
</math><br />
<br />
For each such <math>X = 9k</math> (with <math>k = 1</math> to <math>11</math>), the ages are:<br />
<br />
* Pierre: <math>Q = 3k</math> <br />
* Paul: <math>P = 4k</math><br />
<br />
<br />
* Avec somme future 81 : Paul a 36 ans, Pierre a 27 ans.<br />
<br />
<br />
'''2. Un lac étonnant!'''<br />
<br />
Un lac double de volume chaque jour.<br />
Au bout de 30 jours il est plein.<br />
<br />
Le lac est donc plein à la moitié au bout de combien de jours ?<br />
<br />
'''3. Les 3 prisionniers<br />
<br />
Trois prisonniers sont ensemble. Chacun portant une pancarte au dos tiré au hasard parmi 3 pancartes blanches et 2 noirs. <br />
<br />
Chaque prisonnier ne peut pas voir sa propre pancarte, mais peut voir celles des autres.<br />
<br />
Dans cette situation, les trois prisonniers reçoivent chacun une pancarte blanche.<br />
<br />
Ceux qui devinent la couleur de leur propre pancarte annoncent à tout le monde qu'ils ont trouvé et sont libérés.<br />
<br />
Comment est-il possible de deviner la couleur de sa propre pancarte dans cette situation ?<br />
<br />
* English :<br />
Three prisoners are together. Each of them wears a sign on their back, randomly drawn from a set of 3 white signs and 2 black signs.<br />
<br />
Each prisoner cannot see the color of their own sign but can see the signs of the others.<br />
<br />
In this situation, all three prisoners receive a white sign.<br />
<br />
The first prisoner who correctly guesses the color of their own sign raises their hand and is set free.<br />
<br />
How is it possible to determine the color of their own sign in this scenario?</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Enigme&diff=1914
Enigme
2026-01-22T10:03:27Z
<p>Tcepo: </p>
<hr />
<div>= '''1. Quel est l'âge de Paul ? =<br />
<br />
Paul s'adresse à Pierre :<br />
<br />
"J'ai le double de l'âge que tu avais quand j'avais ton âge.<br />
Quand tu auras mon âge, ensemble nous aurons 72 ans."<br />
<br />
Quel âge a Paul ?<br />
Quel âge a Pierre ?<br />
<br />
* English :<br />
Paul addresses Peter:<br />
<br />
"I am twice as old as you were when I was your age. When you are my age, together we will be 72 years old."<br />
<br />
How old is Paul? How old is Peter?<br />
<br />
<br />
Voici la conversion complète au **format MediaWiki** :<br />
<br />
```mediawiki<br />
== Solution ==<br />
<br />
Let <br />
<br />
* <math>P</math> = Paul’s current age <br />
* <math>Q</math> = Pierre’s current age <br />
* <math>X</math> = the total age “at the moment when Pierre has Paul’s age”<br />
<br />
---<br />
<br />
=== 1. From Paul’s first sentence ===<br />
<br />
''“I have double the age that you had when I was your age.”''<br />
<br />
When Paul was the same age as Pierre '''now''', Pierre’s age was:<br />
<br />
<math><br />
\text{Pierre’s age at that time} = Q - (P - Q) = 2Q - P<br />
</math><br />
<br />
Paul says his current age is twice that:<br />
<br />
<math><br />
P = 2(2Q - P)<br />
</math><br />
<br />
<math><br />
P = 4Q - 2P \Longrightarrow 3P = 4Q<br />
</math><br />
<br />
Hence:<br />
<br />
<math><br />
\boxed{P = \frac{4}{3}Q}<br />
</math><br />
<br />
So <math>Q</math> must be a multiple of 3.<br />
<br />
---<br />
<br />
=== 2. From Paul’s second sentence ===<br />
<br />
''“When you have my age, together we will have <math>X</math> years.”''<br />
<br />
Pierre will reach Paul’s current age in <math>P - Q</math> years. <br />
At that moment:<br />
<br />
* Pierre’s age = <math>P</math> <br />
* Paul’s age = <math>P + (P - Q) = 2P - Q</math><br />
<br />
So:<br />
<br />
<math><br />
X = P + (2P - Q) = 3P - Q<br />
</math><br />
<br />
Insert <math>P = \frac{4}{3}Q</math>:<br />
<br />
<math><br />
X = 3\left(\frac{4}{3}Q\right) - Q = 4Q - Q = 3Q<br />
</math><br />
<br />
Thus:<br />
<br />
<math><br />
\boxed{X = 3Q}<br />
</math><br />
<br />
And because <math>Q = 3k</math> (from the first part), we get:<br />
<br />
<math><br />
\boxed{Q = 3k,\; P = 4k,\; X = 9k} \quad (k \in \mathbb{Z}^+)<br />
</math><br />
<br />
---<br />
<br />
=== 3. Integers between 1 and 100 ===<br />
<br />
We need <math>X \le 100</math>:<br />
<br />
<math><br />
9k \le 100 \Longrightarrow k \le \left\lfloor \frac{100}{9} \right\rfloor = 11<br />
</math><br />
<br />
So <math>k = 1,2,\dots,11</math>.<br />
<br />
{| class="wikitable"<br />
! k !! Q (Pierre) !! P (Paul) !! X = 3Q<br />
|-<br />
| 1 || 3 || 4 || 9<br />
|-<br />
| 2 || 6 || 8 || 18<br />
|-<br />
| 3 || 9 || 12 || 27<br />
|-<br />
| 4 || 12 || 16 || 36<br />
|-<br />
| 5 || 15 || 20 || 45<br />
|-<br />
| 6 || 18 || 24 || 54<br />
|-<br />
| 7 || 21 || 28 || 63<br />
|-<br />
| 8 || 24 || 32 || 72<br />
|-<br />
| 9 || 27 || 36 || 81<br />
|-<br />
| 10 || 30 || 40 || 90<br />
|-<br />
| 11 || 33 || 44 || 99<br />
|}<br />
<br />
All ages are integers, Paul is older than Pierre (since <math>P = 4k > 3k = Q</math>), and the total <math>X</math> lies in the required range 1 to 100.<br />
<br />
---<br />
<br />
== Answer ==<br />
<br />
The possible values of <math>X</math> (1 ≤ X ≤ 100) that give integer ages for Paul and Pierre are:<br />
<br />
<math><br />
\boxed{X \in \{9, 18, 27, 36, 45, 54, 63, 72, 81, 90, 99\}}<br />
</math><br />
<br />
For each such <math>X = 9k</math> (with <math>k = 1</math> to <math>11</math>), the ages are:<br />
<br />
* Pierre: <math>Q = 3k</math> <br />
* Paul: <math>P = 4k</math><br />
```<br />
<br />
<br />
<br />
* Avec somme future 81 : Paul a 36 ans, Pierre a 27 ans.<br />
<br />
<br />
'''2. Un lac étonnant!'''<br />
<br />
Un lac double de volume chaque jour.<br />
Au bout de 30 jours il est plein.<br />
<br />
Le lac est donc plein à la moitié au bout de combien de jours ?<br />
<br />
'''3. Les 3 prisionniers<br />
<br />
Trois prisonniers sont ensemble. Chacun portant une pancarte au dos tiré au hasard parmi 3 pancartes blanches et 2 noirs. <br />
<br />
Chaque prisonnier ne peut pas voir sa propre pancarte, mais peut voir celles des autres.<br />
<br />
Dans cette situation, les trois prisonniers reçoivent chacun une pancarte blanche.<br />
<br />
Ceux qui devinent la couleur de leur propre pancarte annoncent à tout le monde qu'ils ont trouvé et sont libérés.<br />
<br />
Comment est-il possible de deviner la couleur de sa propre pancarte dans cette situation ?<br />
<br />
* English :<br />
Three prisoners are together. Each of them wears a sign on their back, randomly drawn from a set of 3 white signs and 2 black signs.<br />
<br />
Each prisoner cannot see the color of their own sign but can see the signs of the others.<br />
<br />
In this situation, all three prisoners receive a white sign.<br />
<br />
The first prisoner who correctly guesses the color of their own sign raises their hand and is set free.<br />
<br />
How is it possible to determine the color of their own sign in this scenario?</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1913
Human needs
2026-01-15T17:30:29Z
<p>Tcepo: </p>
<hr />
<div>* '''[https://en.wikipedia.org/wiki/Shelter Shelter]''' : g1[[https://en.wikipedia.org/wiki/Particle Particle], [https://en.wikipedia.org/wiki/Motion Motion], [https://en.wikipedia.org/wiki/Fundamental_interaction Fundamental interaction], [https://en.wikipedia.org/wiki/Science Science]]<br />
* '''Structured elementary particles''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Clothes Clothes]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Toilet Toilet]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cleaner Cleaner]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transportation Transportation]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Defense Defense]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Help Help]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transducer Information transducer]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Information_processing Information processing]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Data_storage Information storer]''' (like disks, like books) : g1<br />
* '''[https://en.wikipedia.org/wiki/Search_engine Information seeker]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transmitter Information transmitter]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Stimulants Stimulants]''' (like structured elementary particles, like love, like entertainment...) : g1<br />
* '''[https://en.wikipedia.org/wiki/Particle Particle]'''<br />
* '''[https://en.wikipedia.org/wiki/Leisure Leisure]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Training Training]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cutlery Cutlery]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Mattress Mattress]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Table (furniture)]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Chair Chair]''' : g1<br />
<br><br />
Legend<br />
<br />
AU = Abundant / Universal elements<br />
(CHON + Si, Al, Fe, Ca, Mg, Na, K + S, P, Cl + Cu, Zn)<br />
<br />
RS = Rare / Strategic elements<br />
(Cr, Ni, Ti, Mn, B, F, Li + Nd/Pr/Dy, Co, W, Ag, Au, Ga, As…)<br />
<br />
Score: 0–5 (5 = critical, 0 = negligible)<br />
<br><br />
{| class="wikitable sortable"<br />
! Item (g1)<br />
! AU (elements + score)<br />
! RS (elements + score)<br />
! Notes (one line)<br />
|-<br />
| Shelter<br />
| Si,O,Al,Ca,Fe,Mg,C,H (5)<br />
| Cr,Ni,Ti,B (3)<br />
| Concrete/glass/ceramics + steel; alloys if available<br />
|-<br />
| Structured elementary particles (material structures)<br />
| C,H,O,N,Si,Fe,Al (5)<br />
| Ti,Cr,Ni,B (2)<br />
| Robust structures without high tech<br />
|-<br />
| Clothes<br />
| C,H,O,N,S (5)<br />
| F,Ti (2)<br />
| Fibers + treatments; F for high-performance membranes<br />
|-<br />
| Toilet<br />
| Si,O,Al,Ca,Fe,C,H (5)<br />
| Cr,Ni,Cu (3)<br />
| Ceramics + plumbing; stainless steel preferred<br />
|-<br />
| Cleaner<br />
| H,O,Na,Cl,C,S (5)<br />
| F,P (2)<br />
| Bases/salts/solvents; P for advanced detergents<br />
|-<br />
| Transportation<br />
| Fe,Al,Si,C,O,Cu (5)<br />
| Ni,Cr,Ti,Mn,Li (4)<br />
| Alloys + wiring; Li for electric systems<br />
|-<br />
| Defense<br />
| Fe,C,Al,Si (4)<br />
| Ti,Cr,Ni,W (4)<br />
| W useful for high density (rare)<br />
|-<br />
| Help (tools/basic care)<br />
| Fe,C,H,O,N,Ca,Na,Cl (5)<br />
| Cu,Zn,Ag (2)<br />
| Hygiene + tools; Ag optional antimicrobial<br />
|-<br />
| Information transducer<br />
| Si,O,Al,Fe,C,Cu (4)<br />
| B,Ga (4)<br />
| Sensors/actuators need dopants/semiconductors<br />
|-<br />
| Information processing<br />
| Si,O,Al,C,Cu (4)<br />
| B,P,As,Ga (5)<br />
| Modern computing = doping + fine processes<br />
|-<br />
| Data storage (disks/books)<br />
| C,H,O (paper), Fe (magnetic), Si,O (glass) (4)<br />
| Co,Ni,Nd (4)<br />
| High-end magnetic storage needs Co/Ni + Nd magnets<br />
|-<br />
| Information seeker (search engine)<br />
| Si,Al,Cu,C (4)<br />
| Li,Ni,Co,Nd (4)<br />
| Compute + energy + interconnects<br />
|-<br />
| Information transmitter<br />
| Cu,Al,Si,O (5)<br />
| Ag,Au (2)<br />
| Cu/Al + SiO2 fiber sufficient; nobles for reliability<br />
|-<br />
| Stimulants (broad sense)<br />
| C,H,O,N,S,P (5)<br />
| — (0)<br />
| Organic chemistry + nutrition; rares add little<br />
|-<br />
| Leisure<br />
| C,H,O,N,Si,Fe,Al (4)<br />
| Cu,Li,Nd (2)<br />
| Simple materials; electronics optional<br />
|-<br />
| Training<br />
| C,H,O,N,Si (4)<br />
| Cu,Li (2)<br />
| Media + tools; electronics optional<br />
|-<br />
| Cutlery<br />
| Fe,C (5)<br />
| Cr,Ni (4)<br />
| Stainless steel is a major practical upgrade<br />
|-<br />
| Mattress<br />
| C,H,O,N (5)<br />
| Si (1)<br />
| Foams/latex/fibers; silicones optional<br />
|-<br />
| Table (furniture)<br />
| C,H,O (wood/polymers) or Fe/Al or Si,O (glass) (5)<br />
| Cr,Ni,Ti (1)<br />
| Easily built with abundant elements<br />
|-<br />
| Chair<br />
| C,H,O or Fe/Al (5)<br />
| Ti,Cr (1)<br />
| Same: abundants are sufficient<br />
|}<br />
<br />
<br />
=== Rare/Strategic elements (RS) – where to source (countries) ===<br />
<br />
;Legend<br />
* AU = Abundant/Universal (easy)<br />
* RS = Rare/Strategic (hard / concentrated supply)<br />
<br />
{| class="wikitable sortable"<br />
! RS element<br />
! Main use (why you care)<br />
! Main sourcing countries (today)<br />
|-<br />
| Cr (Chromium)<br />
| Stainless / corrosion resistance<br />
| South Africa (dominant); Kazakhstan, India, Turkey (notable)<br />
|-<br />
| Ni (Nickel)<br />
| Stainless + batteries<br />
| Indonesia (dominant); Philippines, Russia, Canada, Australia, New Caledonia, Brazil<br />
|-<br />
| Ti (Titanium minerals: ilmenite/rutile)<br />
| Light strong alloys, durable structures<br />
| China; Mozambique; South Africa; Australia; Canada; India<br />
|-<br />
| Mn (Manganese)<br />
| Steelmaking + some battery chemistries<br />
| South Africa (largest); Gabon; Australia; China; India<br />
|-<br />
| B (Boron)<br />
| Glass/ceramics; steel/glass modifiers<br />
| Turkey (very large); USA; Argentina; Chile; Russia; China<br />
|-<br />
| F (Fluorspar → fluorochemicals)<br />
| HF/fluoropolymers; batteries; refrigerants<br />
| China (largest); Mexico; Mongolia; South Africa; Vietnam<br />
|-<br />
| Li (Lithium)<br />
| Batteries / energy storage<br />
| Australia; Chile; Argentina; China; Zimbabwe; Canada; Brazil<br />
|-<br />
| Co (Cobalt)<br />
| Batteries / high-performance alloys<br />
| DR Congo (mining dominant); Indonesia (growing); Russia, Australia, Philippines, Cuba (notable); refining largely China<br />
|-<br />
| W (Tungsten)<br />
| High density + hardness (tools/defense)<br />
| China (dominant); Vietnam; Russia; Rwanda; Spain; Austria; Bolivia; Portugal<br />
|-<br />
| Ga (Gallium)<br />
| RF/power electronics (hard bottleneck)<br />
| China (overwhelming); Japan, South Korea, Russia (secondary)<br />
|-<br />
| Nd/Pr/Dy (Rare earths for magnets)<br />
| Motors/actuators/sensors; “high torque density”<br />
| China (largest); Australia; USA; Myanmar (notable); processing often China-centric<br />
|}<br />
<br />
=== Minimal RS subsets (pick what you build) ===<br />
<br />
;RS for “durable everyday civilization” (shelter + toilet + tools + transport)<br />
* Cr + Ni (stainless)<br />
* Ti (light/strong)<br />
* Mn (steel)<br />
* B (glass/steel modifier)<br />
<br />
;RS for “electronics-heavy” (processing + transducer + transmitter + seeker)<br />
* Ga (critical)<br />
* Nd/Pr/Dy (magnets)<br />
* Li (batteries)<br />
* Co (some chemistries)<br />
* (plus Cu from AU for wiring)<br />
<br />
;RS for “extreme density/hardness”<br />
* W<br />
<br />
<br />
==TOP WEBSITES (CATEGORY AWARE)==<br />
# https://en.wikipedia.org/wiki/List_of_most-visited_websites<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing service<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messenger<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Online gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming service<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Pornography<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming service<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting service<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom Communications] || Computers Electronics and Technology - Other<br />
|-<br />
| [https://www.nytimes.com The New York Times] || News & Media Publishers<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming and game development platform<br />
|}<br />
<br />
== Open Source trend ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] : Image generative<br />
<br />
ynotopec at gmail.com (mail me)</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1912
Human needs
2026-01-15T17:25:13Z
<p>Tcepo: </p>
<hr />
<div>* '''[https://en.wikipedia.org/wiki/Shelter Shelter]''' : g1[[https://en.wikipedia.org/wiki/Particle Particle], [https://en.wikipedia.org/wiki/Motion Motion], [https://en.wikipedia.org/wiki/Fundamental_interaction Fundamental interaction], [https://en.wikipedia.org/wiki/Science Science]]<br />
* '''Structured elementary particles''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Clothes Clothes]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Toilet Toilet]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cleaner Cleaner]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transportation Transportation]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Defense Defense]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Help Help]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transducer Information transducer]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Information_processing Information processing]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Data_storage Information storer]''' (like disks, like books) : g1<br />
* '''[https://en.wikipedia.org/wiki/Search_engine Information seeker]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transmitter Information transmitter]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Stimulants Stimulants]''' (like structured elementary particles, like love, like entertainment...) : g1<br />
* '''[https://en.wikipedia.org/wiki/Particle Particle]'''<br />
* '''[https://en.wikipedia.org/wiki/Leisure Leisure]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Training Training]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cutlery Cutlery]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Mattress Mattress]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Table (furniture)]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Chair Chair]''' : g1<br />
<br><br />
Legend<br />
<br />
AU = Abundant / Universal elements<br />
(CHON + Si, Al, Fe, Ca, Mg, Na, K + S, P, Cl + Cu, Zn)<br />
<br />
RS = Rare / Strategic elements<br />
(Cr, Ni, Ti, Mn, B, F, Li + Nd/Pr/Dy, Co, W, Ag, Au, Ga, As…)<br />
<br />
Score: 0–5 (5 = critical, 0 = negligible)<br />
<br><br />
{| class="wikitable sortable"<br />
! Item (g1)<br />
! AU (elements + score)<br />
! RS (elements + score)<br />
! Notes (one line)<br />
|-<br />
| Shelter<br />
| Si,O,Al,Ca,Fe,Mg,C,H (5)<br />
| Cr,Ni,Ti,B (3)<br />
| Concrete/glass/ceramics + steel; alloys if available<br />
|-<br />
| Structured elementary particles (material structures)<br />
| C,H,O,N,Si,Fe,Al (5)<br />
| Ti,Cr,Ni,B (2)<br />
| Robust structures without high tech<br />
|-<br />
| Clothes<br />
| C,H,O,N,S (5)<br />
| F,Ti (2)<br />
| Fibers + treatments; F for high-performance membranes<br />
|-<br />
| Toilet<br />
| Si,O,Al,Ca,Fe,C,H (5)<br />
| Cr,Ni,Cu (3)<br />
| Ceramics + plumbing; stainless steel preferred<br />
|-<br />
| Cleaner<br />
| H,O,Na,Cl,C,S (5)<br />
| F,P (2)<br />
| Bases/salts/solvents; P for advanced detergents<br />
|-<br />
| Transportation<br />
| Fe,Al,Si,C,O,Cu (5)<br />
| Ni,Cr,Ti,Mn,Li (4)<br />
| Alloys + wiring; Li for electric systems<br />
|-<br />
| Defense<br />
| Fe,C,Al,Si (4)<br />
| Ti,Cr,Ni,W (4)<br />
| W useful for high density (rare)<br />
|-<br />
| Help (tools/basic care)<br />
| Fe,C,H,O,N,Ca,Na,Cl (5)<br />
| Cu,Zn,Ag (2)<br />
| Hygiene + tools; Ag optional antimicrobial<br />
|-<br />
| Information transducer<br />
| Si,O,Al,Fe,C,Cu (4)<br />
| B,Ga (4)<br />
| Sensors/actuators need dopants/semiconductors<br />
|-<br />
| Information processing<br />
| Si,O,Al,C,Cu (4)<br />
| B,P,As,Ga (5)<br />
| Modern computing = doping + fine processes<br />
|-<br />
| Data storage (disks/books)<br />
| C,H,O (paper), Fe (magnetic), Si,O (glass) (4)<br />
| Co,Ni,Nd (4)<br />
| High-end magnetic storage needs Co/Ni + Nd magnets<br />
|-<br />
| Information seeker (search engine)<br />
| Si,Al,Cu,C (4)<br />
| Li,Ni,Co,Nd (4)<br />
| Compute + energy + interconnects<br />
|-<br />
| Information transmitter<br />
| Cu,Al,Si,O (5)<br />
| Ag,Au (2)<br />
| Cu/Al + SiO2 fiber sufficient; nobles for reliability<br />
|-<br />
| Stimulants (broad sense)<br />
| C,H,O,N,S,P (5)<br />
| — (0)<br />
| Organic chemistry + nutrition; rares add little<br />
|-<br />
| Leisure<br />
| C,H,O,N,Si,Fe,Al (4)<br />
| Cu,Li,Nd (2)<br />
| Simple materials; electronics optional<br />
|-<br />
| Training<br />
| C,H,O,N,Si (4)<br />
| Cu,Li (2)<br />
| Media + tools; electronics optional<br />
|-<br />
| Cutlery<br />
| Fe,C (5)<br />
| Cr,Ni (4)<br />
| Stainless steel is a major practical upgrade<br />
|-<br />
| Mattress<br />
| C,H,O,N (5)<br />
| Si (1)<br />
| Foams/latex/fibers; silicones optional<br />
|-<br />
| Table (furniture)<br />
| C,H,O (wood/polymers) or Fe/Al or Si,O (glass) (5)<br />
| Cr,Ni,Ti (1)<br />
| Easily built with abundant elements<br />
|-<br />
| Chair<br />
| C,H,O or Fe/Al (5)<br />
| Ti,Cr (1)<br />
| Same: abundants are sufficient<br />
|}<br />
<br />
<br />
<br />
==TOP WEBSITES (CATEGORY AWARE)==<br />
# https://en.wikipedia.org/wiki/List_of_most-visited_websites<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing service<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messenger<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Online gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming service<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Pornography<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming service<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting service<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom Communications] || Computers Electronics and Technology - Other<br />
|-<br />
| [https://www.nytimes.com The New York Times] || News & Media Publishers<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming and game development platform<br />
|}<br />
<br />
== Open Source trend ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] : Image generative<br />
<br />
ynotopec at gmail.com (mail me)</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1911
Human needs
2026-01-15T17:23:18Z
<p>Tcepo: </p>
<hr />
<div>* '''[https://en.wikipedia.org/wiki/Shelter Shelter]''' : g1[[https://en.wikipedia.org/wiki/Particle Particle], [https://en.wikipedia.org/wiki/Motion Motion], [https://en.wikipedia.org/wiki/Fundamental_interaction Fundamental interaction], [https://en.wikipedia.org/wiki/Science Science]]<br />
* '''Structured elementary particles''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Clothes Clothes]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Toilet Toilet]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cleaner Cleaner]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transportation Transportation]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Defense Defense]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Help Help]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transducer Information transducer]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Information_processing Information processing]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Data_storage Information storer]''' (like disks, like books) : g1<br />
* '''[https://en.wikipedia.org/wiki/Search_engine Information seeker]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transmitter Information transmitter]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Stimulants Stimulants]''' (like structured elementary particles, like love, like entertainment...) : g1<br />
* '''[https://en.wikipedia.org/wiki/Particle Particle]'''<br />
* '''[https://en.wikipedia.org/wiki/Leisure Leisure]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Training Training]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cutlery Cutlery]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Mattress Mattress]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Table (furniture)]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Chair Chair]''' : g1<br />
<br><br />
{| class="wikitable sortable"<br />
! Item (g1)<br />
! AU (éléments + score)<br />
! RS (éléments + score)<br />
! Notes (1 ligne)<br />
|-<br />
| Shelter<br />
| Si,O,Al,Ca,Fe,Mg,C,H (5)<br />
| Cr,Ni,Ti,B (3)<br />
| Béton/verre/céramique + acier; inox/alliages si besoin<br />
|-<br />
| Structured elementary particles (structures matérielles)<br />
| C,H,O,N,Si,Fe,Al (5)<br />
| Ti,Cr,Ni,B (2)<br />
| Structures robustes sans haute techno<br />
|-<br />
| Clothes<br />
| C,H,O,N,S (5)<br />
| F,Ti (2)<br />
| Fibres + traitements; F pour membranes “haute perf”<br />
|-<br />
| Toilet<br />
| Si,O,Al,Ca,Fe,C,H (5)<br />
| Cr,Ni,Cu (3)<br />
| Céramique + plomberie; inox recommandé<br />
|-<br />
| Cleaner<br />
| H,O,Na,Cl,C,S (5)<br />
| F,P (2)<br />
| Bases/sels/solvants; P pour détergents avancés<br />
|-<br />
| Transportation<br />
| Fe,Al,Si,C,O,Cu (5)<br />
| Ni,Cr,Ti,Mn,Li (4)<br />
| Alliages + câbles; Li si électrique<br />
|-<br />
| Defense<br />
| Fe,C,Al,Si (4)<br />
| Ti,Cr,Ni,W (4)<br />
| W utile si haute densité (rare)<br />
|-<br />
| Help (outillage/soins de base)<br />
| Fe,C,H,O,N,Ca,Na,Cl (5)<br />
| Cu,Zn,Ag (2)<br />
| Hygiène + outils; Ag optionnel antimicrobien<br />
|-<br />
| Information transducer<br />
| Si,O,Al,Fe,C,Cu (4)<br />
| B,Ga (4)<br />
| Capteurs/actuateurs avancés demandent dopants/semiconducteurs<br />
|-<br />
| Information processing<br />
| Si,O,Al,C,Cu (4)<br />
| B,P,As,Ga (5)<br />
| Calcul moderne = dopage + procédés fins<br />
|-<br />
| Data storage (disks/books)<br />
| C,H,O (papier), Fe (magnétique), Si,O (verre) (4)<br />
| Co,Ni,Nd (4)<br />
| Magnétique “haut niveau” = Co/Ni + aimants Nd<br />
|-<br />
| Information seeker (search engine)<br />
| Si,Al,Cu,C (4)<br />
| Li,Ni,Co,Nd (4)<br />
| Besoin de compute + énergie + interconnexions<br />
|-<br />
| Information transmitter<br />
| Cu,Al,Si,O (5)<br />
| Ag,Au (2)<br />
| Cu/Al + fibre (SiO2) suffisent; nobles pour fiabilité<br />
|-<br />
| Stimulants (au sens large)<br />
| C,H,O,N,S,P (5)<br />
| — (0)<br />
| Chimie organique + nutrition; le rare n’apporte pas grand-chose<br />
|-<br />
| Leisure<br />
| C,H,O,N,Si,Fe,Al (4)<br />
| Cu,Li,Nd (2)<br />
| Matériaux simples; électronique optionnelle<br />
|-<br />
| Training<br />
| C,H,O,N,Si (4)<br />
| Cu,Li (2)<br />
| Supports + outils; électronique optionnelle<br />
|-<br />
| Cutlery<br />
| Fe,C (5)<br />
| Cr,Ni (4)<br />
| Inox = énorme gain pratique<br />
|-<br />
| Mattress<br />
| C,H,O,N (5)<br />
| Si (1)<br />
| Mousses/latex/fibres; silicones optionnels<br />
|-<br />
| Table (furniture)<br />
| C,H,O (bois/polymères) ou Fe/Al ou Si,O (verre) (5)<br />
| Cr,Ni,Ti (1)<br />
| Très faisable avec abondants<br />
|-<br />
| Chair<br />
| C,H,O ou Fe/Al (5)<br />
| Ti,Cr (1)<br />
| Pareil : abondants suffisent<br />
|}<br />
<br />
<br />
<br />
==TOP WEBSITES (CATEGORY AWARE)==<br />
# https://en.wikipedia.org/wiki/List_of_most-visited_websites<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing service<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messenger<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Online gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming service<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Pornography<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming service<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting service<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom Communications] || Computers Electronics and Technology - Other<br />
|-<br />
| [https://www.nytimes.com The New York Times] || News & Media Publishers<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming and game development platform<br />
|}<br />
<br />
== Open Source trend ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] : Image generative<br />
<br />
ynotopec at gmail.com (mail me)</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1910
Main Page
2026-01-08T10:19:08Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Welcome to the '''infocepo.com''' portal.<br />
<br />
This wiki is intended for system administrators, cloud engineers, developers, students, and enthusiasts who want to:<br />
<br />
* Understand modern architectures (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Deploy private AI assistants and productivity tools<br />
* Build hands-on labs to learn by doing<br />
* Prepare large-scale audits, migrations, and automations<br />
<br />
The goal: turn theory into '''reusable scripts, diagrams, and architectures'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Getting started quickly =<br />
<br />
== Recommended paths ==<br />
<br />
; 1. Build a private AI assistant<br />
* Deploy a typical stack: '''Open WebUI + Ollama + GPU''' (H100 or consumer-grade GPU)<br />
* Add a chat model and a summarization model<br />
* Integrate internal data (RAG, embeddings)<br />
<br />
; 2. Launch a Cloud lab<br />
* Create a small cluster (Kubernetes, OpenStack, or bare-metal)<br />
* Set up a deployment pipeline (Helm, Ansible, Terraform…)<br />
* Add an AI service (transcription, summarization, chatbot…)<br />
<br />
; 3. Prepare an audit / migration<br />
* Inventory servers with '''ServerDiff.sh'''<br />
* Design the target architecture (cloud diagrams)<br />
* Automate the migration with reproducible scripts<br />
<br />
== Content overview ==<br />
<br />
* '''AI guides & tools''' : assistants, models, evaluations, GPUs<br />
* '''Cloud & infrastructure''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Comparison tables''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= future =<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
= AI Assistants & Cloud Tools =<br />
<br />
== AI Assistants ==<br />
<br />
; '''ChatGPT'''<br />
* https://chatgpt.com ChatGPT – Public conversational assistant, suited for exploration, writing, and rapid experimentation.<br />
<br />
; '''Self-hosted AI assistants'''<br />
* https://github.com/open-webui/open-webui Open WebUI + https://www.scaleway.com/en/h100-pcie-try-it-now/ H100 GPU + https://ollama.com Ollama <br />
: Typical stack for private assistants, self-hosted LLMs, and OpenAI-compatible APIs.<br />
* https://github.com/ynotopec/summarize Private summary – Local, fast, offline summarizer for your own data.<br />
<br />
== Development, models & tracking ==<br />
<br />
; '''Discovering and tracking models'''<br />
* https://ollama.com/library LLM Trending – Model library (chat, code, RAG…) for local deployment.<br />
* https://huggingface.co/models Models Trending – Model marketplace, filterable by task, size, and license.<br />
* https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending – Vision-language models (image → text).<br />
* https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation – Image generation model comparisons.<br />
<br />
; '''Evaluation & benchmarks'''<br />
* https://lmarena.ai/leaderboard ChatBot Evaluation – Chatbot rankings (open-source and proprietary models).<br />
* https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard – Benchmark of embedding models for RAG and semantic search.<br />
* https://ann-benchmarks.com Vectors DB Ranking – Vector database comparison (latency, memory, features).<br />
* https://top500.org/lists/green500/ HPC Efficiency – Ranking of the most energy-efficient supercomputers.<br />
<br />
; '''Development & fine-tuning tools'''<br />
* https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending – Major recent open-source projects, sorted by popularity and activity.<br />
* https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning – Advanced framework for LLM fine-tuning (instruction tuning, LoRA, etc.).<br />
* https://www.perplexity.ai Perplexity AI – Advanced research and synthesis oriented as a “research copilot”.<br />
<br />
== AI Hardware & GPUs ==<br />
<br />
; '''GPUs & accelerators'''<br />
* https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100 – Datacenter GPU for Kubernetes clusters and intensive AI workloads.<br />
* NVIDIA 5080 – Consumer GPU for lower-cost private LLM deployments.<br />
* https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator – Hardware accelerator dedicated to LLM inference.<br />
<br />
----<br />
<br />
= Open models & internal endpoints =<br />
<br />
''(Last update: 08/12/2025)''<br />
<br />
The models below correspond to '''logical endpoints''' (for example via a proxy or gateway), selected for specific use cases.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Primary use case<br />
|-<br />
| '''ai-chat''' || Based on '''gpt-oss-20b''' – General-purpose chat, good cost / quality balance.<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, temperature = 0 – Deterministic, reproducible translation (FR, EN, other languages).<br />
|-<br />
| '''ai-summary''' || qwen3 – Model optimized for summarizing long texts (reports, documents, transcriptions).<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – Code reasoning, explanation, and refactoring.<br />
|-<br />
| '''ai-code-completion''' || granite-2b – Fast code completion, designed for IDE auto-completion.<br />
|-<br />
| '''ai-parse''' || gemma2-simpo – Structured extraction, log / JSON / table parsing.<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG usage in French (business knowledge, internal FAQs).<br />
|-<br />
| '''gpt-oss-20b''' || Agentic tasks.<br />
|}<br />
<br />
Usage idea: each endpoint is associated with one or more labs (chat, summary, parsing, RAG, etc.) in the Cloud Lab section.<br />
<br />
----<br />
<br />
= News & Trends =<br />
<br />
* https://www.youtube.com/@lev-selector/videos Top AI News – Curated AI news videos.<br />
* https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper – Example of real-time transcription with speaker detection.<br />
* https://github.com/openai-translator/openai-translator OpenAI Translator – Modern extension / client for LLM-assisted translation.<br />
* https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM – Conversational search based on LLMs and OpenSearch.<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained – Introduction to Transformers, the core architecture of LLMs.<br />
* Hands-on labs, scripts, and real-world feedback in the [[LAB project|CLOUD LAB]] project below.<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
The '''Cloud Lab''' provides reproducible scenarios: infrastructure audits, cloud migration, automation, high availability.<br />
<br />
== Audit project – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Bash audit script to:<br />
<br />
* detect configuration drift,<br />
* compare multiple environments,<br />
* prepare a migration or remediation plan.<br />
<br />
== Example of Cloud migration ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
Example: migration of virtual environments to a modernized cloud, including audit, architecture design, and automation.<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Infrastructure audit || 82 services, automated audit via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Cloud architecture diagram || Visual design and documentation || 1.5<br />
|-<br />
| Compliance checks || 2 clouds, 6 hypervisors, 6 TB of RAM || 1.5<br />
|-<br />
| Cloud platform installation || Deployment of main target environments || 1.0<br />
|-<br />
| Stability verification || Early functional tests || 0.5<br />
|-<br />
| Automation study || Identification and automation of repetitive tasks || 1.5<br />
|-<br />
| Template development || 6 templates, 8 environments, 2 clouds / OS || 1.5<br />
|-<br />
| Migration diagram || Illustration of the migration process || 1.0<br />
|-<br />
| Migration code writing || 138 lines (see '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Process stabilization || Validation that migration is reproducible || 1.5<br />
|-<br />
| Cloud benchmarking || Performance comparison vs legacy infrastructure || 1.5<br />
|-<br />
| Downtime tuning || Calculation of outage time per migration || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs per VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 person-days<br />
|}<br />
<br />
=== Stability checks (minimal HA) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Expected result<br />
|-<br />
| Shutdown of one node || All services must automatically restart on remaining nodes.<br />
|-<br />
| Simultaneous shutdown / restart of all nodes || All services must recover correctly after reboot.<br />
|}<br />
<br />
----<br />
<br />
= Web Architecture & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principles for designing scalable and portable web architectures:<br />
<br />
* Favor '''simple, modular, and flexible''' infrastructure.<br />
* Follow client location (GDNS or equivalent) to bring content closer.<br />
* Use network load balancers (LVS, IPVS) for scalability.<br />
* Systematically compare costs and beware of '''vendor lock-in'''.<br />
* TLS:<br />
** HAProxy for fast frontends,<br />
** Envoy for compatibility and advanced use cases (mTLS, HTTP/2/3).<br />
* Caching:<br />
** Varnish, Apache Traffic Server for large content volumes.<br />
* Favor open-source stacks and database caches (e.g., Memcached).<br />
* Use message queues, buffers, and quotas to smooth traffic spikes.<br />
* For complete architectures:<br />
** https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture<br />
** https://github.com/systemdesign42/system-design System Design GitHub<br />
<br />
----<br />
<br />
= Comparison of major Cloud platforms =<br />
<br />
{| class="wikitable"<br />
! Feature !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router control''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Network virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
This table serves as a starting point for choosing the right stack based on:<br />
<br />
* Desired level of control (API vs bare-metal),<br />
* Context (on-prem, public cloud, HPC, CRM…),<br />
* Existing automation tooling.<br />
<br />
----<br />
<br />
= Useful Cloud & IT links =<br />
<br />
* https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared – AWS / Azure / GCP service mapping.<br />
* https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map – Global Internet mapping.<br />
* https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape – Overview of cloud-native projects (CNCF).<br />
* https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki – Wikimedia infrastructure, real large-scale example.<br />
* https://openapm.io OpenAPM – SRE Tools – APM / observability tooling.<br />
* https://access.redhat.com/downloads/content/package-browser RedHat Package Browser – Package and version search at Red Hat.<br />
* https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Barometer of IT freelance daily rates.<br />
* https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor) – Salary indicators.<br />
<br />
----<br />
<br />
= Advanced: High Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Basic principles:<br />
<br />
* Multi-node or multi-site clusters for redundancy.<br />
* Use of IPMI for fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* For a 2-node cluster:<br />
– carefully sequence fencing to avoid split-brain,<br />
– 3 or more nodes remain recommended for production.<br />
<br />
=== Common resource patterns ===<br />
<br />
* Multipath storage, LUNs, LVM, NFS.<br />
* User resources and application processes.<br />
* Virtual IPs, DNS records, network listeners.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* Job orchestration (SLURM or equivalent).<br />
* High-performance shared storage (GPFS, Lustre…).<br />
* Possible integration with AI workloads (large-scale training, GPU inference).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD pipelines with built-in security checks (linting, SAST, DAST, SBOM).<br />
* Observability (logs, metrics, traces) integrated from design time.<br />
* Automated vulnerability scanning, secret management, policy-as-code.<br />
<br />
----<br />
<br />
= About & Contributions =<br />
<br />
For more examples, scripts, diagrams, and feedback, see:<br />
<br />
* https://infocepo.com infocepo.com<br />
<br />
Suggestions for corrections, diagram improvements, or new labs are welcome. <br />
This wiki aims to remain a '''living laboratory''' for AI, cloud, and automation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1909
Main Page
2026-01-08T10:16:40Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud and AI on infocepo.com]]<br />
<br />
= infocepo.com – Cloud, AI & Labs =<br />
<br />
Welcome to the '''infocepo.com''' portal.<br />
<br />
This wiki is intended for system administrators, cloud engineers, developers, students, and enthusiasts who want to:<br />
<br />
* Understand modern architectures (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Deploy private AI assistants and productivity tools<br />
* Build hands-on labs to learn by doing<br />
* Prepare large-scale audits, migrations, and automations<br />
<br />
The goal: turn theory into '''reusable scripts, diagrams, and architectures'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Getting started quickly =<br />
<br />
== Recommended paths ==<br />
<br />
; 1. Build a private AI assistant<br />
* Deploy a typical stack: '''Open WebUI + Ollama + GPU''' (H100 or consumer-grade GPU)<br />
* Add a chat model and a summarization model<br />
* Integrate internal data (RAG, embeddings)<br />
<br />
; 2. Launch a Cloud lab<br />
* Create a small cluster (Kubernetes, OpenStack, or bare-metal)<br />
* Set up a deployment pipeline (Helm, Ansible, Terraform…)<br />
* Add an AI service (transcription, summarization, chatbot…)<br />
<br />
; 3. Prepare an audit / migration<br />
* Inventory servers with '''ServerDiff.sh'''<br />
* Design the target architecture (cloud diagrams)<br />
* Automate the migration with reproducible scripts<br />
<br />
== Content overview ==<br />
<br />
* '''AI guides & tools''' : assistants, models, evaluations, GPUs<br />
* '''Cloud & infrastructure''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Comparison tables''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= future =<br />
[[File:Automation-full-vs-humans.png|thumb|right|The world after automation]]<br />
<br />
= AI Assistants & Cloud Tools =<br />
<br />
== AI Assistants ==<br />
<br />
; '''ChatGPT'''<br />
* https://chat.openai.com ChatGPT – Public conversational assistant, suited for exploration, writing, and rapid experimentation.<br />
<br />
; '''Self-hosted AI assistants'''<br />
* https://github.com/open-webui/open-webui Open WebUI + https://www.scaleway.com/en/h100-pcie-try-it-now/ H100 GPU + https://ollama.com Ollama <br />
: Typical stack for private assistants, self-hosted LLMs, and OpenAI-compatible APIs.<br />
* https://github.com/ynotopec/summarize Private summary – Local, fast, offline summarizer for your own data.<br />
<br />
== Development, models & tracking ==<br />
<br />
; '''Discovering and tracking models'''<br />
* https://ollama.com/library LLM Trending – Model library (chat, code, RAG…) for local deployment.<br />
* https://huggingface.co/models Models Trending – Model marketplace, filterable by task, size, and license.<br />
* https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending – Vision-language models (image → text).<br />
* https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation – Image generation model comparisons.<br />
<br />
; '''Evaluation & benchmarks'''<br />
* https://lmarena.ai/leaderboard ChatBot Evaluation – Chatbot rankings (open-source and proprietary models).<br />
* https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard – Benchmark of embedding models for RAG and semantic search.<br />
* https://ann-benchmarks.com Vectors DB Ranking – Vector database comparison (latency, memory, features).<br />
* https://top500.org/lists/green500/ HPC Efficiency – Ranking of the most energy-efficient supercomputers.<br />
<br />
; '''Development & fine-tuning tools'''<br />
* https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending – Major recent open-source projects, sorted by popularity and activity.<br />
* https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning – Advanced framework for LLM fine-tuning (instruction tuning, LoRA, etc.).<br />
* https://www.perplexity.ai Perplexity AI – Advanced research and synthesis oriented as a “research copilot”.<br />
<br />
== AI Hardware & GPUs ==<br />
<br />
; '''GPUs & accelerators'''<br />
* https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100 – Datacenter GPU for Kubernetes clusters and intensive AI workloads.<br />
* NVIDIA 5080 – Consumer GPU for lower-cost private LLM deployments.<br />
* https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator – Hardware accelerator dedicated to LLM inference.<br />
<br />
----<br />
<br />
= Open models & internal endpoints =<br />
<br />
''(Last update: 08/12/2025)''<br />
<br />
The models below correspond to '''logical endpoints''' (for example via a proxy or gateway), selected for specific use cases.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Primary use case<br />
|-<br />
| '''ai-chat''' || Based on '''gpt-oss-20b''' – General-purpose chat, good cost / quality balance.<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, temperature = 0 – Deterministic, reproducible translation (FR, EN, other languages).<br />
|-<br />
| '''ai-summary''' || qwen3 – Model optimized for summarizing long texts (reports, documents, transcriptions).<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – Code reasoning, explanation, and refactoring.<br />
|-<br />
| '''ai-code-completion''' || granite-2b – Fast code completion, designed for IDE auto-completion.<br />
|-<br />
| '''ai-parse''' || gemma2-simpo – Structured extraction, log / JSON / table parsing.<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – RAG usage in French (business knowledge, internal FAQs).<br />
|-<br />
| '''gpt-oss-20b''' || Agentic tasks.<br />
|}<br />
<br />
Usage idea: each endpoint is associated with one or more labs (chat, summary, parsing, RAG, etc.) in the Cloud Lab section.<br />
<br />
----<br />
<br />
= News & Trends =<br />
<br />
* https://www.youtube.com/@lev-selector/videos Top AI News – Curated AI news videos.<br />
* https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper – Example of real-time transcription with speaker detection.<br />
* https://github.com/openai-translator/openai-translator OpenAI Translator – Modern extension / client for LLM-assisted translation.<br />
* https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM – Conversational search based on LLMs and OpenSearch.<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained – Introduction to Transformers, the core architecture of LLMs.<br />
* Hands-on labs, scripts, and real-world feedback in the [[LAB project|CLOUD LAB]] project below.<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab reference diagram]]<br />
<br />
The '''Cloud Lab''' provides reproducible scenarios: infrastructure audits, cloud migration, automation, high availability.<br />
<br />
== Audit project – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Bash audit script to:<br />
<br />
* detect configuration drift,<br />
* compare multiple environments,<br />
* prepare a migration or remediation plan.<br />
<br />
== Example of Cloud migration ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud migration diagram]]<br />
<br />
Example: migration of virtual environments to a modernized cloud, including audit, architecture design, and automation.<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Infrastructure audit || 82 services, automated audit via '''ServerDiff.sh''' || 1.5<br />
|-<br />
| Cloud architecture diagram || Visual design and documentation || 1.5<br />
|-<br />
| Compliance checks || 2 clouds, 6 hypervisors, 6 TB of RAM || 1.5<br />
|-<br />
| Cloud platform installation || Deployment of main target environments || 1.0<br />
|-<br />
| Stability verification || Early functional tests || 0.5<br />
|-<br />
| Automation study || Identification and automation of repetitive tasks || 1.5<br />
|-<br />
| Template development || 6 templates, 8 environments, 2 clouds / OS || 1.5<br />
|-<br />
| Migration diagram || Illustration of the migration process || 1.0<br />
|-<br />
| Migration code writing || 138 lines (see '''MigrationApp.sh''') || 1.5<br />
|-<br />
| Process stabilization || Validation that migration is reproducible || 1.5<br />
|-<br />
| Cloud benchmarking || Performance comparison vs legacy infrastructure || 1.5<br />
|-<br />
| Downtime tuning || Calculation of outage time per migration || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs per VM || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 person-days<br />
|}<br />
<br />
=== Stability checks (minimal HA) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Expected result<br />
|-<br />
| Shutdown of one node || All services must automatically restart on remaining nodes.<br />
|-<br />
| Simultaneous shutdown / restart of all nodes || All services must recover correctly after reboot.<br />
|}<br />
<br />
----<br />
<br />
= Web Architecture & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Reference web architecture]]<br />
<br />
Principles for designing scalable and portable web architectures:<br />
<br />
* Favor '''simple, modular, and flexible''' infrastructure.<br />
* Follow client location (GDNS or equivalent) to bring content closer.<br />
* Use network load balancers (LVS, IPVS) for scalability.<br />
* Systematically compare costs and beware of '''vendor lock-in'''.<br />
* TLS:<br />
** HAProxy for fast frontends,<br />
** Envoy for compatibility and advanced use cases (mTLS, HTTP/2/3).<br />
* Caching:<br />
** Varnish, Apache Traffic Server for large content volumes.<br />
* Favor open-source stacks and database caches (e.g., Memcached).<br />
* Use message queues, buffers, and quotas to smooth traffic spikes.<br />
* For complete architectures:<br />
** https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture<br />
** https://github.com/systemdesign42/system-design System Design GitHub<br />
<br />
----<br />
<br />
= Comparison of major Cloud platforms =<br />
<br />
{| class="wikitable"<br />
! Feature !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router control''' || Kube-router || Router/Subnet API || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux firewall || Linux firewall || Linux firewall || API<br />
|-<br />
| '''Network virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
This table serves as a starting point for choosing the right stack based on:<br />
<br />
* Desired level of control (API vs bare-metal),<br />
* Context (on-prem, public cloud, HPC, CRM…),<br />
* Existing automation tooling.<br />
<br />
----<br />
<br />
= Useful Cloud & IT links =<br />
<br />
* https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared – AWS / Azure / GCP service mapping.<br />
* https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map – Global Internet mapping.<br />
* https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape – Overview of cloud-native projects (CNCF).<br />
* https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki – Wikimedia infrastructure, real large-scale example.<br />
* https://openapm.io OpenAPM – SRE Tools – APM / observability tooling.<br />
* https://access.redhat.com/downloads/content/package-browser RedHat Package Browser – Package and version search at Red Hat.<br />
* https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Barometer of IT freelance daily rates.<br />
* https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor) – Salary indicators.<br />
<br />
----<br />
<br />
= Advanced: High Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|HA cluster architecture]]<br />
<br />
Basic principles:<br />
<br />
* Multi-node or multi-site clusters for redundancy.<br />
* Use of IPMI for fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* For a 2-node cluster:<br />
– carefully sequence fencing to avoid split-brain,<br />
– 3 or more nodes remain recommended for production.<br />
<br />
=== Common resource patterns ===<br />
<br />
* Multipath storage, LUNs, LVM, NFS.<br />
* User resources and application processes.<br />
* Virtual IPs, DNS records, network listeners.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Overview of an HPC cluster]]<br />
<br />
* Job orchestration (SLURM or equivalent).<br />
* High-performance shared storage (GPFS, Lustre…).<br />
* Possible integration with AI workloads (large-scale training, GPU inference).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps reference design]]<br />
<br />
* CI/CD pipelines with built-in security checks (linting, SAST, DAST, SBOM).<br />
* Observability (logs, metrics, traces) integrated from design time.<br />
* Automated vulnerability scanning, secret management, policy-as-code.<br />
<br />
----<br />
<br />
= About & Contributions =<br />
<br />
For more examples, scripts, diagrams, and feedback, see:<br />
<br />
* https://infocepo.com infocepo.com<br />
<br />
Suggestions for corrections, diagram improvements, or new labs are welcome. <br />
This wiki aims to remain a '''living laboratory''' for AI, cloud, and automation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1908
Main Page
2026-01-02T12:02:48Z
<p>Tcepo: /* futur */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Découvrir le cloud et l’IA sur infocepo.com]]<br />
<br />
= infocepo.com – Cloud, IA & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki s’adresse aux administrateurs système, ingénieurs cloud, développeurs, étudiants et passionnés qui veulent :<br />
<br />
* Comprendre les architectures modernes (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Mettre en place des assistants IA privés et des outils de productivité<br />
* Construire des labs concrets pour apprendre en pratiquant<br />
* Préparer des audits, migrations et automatisations à grande échelle<br />
<br />
L’objectif : transformer la théorie en '''scripts, diagrammes et architectures réutilisables'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Monter un assistant IA privé<br />
* Déployer un stack type : '''Open WebUI + Ollama + GPU''' (H100 ou GPU grand public)<br />
* Ajouter un modèle de chat + un modèle de résumé<br />
* Intégrer des données internes (RAG, embeddings)<br />
<br />
; 2. Lancer un lab Cloud<br />
* Créer un petit cluster (Kubernetes, OpenStack ou bare-metal)<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA (transcription, résumé, chatbot…)<br />
<br />
; 3. Préparer un audit/migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Dessiner l’architecture cible (diagrammes cloud)<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Résumé des contenus ==<br />
<br />
* '''Guides & outils IA''' : assistants, modèles, évaluations, GPU<br />
* '''Cloud & infra''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Tableaux comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= futur =<br />
[[File:Automation-full-vs-humans.png|thumb|right|Le monde après l'automatisation]]<br />
<br />
= Assistants IA & Outils Cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chat.openai.com ChatGPT] – Assistant conversationnel public, adapté à l’exploration, à la rédaction et à l’expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] <br />
: Stack type pour assistants privés, LLMs auto-hébergés et API compatibles OpenAI.<br />
* [https://github.com/ynotopec/summarize Private summary] – Résumeur local, rapide et hors-ligne pour vos propres données.<br />
<br />
== Développement, modèles & suivi ==<br />
<br />
; '''Découvrir et suivre les modèles'''<br />
* [https://ollama.com/library LLM Trending] – Bibliothèque de modèles (chat, code, RAG…) pour déploiement local.<br />
* [https://huggingface.co/models Models Trending] – Marketplace de modèles, filtrage par tâche, taille, licence.<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Modèles vision-langage (image → texte).<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Comparaison de modèles de génération d’images.<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Classement de chatbots (modèles open-source et propriétaires).<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Benchmark des modèles d’embedding pour RAG et recherche sémantique.<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Comparatif de bases vectorielles (latence, mémoire, fonctionnalités).<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Classement des supercalculateurs les plus efficaces énergétiquement.<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Grands projets open source récents, triés par popularité et activité.<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Framework avancé pour fine-tuning de LLMs (instruction-tuning, LoRA, etc.).<br />
* [https://www.perplexity.ai Perplexity AI] – Recherche avancée et synthèse orientée “copilote de recherche”.<br />
<br />
== Matériel IA & GPU ==<br />
<br />
; '''GPU & accélérateurs'''<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – GPU datacenter pour clusters Kubernetes et workloads IA intensifs.<br />
* NVIDIA 5080 – GPU grand public pour déploiements privés de LLMs à moindre coût.<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Accélérateur matériel dédié à l’inférence LLM.<br />
<br />
----<br />
<br />
= Modèles ouverts & Endpoints internes =<br />
<br />
''(Dernière mise à jour : 08/12/2025)''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' (par exemple via un proxy ou une gateway), choisis pour des usages précis.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Cas d’usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – Chat généraliste, bon compromis coût / qualité.<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – Traduction déterministe, reproductible (FR, EN, autres langues).<br />
|-<br />
| '''ai-summary''' || qwen3 – Modèle optimisé pour la synthèse de textes longs (rapports, docs, transcriptions).<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – Raisonnement sur le code, explication et refactorisation.<br />
|-<br />
| '''ai-code-completion''' || granite-2b – Complétion de code rapide, pensée pour l’auto-complétion IDE.<br />
|-<br />
| '''ai-parse''' || gemma2-simpo – Extraction structurée, parsing de logs / JSON / tableaux.<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – Usage RAG en français (connaissance métier, FAQ internes).<br />
|-<br />
| '''gpt-oss-20b''' || Tâches agentiques.<br />
|}<br />
<br />
Idée d’usage : chaque endpoint est associé à un ou plusieurs labs (chat, résumé, parsing, RAG, etc.) dans la section Cloud Lab.<br />
<br />
----<br />
<br />
= Actualités & Tendances =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Sélection vidéo des actus IA.<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Exemple de transcription temps réel avec détection de locuteurs.<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Extension / client moderne pour traduction assistée par LLM.<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Recherche conversationnelle basée sur LLMs et OpenSearch.<br />
<br />
----<br />
<br />
= Formation & Apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Introduction aux Transformers, architecture clé des LLMs.<br />
* Labs, scripts et retours d’expérience concrets dans le projet [[LAB project|CLOUD LAB]] ci-dessous.<br />
<br />
----<br />
<br />
= Cloud Lab & Projets d’Audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Diagramme de référence Cloud Lab]]<br />
<br />
Le '''Cloud Lab''' propose des scénarios reproductibles : audit d’infrastructure, migration vers le cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit de serveurs pour :<br />
<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer une migration ou un plan de remédiation.<br />
<br />
== Exemple de migration Cloud ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Diagramme de migration Cloud]]<br />
<br />
Exemple : migration d’environnements virtuels vers un cloud modernisé, avec audit, design d’architecture et automatisation.<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit de l’infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1,5<br />
|-<br />
| Diagramme d’architecture cloud || Conception et documentation visuelle || 1,5<br />
|-<br />
| Contrôle de conformité || 2 clouds, 6 hyperviseurs, 6 To de RAM || 1,5<br />
|-<br />
| Installation des plateformes cloud || Déploiement des environnements cibles principaux || 1,0<br />
|-<br />
| Vérification de stabilité || Tests de fonctionnement précoce || 0,5<br />
|-<br />
| Étude d’automatisation || Identification et automatisation des tâches répétitives || 1,5<br />
|-<br />
| Développement de templates || 6 templates, 8 environnements, 2 clouds / OS || 1,5<br />
|-<br />
| Diagramme de migration || Illustration du processus de migration || 1,0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1,5<br />
|-<br />
| Stabilisation du process || Validation que la migration est reproductible || 1,5<br />
|-<br />
| Benchmarking cloud || Comparaison de performance vs infrastructure legacy || 1,5<br />
|-<br />
| Calage du downtime || Calcul du temps de coupure par migration || 0,5<br />
|-<br />
| Chargement des VMs || 82 VMs : OS, code, 2 IP par VM || 0,1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours·homme<br />
|}<br />
<br />
=== Vérification de stabilité (HA minimale) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services doivent redémarrer automatiquement sur les nœuds restants.<br />
|-<br />
| Extinction / rallumage simultané de tous les nœuds || Tous les services doivent remonter correctement après redémarrage.<br />
|}<br />
<br />
----<br />
<br />
= Architecture Web & Bonnes Pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Architecture Web de référence]]<br />
<br />
Principes pour concevoir des architectures web scalables et portables :<br />
<br />
* Favoriser une infrastructure '''simple, modulaire et flexible'''.<br />
* Suivre la localisation des clients (GDNS ou équivalent) pour rapprocher les contenus.<br />
* Utiliser des load balancers réseau (LVS, IPVS) pour la montée en charge.<br />
* Comparer systématiquement les coûts, et se méfier du '''vendor lock-in'''.<br />
* TLS :<br />
** HAProxy pour les frontends rapides,<br />
** Envoy pour la compatibilité et les cas avancés (mTLS, HTTP/2/3).<br />
* Cache :<br />
** Varnish, Apache Traffic Server pour les gros volumes de contenu.<br />
* Privilégier les stacks open source et les caches de base de données (ex. Memcached).<br />
* Utiliser des files de messages, buffers et quotas pour lisser les pics de charge.<br />
* Pour des architectures complètes :<br />
** [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture]<br />
** [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes Cloud =<br />
<br />
{| class="wikitable"<br />
! Fonction !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle des routeurs''' || Kube-router || API Routeur/Sous-réseau || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle du pare-feu''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Pare-feu Linux || Pare-feu Linux || Pare-feu Linux || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN, autres || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Options de stockage''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Ce tableau sert de point de départ pour choisir la bonne pile selon :<br />
<br />
* Niveau de contrôle souhaité (API vs bare-metal),<br />
* Contexte (on-prem, cloud public, HPC, CRM…),<br />
* Outillage d’automatisation déjà en place.<br />
<br />
----<br />
<br />
= Liens utiles Cloud & IT =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared] – Correspondance des services AWS / Azure / GCP.<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map] – Cartographie globale de l’Internet.<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape] – Panorama des projets cloud-native (CNCF).<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki] – Infrastructure Wikimedia, exemple réel à grande échelle.<br />
* [https://openapm.io OpenAPM – SRE Tools] – Outillage APM / observabilité.<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser] – Recherche de paquets et versions chez Red Hat.<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM Freelance IT] – Référentiel de tarifs freelances.<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)] – Indications de salaires IT.<br />
<br />
----<br />
<br />
= Avancé : Haute Disponibilité, HPC & DevSecOps =<br />
<br />
== Haute Disponibilité avec Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|Architecture de cluster HA]]<br />
<br />
Principes de base :<br />
<br />
* Clusters multi-nœuds ou bi-sites pour la redondance.<br />
* Utilisation d’IPMI pour le fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* Pour un cluster 2 nœuds : <br />
– séquencer le fencing pour éviter les split-brains, <br />
– 3 nœuds ou plus restent recommandés pour la production.<br />
<br />
=== Pattern de ressources courant ===<br />
<br />
* Stockage multipath, LUN, LVM, NFS.<br />
* Ressources utilisateurs et processus applicatifs.<br />
* IP virtuelle, enregistrements DNS, listeners réseau.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Vue d’ensemble d’un cluster HPC]]<br />
<br />
* Orchestration de jobs (SLURM ou équivalent).<br />
* Partage de stockage haute performance (GPFS, Lustre…).<br />
* Intégration possible avec des workloads IA (entraînement massif, inference sur GPU).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|Design de référence DevSecOps]]<br />
<br />
* Pipelines CI/CD avec contrôles de sécurité intégrés (linting, SAST, DAST, SBOM).<br />
* Observabilité (logs, metrics, traces) intégrée dès la conception.<br />
* Automatisation des scans de vulnérabilités, gestion des secrets, policy-as-code.<br />
<br />
----<br />
<br />
= À propos & Contributions =<br />
<br />
Pour plus d’exemples, de scripts, de diagrammes et de retours d’expérience, consulter :<br />
<br />
* [https://infocepo.com infocepo.com]<br />
<br />
Les suggestions de correction, d’amélioration de diagrammes, ou de nouveaux labs sont les bienvenues. <br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1907
Main Page
2026-01-02T12:02:26Z
<p>Tcepo: /* futur */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Découvrir le cloud et l’IA sur infocepo.com]]<br />
<br />
= infocepo.com – Cloud, IA & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki s’adresse aux administrateurs système, ingénieurs cloud, développeurs, étudiants et passionnés qui veulent :<br />
<br />
* Comprendre les architectures modernes (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Mettre en place des assistants IA privés et des outils de productivité<br />
* Construire des labs concrets pour apprendre en pratiquant<br />
* Préparer des audits, migrations et automatisations à grande échelle<br />
<br />
L’objectif : transformer la théorie en '''scripts, diagrammes et architectures réutilisables'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Monter un assistant IA privé<br />
* Déployer un stack type : '''Open WebUI + Ollama + GPU''' (H100 ou GPU grand public)<br />
* Ajouter un modèle de chat + un modèle de résumé<br />
* Intégrer des données internes (RAG, embeddings)<br />
<br />
; 2. Lancer un lab Cloud<br />
* Créer un petit cluster (Kubernetes, OpenStack ou bare-metal)<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA (transcription, résumé, chatbot…)<br />
<br />
; 3. Préparer un audit/migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Dessiner l’architecture cible (diagrammes cloud)<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Résumé des contenus ==<br />
<br />
* '''Guides & outils IA''' : assistants, modèles, évaluations, GPU<br />
* '''Cloud & infra''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Tableaux comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= futur =<br />
[[File:Automation-full-vs-humans.png|thumb|right|Le monde après l'automatisation]]<br />
<br />
---<br />
<br />
= Assistants IA & Outils Cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chat.openai.com ChatGPT] – Assistant conversationnel public, adapté à l’exploration, à la rédaction et à l’expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] <br />
: Stack type pour assistants privés, LLMs auto-hébergés et API compatibles OpenAI.<br />
* [https://github.com/ynotopec/summarize Private summary] – Résumeur local, rapide et hors-ligne pour vos propres données.<br />
<br />
== Développement, modèles & suivi ==<br />
<br />
; '''Découvrir et suivre les modèles'''<br />
* [https://ollama.com/library LLM Trending] – Bibliothèque de modèles (chat, code, RAG…) pour déploiement local.<br />
* [https://huggingface.co/models Models Trending] – Marketplace de modèles, filtrage par tâche, taille, licence.<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Modèles vision-langage (image → texte).<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Comparaison de modèles de génération d’images.<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Classement de chatbots (modèles open-source et propriétaires).<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Benchmark des modèles d’embedding pour RAG et recherche sémantique.<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Comparatif de bases vectorielles (latence, mémoire, fonctionnalités).<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Classement des supercalculateurs les plus efficaces énergétiquement.<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Grands projets open source récents, triés par popularité et activité.<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Framework avancé pour fine-tuning de LLMs (instruction-tuning, LoRA, etc.).<br />
* [https://www.perplexity.ai Perplexity AI] – Recherche avancée et synthèse orientée “copilote de recherche”.<br />
<br />
== Matériel IA & GPU ==<br />
<br />
; '''GPU & accélérateurs'''<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – GPU datacenter pour clusters Kubernetes et workloads IA intensifs.<br />
* NVIDIA 5080 – GPU grand public pour déploiements privés de LLMs à moindre coût.<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Accélérateur matériel dédié à l’inférence LLM.<br />
<br />
----<br />
<br />
= Modèles ouverts & Endpoints internes =<br />
<br />
''(Dernière mise à jour : 08/12/2025)''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' (par exemple via un proxy ou une gateway), choisis pour des usages précis.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Cas d’usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – Chat généraliste, bon compromis coût / qualité.<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – Traduction déterministe, reproductible (FR, EN, autres langues).<br />
|-<br />
| '''ai-summary''' || qwen3 – Modèle optimisé pour la synthèse de textes longs (rapports, docs, transcriptions).<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – Raisonnement sur le code, explication et refactorisation.<br />
|-<br />
| '''ai-code-completion''' || granite-2b – Complétion de code rapide, pensée pour l’auto-complétion IDE.<br />
|-<br />
| '''ai-parse''' || gemma2-simpo – Extraction structurée, parsing de logs / JSON / tableaux.<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – Usage RAG en français (connaissance métier, FAQ internes).<br />
|-<br />
| '''gpt-oss-20b''' || Tâches agentiques.<br />
|}<br />
<br />
Idée d’usage : chaque endpoint est associé à un ou plusieurs labs (chat, résumé, parsing, RAG, etc.) dans la section Cloud Lab.<br />
<br />
----<br />
<br />
= Actualités & Tendances =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Sélection vidéo des actus IA.<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Exemple de transcription temps réel avec détection de locuteurs.<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Extension / client moderne pour traduction assistée par LLM.<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Recherche conversationnelle basée sur LLMs et OpenSearch.<br />
<br />
----<br />
<br />
= Formation & Apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Introduction aux Transformers, architecture clé des LLMs.<br />
* Labs, scripts et retours d’expérience concrets dans le projet [[LAB project|CLOUD LAB]] ci-dessous.<br />
<br />
----<br />
<br />
= Cloud Lab & Projets d’Audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Diagramme de référence Cloud Lab]]<br />
<br />
Le '''Cloud Lab''' propose des scénarios reproductibles : audit d’infrastructure, migration vers le cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit de serveurs pour :<br />
<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer une migration ou un plan de remédiation.<br />
<br />
== Exemple de migration Cloud ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Diagramme de migration Cloud]]<br />
<br />
Exemple : migration d’environnements virtuels vers un cloud modernisé, avec audit, design d’architecture et automatisation.<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit de l’infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1,5<br />
|-<br />
| Diagramme d’architecture cloud || Conception et documentation visuelle || 1,5<br />
|-<br />
| Contrôle de conformité || 2 clouds, 6 hyperviseurs, 6 To de RAM || 1,5<br />
|-<br />
| Installation des plateformes cloud || Déploiement des environnements cibles principaux || 1,0<br />
|-<br />
| Vérification de stabilité || Tests de fonctionnement précoce || 0,5<br />
|-<br />
| Étude d’automatisation || Identification et automatisation des tâches répétitives || 1,5<br />
|-<br />
| Développement de templates || 6 templates, 8 environnements, 2 clouds / OS || 1,5<br />
|-<br />
| Diagramme de migration || Illustration du processus de migration || 1,0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1,5<br />
|-<br />
| Stabilisation du process || Validation que la migration est reproductible || 1,5<br />
|-<br />
| Benchmarking cloud || Comparaison de performance vs infrastructure legacy || 1,5<br />
|-<br />
| Calage du downtime || Calcul du temps de coupure par migration || 0,5<br />
|-<br />
| Chargement des VMs || 82 VMs : OS, code, 2 IP par VM || 0,1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours·homme<br />
|}<br />
<br />
=== Vérification de stabilité (HA minimale) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services doivent redémarrer automatiquement sur les nœuds restants.<br />
|-<br />
| Extinction / rallumage simultané de tous les nœuds || Tous les services doivent remonter correctement après redémarrage.<br />
|}<br />
<br />
----<br />
<br />
= Architecture Web & Bonnes Pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Architecture Web de référence]]<br />
<br />
Principes pour concevoir des architectures web scalables et portables :<br />
<br />
* Favoriser une infrastructure '''simple, modulaire et flexible'''.<br />
* Suivre la localisation des clients (GDNS ou équivalent) pour rapprocher les contenus.<br />
* Utiliser des load balancers réseau (LVS, IPVS) pour la montée en charge.<br />
* Comparer systématiquement les coûts, et se méfier du '''vendor lock-in'''.<br />
* TLS :<br />
** HAProxy pour les frontends rapides,<br />
** Envoy pour la compatibilité et les cas avancés (mTLS, HTTP/2/3).<br />
* Cache :<br />
** Varnish, Apache Traffic Server pour les gros volumes de contenu.<br />
* Privilégier les stacks open source et les caches de base de données (ex. Memcached).<br />
* Utiliser des files de messages, buffers et quotas pour lisser les pics de charge.<br />
* Pour des architectures complètes :<br />
** [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture]<br />
** [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes Cloud =<br />
<br />
{| class="wikitable"<br />
! Fonction !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle des routeurs''' || Kube-router || API Routeur/Sous-réseau || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle du pare-feu''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Pare-feu Linux || Pare-feu Linux || Pare-feu Linux || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN, autres || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Options de stockage''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Ce tableau sert de point de départ pour choisir la bonne pile selon :<br />
<br />
* Niveau de contrôle souhaité (API vs bare-metal),<br />
* Contexte (on-prem, cloud public, HPC, CRM…),<br />
* Outillage d’automatisation déjà en place.<br />
<br />
----<br />
<br />
= Liens utiles Cloud & IT =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared] – Correspondance des services AWS / Azure / GCP.<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map] – Cartographie globale de l’Internet.<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape] – Panorama des projets cloud-native (CNCF).<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki] – Infrastructure Wikimedia, exemple réel à grande échelle.<br />
* [https://openapm.io OpenAPM – SRE Tools] – Outillage APM / observabilité.<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser] – Recherche de paquets et versions chez Red Hat.<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM Freelance IT] – Référentiel de tarifs freelances.<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)] – Indications de salaires IT.<br />
<br />
----<br />
<br />
= Avancé : Haute Disponibilité, HPC & DevSecOps =<br />
<br />
== Haute Disponibilité avec Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|Architecture de cluster HA]]<br />
<br />
Principes de base :<br />
<br />
* Clusters multi-nœuds ou bi-sites pour la redondance.<br />
* Utilisation d’IPMI pour le fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* Pour un cluster 2 nœuds : <br />
– séquencer le fencing pour éviter les split-brains, <br />
– 3 nœuds ou plus restent recommandés pour la production.<br />
<br />
=== Pattern de ressources courant ===<br />
<br />
* Stockage multipath, LUN, LVM, NFS.<br />
* Ressources utilisateurs et processus applicatifs.<br />
* IP virtuelle, enregistrements DNS, listeners réseau.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Vue d’ensemble d’un cluster HPC]]<br />
<br />
* Orchestration de jobs (SLURM ou équivalent).<br />
* Partage de stockage haute performance (GPFS, Lustre…).<br />
* Intégration possible avec des workloads IA (entraînement massif, inference sur GPU).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|Design de référence DevSecOps]]<br />
<br />
* Pipelines CI/CD avec contrôles de sécurité intégrés (linting, SAST, DAST, SBOM).<br />
* Observabilité (logs, metrics, traces) intégrée dès la conception.<br />
* Automatisation des scans de vulnérabilités, gestion des secrets, policy-as-code.<br />
<br />
----<br />
<br />
= À propos & Contributions =<br />
<br />
Pour plus d’exemples, de scripts, de diagrammes et de retours d’expérience, consulter :<br />
<br />
* [https://infocepo.com infocepo.com]<br />
<br />
Les suggestions de correction, d’amélioration de diagrammes, ou de nouveaux labs sont les bienvenues. <br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1906
Main Page
2026-01-02T12:02:06Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Découvrir le cloud et l’IA sur infocepo.com]]<br />
<br />
= infocepo.com – Cloud, IA & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki s’adresse aux administrateurs système, ingénieurs cloud, développeurs, étudiants et passionnés qui veulent :<br />
<br />
* Comprendre les architectures modernes (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Mettre en place des assistants IA privés et des outils de productivité<br />
* Construire des labs concrets pour apprendre en pratiquant<br />
* Préparer des audits, migrations et automatisations à grande échelle<br />
<br />
L’objectif : transformer la théorie en '''scripts, diagrammes et architectures réutilisables'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Monter un assistant IA privé<br />
* Déployer un stack type : '''Open WebUI + Ollama + GPU''' (H100 ou GPU grand public)<br />
* Ajouter un modèle de chat + un modèle de résumé<br />
* Intégrer des données internes (RAG, embeddings)<br />
<br />
; 2. Lancer un lab Cloud<br />
* Créer un petit cluster (Kubernetes, OpenStack ou bare-metal)<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA (transcription, résumé, chatbot…)<br />
<br />
; 3. Préparer un audit/migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Dessiner l’architecture cible (diagrammes cloud)<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Résumé des contenus ==<br />
<br />
* '''Guides & outils IA''' : assistants, modèles, évaluations, GPU<br />
* '''Cloud & infra''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Tableaux comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= futur =<br />
[[File:Automation-full-vs-humans.png|thumb|right|Le monde après l'automatisation]<br />
<br />
---<br />
<br />
= Assistants IA & Outils Cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chat.openai.com ChatGPT] – Assistant conversationnel public, adapté à l’exploration, à la rédaction et à l’expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] <br />
: Stack type pour assistants privés, LLMs auto-hébergés et API compatibles OpenAI.<br />
* [https://github.com/ynotopec/summarize Private summary] – Résumeur local, rapide et hors-ligne pour vos propres données.<br />
<br />
== Développement, modèles & suivi ==<br />
<br />
; '''Découvrir et suivre les modèles'''<br />
* [https://ollama.com/library LLM Trending] – Bibliothèque de modèles (chat, code, RAG…) pour déploiement local.<br />
* [https://huggingface.co/models Models Trending] – Marketplace de modèles, filtrage par tâche, taille, licence.<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Modèles vision-langage (image → texte).<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Comparaison de modèles de génération d’images.<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Classement de chatbots (modèles open-source et propriétaires).<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Benchmark des modèles d’embedding pour RAG et recherche sémantique.<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Comparatif de bases vectorielles (latence, mémoire, fonctionnalités).<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Classement des supercalculateurs les plus efficaces énergétiquement.<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Grands projets open source récents, triés par popularité et activité.<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Framework avancé pour fine-tuning de LLMs (instruction-tuning, LoRA, etc.).<br />
* [https://www.perplexity.ai Perplexity AI] – Recherche avancée et synthèse orientée “copilote de recherche”.<br />
<br />
== Matériel IA & GPU ==<br />
<br />
; '''GPU & accélérateurs'''<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – GPU datacenter pour clusters Kubernetes et workloads IA intensifs.<br />
* NVIDIA 5080 – GPU grand public pour déploiements privés de LLMs à moindre coût.<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Accélérateur matériel dédié à l’inférence LLM.<br />
<br />
----<br />
<br />
= Modèles ouverts & Endpoints internes =<br />
<br />
''(Dernière mise à jour : 08/12/2025)''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' (par exemple via un proxy ou une gateway), choisis pour des usages précis.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Cas d’usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – Chat généraliste, bon compromis coût / qualité.<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – Traduction déterministe, reproductible (FR, EN, autres langues).<br />
|-<br />
| '''ai-summary''' || qwen3 – Modèle optimisé pour la synthèse de textes longs (rapports, docs, transcriptions).<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – Raisonnement sur le code, explication et refactorisation.<br />
|-<br />
| '''ai-code-completion''' || granite-2b – Complétion de code rapide, pensée pour l’auto-complétion IDE.<br />
|-<br />
| '''ai-parse''' || gemma2-simpo – Extraction structurée, parsing de logs / JSON / tableaux.<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – Usage RAG en français (connaissance métier, FAQ internes).<br />
|-<br />
| '''gpt-oss-20b''' || Tâches agentiques.<br />
|}<br />
<br />
Idée d’usage : chaque endpoint est associé à un ou plusieurs labs (chat, résumé, parsing, RAG, etc.) dans la section Cloud Lab.<br />
<br />
----<br />
<br />
= Actualités & Tendances =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Sélection vidéo des actus IA.<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Exemple de transcription temps réel avec détection de locuteurs.<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Extension / client moderne pour traduction assistée par LLM.<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Recherche conversationnelle basée sur LLMs et OpenSearch.<br />
<br />
----<br />
<br />
= Formation & Apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Introduction aux Transformers, architecture clé des LLMs.<br />
* Labs, scripts et retours d’expérience concrets dans le projet [[LAB project|CLOUD LAB]] ci-dessous.<br />
<br />
----<br />
<br />
= Cloud Lab & Projets d’Audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Diagramme de référence Cloud Lab]]<br />
<br />
Le '''Cloud Lab''' propose des scénarios reproductibles : audit d’infrastructure, migration vers le cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit de serveurs pour :<br />
<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer une migration ou un plan de remédiation.<br />
<br />
== Exemple de migration Cloud ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Diagramme de migration Cloud]]<br />
<br />
Exemple : migration d’environnements virtuels vers un cloud modernisé, avec audit, design d’architecture et automatisation.<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit de l’infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1,5<br />
|-<br />
| Diagramme d’architecture cloud || Conception et documentation visuelle || 1,5<br />
|-<br />
| Contrôle de conformité || 2 clouds, 6 hyperviseurs, 6 To de RAM || 1,5<br />
|-<br />
| Installation des plateformes cloud || Déploiement des environnements cibles principaux || 1,0<br />
|-<br />
| Vérification de stabilité || Tests de fonctionnement précoce || 0,5<br />
|-<br />
| Étude d’automatisation || Identification et automatisation des tâches répétitives || 1,5<br />
|-<br />
| Développement de templates || 6 templates, 8 environnements, 2 clouds / OS || 1,5<br />
|-<br />
| Diagramme de migration || Illustration du processus de migration || 1,0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1,5<br />
|-<br />
| Stabilisation du process || Validation que la migration est reproductible || 1,5<br />
|-<br />
| Benchmarking cloud || Comparaison de performance vs infrastructure legacy || 1,5<br />
|-<br />
| Calage du downtime || Calcul du temps de coupure par migration || 0,5<br />
|-<br />
| Chargement des VMs || 82 VMs : OS, code, 2 IP par VM || 0,1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours·homme<br />
|}<br />
<br />
=== Vérification de stabilité (HA minimale) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services doivent redémarrer automatiquement sur les nœuds restants.<br />
|-<br />
| Extinction / rallumage simultané de tous les nœuds || Tous les services doivent remonter correctement après redémarrage.<br />
|}<br />
<br />
----<br />
<br />
= Architecture Web & Bonnes Pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Architecture Web de référence]]<br />
<br />
Principes pour concevoir des architectures web scalables et portables :<br />
<br />
* Favoriser une infrastructure '''simple, modulaire et flexible'''.<br />
* Suivre la localisation des clients (GDNS ou équivalent) pour rapprocher les contenus.<br />
* Utiliser des load balancers réseau (LVS, IPVS) pour la montée en charge.<br />
* Comparer systématiquement les coûts, et se méfier du '''vendor lock-in'''.<br />
* TLS :<br />
** HAProxy pour les frontends rapides,<br />
** Envoy pour la compatibilité et les cas avancés (mTLS, HTTP/2/3).<br />
* Cache :<br />
** Varnish, Apache Traffic Server pour les gros volumes de contenu.<br />
* Privilégier les stacks open source et les caches de base de données (ex. Memcached).<br />
* Utiliser des files de messages, buffers et quotas pour lisser les pics de charge.<br />
* Pour des architectures complètes :<br />
** [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture]<br />
** [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes Cloud =<br />
<br />
{| class="wikitable"<br />
! Fonction !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle des routeurs''' || Kube-router || API Routeur/Sous-réseau || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle du pare-feu''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Pare-feu Linux || Pare-feu Linux || Pare-feu Linux || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN, autres || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Options de stockage''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Ce tableau sert de point de départ pour choisir la bonne pile selon :<br />
<br />
* Niveau de contrôle souhaité (API vs bare-metal),<br />
* Contexte (on-prem, cloud public, HPC, CRM…),<br />
* Outillage d’automatisation déjà en place.<br />
<br />
----<br />
<br />
= Liens utiles Cloud & IT =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared] – Correspondance des services AWS / Azure / GCP.<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map] – Cartographie globale de l’Internet.<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape] – Panorama des projets cloud-native (CNCF).<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki] – Infrastructure Wikimedia, exemple réel à grande échelle.<br />
* [https://openapm.io OpenAPM – SRE Tools] – Outillage APM / observabilité.<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser] – Recherche de paquets et versions chez Red Hat.<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM Freelance IT] – Référentiel de tarifs freelances.<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)] – Indications de salaires IT.<br />
<br />
----<br />
<br />
= Avancé : Haute Disponibilité, HPC & DevSecOps =<br />
<br />
== Haute Disponibilité avec Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|Architecture de cluster HA]]<br />
<br />
Principes de base :<br />
<br />
* Clusters multi-nœuds ou bi-sites pour la redondance.<br />
* Utilisation d’IPMI pour le fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* Pour un cluster 2 nœuds : <br />
– séquencer le fencing pour éviter les split-brains, <br />
– 3 nœuds ou plus restent recommandés pour la production.<br />
<br />
=== Pattern de ressources courant ===<br />
<br />
* Stockage multipath, LUN, LVM, NFS.<br />
* Ressources utilisateurs et processus applicatifs.<br />
* IP virtuelle, enregistrements DNS, listeners réseau.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Vue d’ensemble d’un cluster HPC]]<br />
<br />
* Orchestration de jobs (SLURM ou équivalent).<br />
* Partage de stockage haute performance (GPFS, Lustre…).<br />
* Intégration possible avec des workloads IA (entraînement massif, inference sur GPU).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|Design de référence DevSecOps]]<br />
<br />
* Pipelines CI/CD avec contrôles de sécurité intégrés (linting, SAST, DAST, SBOM).<br />
* Observabilité (logs, metrics, traces) intégrée dès la conception.<br />
* Automatisation des scans de vulnérabilités, gestion des secrets, policy-as-code.<br />
<br />
----<br />
<br />
= À propos & Contributions =<br />
<br />
Pour plus d’exemples, de scripts, de diagrammes et de retours d’expérience, consulter :<br />
<br />
* [https://infocepo.com infocepo.com]<br />
<br />
Les suggestions de correction, d’amélioration de diagrammes, ou de nouveaux labs sont les bienvenues. <br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=File:Automation-full-vs-humans.png&diff=1905
File:Automation-full-vs-humans.png
2026-01-02T12:00:26Z
<p>Tcepo: Tcepo uploaded a new version of File:Automation-full-vs-humans.png</p>
<hr />
<div></div>
Tcepo
https://infocepo.com/wiki/index.php?title=File:Automation-full-vs-humans.png&diff=1904
File:Automation-full-vs-humans.png
2026-01-02T11:55:54Z
<p>Tcepo: </p>
<hr />
<div></div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1903
Main Page
2025-12-08T16:49:30Z
<p>Tcepo: /* Modèles ouverts & Endpoints internes */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Découvrir le cloud et l’IA sur infocepo.com]]<br />
<br />
= infocepo.com – Cloud, IA & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki s’adresse aux administrateurs système, ingénieurs cloud, développeurs, étudiants et passionnés qui veulent :<br />
<br />
* Comprendre les architectures modernes (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Mettre en place des assistants IA privés et des outils de productivité<br />
* Construire des labs concrets pour apprendre en pratiquant<br />
* Préparer des audits, migrations et automatisations à grande échelle<br />
<br />
L’objectif : transformer la théorie en '''scripts, diagrammes et architectures réutilisables'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Monter un assistant IA privé<br />
* Déployer un stack type : '''Open WebUI + Ollama + GPU''' (H100 ou GPU grand public)<br />
* Ajouter un modèle de chat + un modèle de résumé<br />
* Intégrer des données internes (RAG, embeddings)<br />
<br />
; 2. Lancer un lab Cloud<br />
* Créer un petit cluster (Kubernetes, OpenStack ou bare-metal)<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA (transcription, résumé, chatbot…)<br />
<br />
; 3. Préparer un audit/migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Dessiner l’architecture cible (diagrammes cloud)<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Résumé des contenus ==<br />
<br />
* '''Guides & outils IA''' : assistants, modèles, évaluations, GPU<br />
* '''Cloud & infra''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Tableaux comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Assistants IA & Outils Cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chat.openai.com ChatGPT] – Assistant conversationnel public, adapté à l’exploration, à la rédaction et à l’expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] <br />
: Stack type pour assistants privés, LLMs auto-hébergés et API compatibles OpenAI.<br />
* [https://github.com/ynotopec/summarize Private summary] – Résumeur local, rapide et hors-ligne pour vos propres données.<br />
<br />
== Développement, modèles & suivi ==<br />
<br />
; '''Découvrir et suivre les modèles'''<br />
* [https://ollama.com/library LLM Trending] – Bibliothèque de modèles (chat, code, RAG…) pour déploiement local.<br />
* [https://huggingface.co/models Models Trending] – Marketplace de modèles, filtrage par tâche, taille, licence.<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Modèles vision-langage (image → texte).<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Comparaison de modèles de génération d’images.<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Classement de chatbots (modèles open-source et propriétaires).<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Benchmark des modèles d’embedding pour RAG et recherche sémantique.<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Comparatif de bases vectorielles (latence, mémoire, fonctionnalités).<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Classement des supercalculateurs les plus efficaces énergétiquement.<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Grands projets open source récents, triés par popularité et activité.<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Framework avancé pour fine-tuning de LLMs (instruction-tuning, LoRA, etc.).<br />
* [https://www.perplexity.ai Perplexity AI] – Recherche avancée et synthèse orientée “copilote de recherche”.<br />
<br />
== Matériel IA & GPU ==<br />
<br />
; '''GPU & accélérateurs'''<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – GPU datacenter pour clusters Kubernetes et workloads IA intensifs.<br />
* NVIDIA 5080 – GPU grand public pour déploiements privés de LLMs à moindre coût.<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Accélérateur matériel dédié à l’inférence LLM.<br />
<br />
----<br />
<br />
= Modèles ouverts & Endpoints internes =<br />
<br />
''(Dernière mise à jour : 08/12/2025)''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' (par exemple via un proxy ou une gateway), choisis pour des usages précis.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Cas d’usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gpt-oss-20b''' – Chat généraliste, bon compromis coût / qualité.<br />
|-<br />
| '''ai-translate''' || gpt-oss-20b, température = 0 – Traduction déterministe, reproductible (FR, EN, autres langues).<br />
|-<br />
| '''ai-summary''' || qwen3 – Modèle optimisé pour la synthèse de textes longs (rapports, docs, transcriptions).<br />
|-<br />
| '''ai-code''' || gpt-oss-20b – Raisonnement sur le code, explication et refactorisation.<br />
|-<br />
| '''ai-code-completion''' || granite-2b – Complétion de code rapide, pensée pour l’auto-complétion IDE.<br />
|-<br />
| '''ai-parse''' || gemma2-simpo – Extraction structurée, parsing de logs / JSON / tableaux.<br />
|-<br />
| '''ai-RAG-FR''' || qwen3 – Usage RAG en français (connaissance métier, FAQ internes).<br />
|-<br />
| '''gpt-oss-20b''' || Tâches agentiques.<br />
|}<br />
<br />
Idée d’usage : chaque endpoint est associé à un ou plusieurs labs (chat, résumé, parsing, RAG, etc.) dans la section Cloud Lab.<br />
<br />
----<br />
<br />
= Actualités & Tendances =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Sélection vidéo des actus IA.<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Exemple de transcription temps réel avec détection de locuteurs.<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Extension / client moderne pour traduction assistée par LLM.<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Recherche conversationnelle basée sur LLMs et OpenSearch.<br />
<br />
----<br />
<br />
= Formation & Apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Introduction aux Transformers, architecture clé des LLMs.<br />
* Labs, scripts et retours d’expérience concrets dans le projet [[LAB project|CLOUD LAB]] ci-dessous.<br />
<br />
----<br />
<br />
= Cloud Lab & Projets d’Audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Diagramme de référence Cloud Lab]]<br />
<br />
Le '''Cloud Lab''' propose des scénarios reproductibles : audit d’infrastructure, migration vers le cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit de serveurs pour :<br />
<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer une migration ou un plan de remédiation.<br />
<br />
== Exemple de migration Cloud ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Diagramme de migration Cloud]]<br />
<br />
Exemple : migration d’environnements virtuels vers un cloud modernisé, avec audit, design d’architecture et automatisation.<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit de l’infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1,5<br />
|-<br />
| Diagramme d’architecture cloud || Conception et documentation visuelle || 1,5<br />
|-<br />
| Contrôle de conformité || 2 clouds, 6 hyperviseurs, 6 To de RAM || 1,5<br />
|-<br />
| Installation des plateformes cloud || Déploiement des environnements cibles principaux || 1,0<br />
|-<br />
| Vérification de stabilité || Tests de fonctionnement précoce || 0,5<br />
|-<br />
| Étude d’automatisation || Identification et automatisation des tâches répétitives || 1,5<br />
|-<br />
| Développement de templates || 6 templates, 8 environnements, 2 clouds / OS || 1,5<br />
|-<br />
| Diagramme de migration || Illustration du processus de migration || 1,0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1,5<br />
|-<br />
| Stabilisation du process || Validation que la migration est reproductible || 1,5<br />
|-<br />
| Benchmarking cloud || Comparaison de performance vs infrastructure legacy || 1,5<br />
|-<br />
| Calage du downtime || Calcul du temps de coupure par migration || 0,5<br />
|-<br />
| Chargement des VMs || 82 VMs : OS, code, 2 IP par VM || 0,1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours·homme<br />
|}<br />
<br />
=== Vérification de stabilité (HA minimale) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services doivent redémarrer automatiquement sur les nœuds restants.<br />
|-<br />
| Extinction / rallumage simultané de tous les nœuds || Tous les services doivent remonter correctement après redémarrage.<br />
|}<br />
<br />
----<br />
<br />
= Architecture Web & Bonnes Pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Architecture Web de référence]]<br />
<br />
Principes pour concevoir des architectures web scalables et portables :<br />
<br />
* Favoriser une infrastructure '''simple, modulaire et flexible'''.<br />
* Suivre la localisation des clients (GDNS ou équivalent) pour rapprocher les contenus.<br />
* Utiliser des load balancers réseau (LVS, IPVS) pour la montée en charge.<br />
* Comparer systématiquement les coûts, et se méfier du '''vendor lock-in'''.<br />
* TLS :<br />
** HAProxy pour les frontends rapides,<br />
** Envoy pour la compatibilité et les cas avancés (mTLS, HTTP/2/3).<br />
* Cache :<br />
** Varnish, Apache Traffic Server pour les gros volumes de contenu.<br />
* Privilégier les stacks open source et les caches de base de données (ex. Memcached).<br />
* Utiliser des files de messages, buffers et quotas pour lisser les pics de charge.<br />
* Pour des architectures complètes :<br />
** [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture]<br />
** [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes Cloud =<br />
<br />
{| class="wikitable"<br />
! Fonction !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle des routeurs''' || Kube-router || API Routeur/Sous-réseau || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle du pare-feu''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Pare-feu Linux || Pare-feu Linux || Pare-feu Linux || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN, autres || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Options de stockage''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Ce tableau sert de point de départ pour choisir la bonne pile selon :<br />
<br />
* Niveau de contrôle souhaité (API vs bare-metal),<br />
* Contexte (on-prem, cloud public, HPC, CRM…),<br />
* Outillage d’automatisation déjà en place.<br />
<br />
----<br />
<br />
= Liens utiles Cloud & IT =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared] – Correspondance des services AWS / Azure / GCP.<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map] – Cartographie globale de l’Internet.<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape] – Panorama des projets cloud-native (CNCF).<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki] – Infrastructure Wikimedia, exemple réel à grande échelle.<br />
* [https://openapm.io OpenAPM – SRE Tools] – Outillage APM / observabilité.<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser] – Recherche de paquets et versions chez Red Hat.<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM Freelance IT] – Référentiel de tarifs freelances.<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)] – Indications de salaires IT.<br />
<br />
----<br />
<br />
= Avancé : Haute Disponibilité, HPC & DevSecOps =<br />
<br />
== Haute Disponibilité avec Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|Architecture de cluster HA]]<br />
<br />
Principes de base :<br />
<br />
* Clusters multi-nœuds ou bi-sites pour la redondance.<br />
* Utilisation d’IPMI pour le fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* Pour un cluster 2 nœuds : <br />
– séquencer le fencing pour éviter les split-brains, <br />
– 3 nœuds ou plus restent recommandés pour la production.<br />
<br />
=== Pattern de ressources courant ===<br />
<br />
* Stockage multipath, LUN, LVM, NFS.<br />
* Ressources utilisateurs et processus applicatifs.<br />
* IP virtuelle, enregistrements DNS, listeners réseau.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Vue d’ensemble d’un cluster HPC]]<br />
<br />
* Orchestration de jobs (SLURM ou équivalent).<br />
* Partage de stockage haute performance (GPFS, Lustre…).<br />
* Intégration possible avec des workloads IA (entraînement massif, inference sur GPU).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|Design de référence DevSecOps]]<br />
<br />
* Pipelines CI/CD avec contrôles de sécurité intégrés (linting, SAST, DAST, SBOM).<br />
* Observabilité (logs, metrics, traces) intégrée dès la conception.<br />
* Automatisation des scans de vulnérabilités, gestion des secrets, policy-as-code.<br />
<br />
----<br />
<br />
= À propos & Contributions =<br />
<br />
Pour plus d’exemples, de scripts, de diagrammes et de retours d’expérience, consulter :<br />
<br />
* [https://infocepo.com infocepo.com]<br />
<br />
Les suggestions de correction, d’amélioration de diagrammes, ou de nouveaux labs sont les bienvenues. <br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Human_needs&diff=1902
Human needs
2025-12-07T13:38:33Z
<p>Tcepo: /* TOP WEBSITES (CATEGORY AWARE) */</p>
<hr />
<div>* '''[https://en.wikipedia.org/wiki/Shelter Shelter]''' : g1[[https://en.wikipedia.org/wiki/Particle Particle], [https://en.wikipedia.org/wiki/Motion Motion], [https://en.wikipedia.org/wiki/Fundamental_interaction Fundamental interaction], [https://en.wikipedia.org/wiki/Science Science]]<br />
* '''Structured elementary particles''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Clothes Clothes]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Toilet Toilet]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cleaner Cleaner]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transportation Transportation]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Defense Defense]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Help Help]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transducer Information transducer]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Information_processing Information processing]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Data_storage Information storer]''' (like disks, like books) : g1<br />
* '''[https://en.wikipedia.org/wiki/Search_engine Information seeker]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Transmitter Information transmitter]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Stimulants Stimulants]''' (like structured elementary particles, like love, like entertainment...) : g1<br />
* '''[https://en.wikipedia.org/wiki/Particle Particle]'''<br />
* '''[https://en.wikipedia.org/wiki/Leisure Leisure]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Training Training]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Cutlery Cutlery]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Mattress Mattress]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Table (furniture)]''' : g1<br />
* '''[https://en.wikipedia.org/wiki/Chair Chair]''' : g1<br />
<br><br />
<br />
==TOP WEBSITES (CATEGORY AWARE)==<br />
# https://en.wikipedia.org/wiki/List_of_most-visited_websites<br />
{| class="wikitable"<br />
! Website !! Type<br />
|-<br />
| [https://www.google.com Google] || Search engine<br />
|-<br />
| [https://www.youtube.com YouTube] || Video sharing service<br />
|-<br />
| [https://www.facebook.com Facebook] || Social media<br />
|-<br />
| [https://www.chatgpt.com ChatGPT] || Chatbot<br />
|-<br />
| [https://www.whatsapp.com WhatsApp] || Instant messenger<br />
|-<br />
| [https://www.wikipedia.org Wikipedia] || Encyclopedia<br />
|-<br />
| [https://www.yahoo.co.jp Yahoo! Japan] || News<br />
|-<br />
| [https://www.amazon.com Amazon] || Marketplace<br />
|-<br />
| [https://bet.br BET.br] || Online gambling<br />
|-<br />
| [https://www.office.com Microsoft 365] || Software<br />
|-<br />
| [https://www.netflix.com Netflix] || Streaming service<br />
|-<br />
| [https://www.pornhub.com Pornhub] || Pornography<br />
|-<br />
| [https://www.live.com Live] || Email<br />
|-<br />
| [https://www.twitch.tv Twitch] || Livestreaming service<br />
|-<br />
| [https://www.samsung.com Samsung] || Consumer electronics<br />
|-<br />
| [https://www.weather.com Weather] || Weather<br />
|-<br />
| [https://www.fandom.com Fandom] || Wiki hosting service<br />
|-<br />
| [https://www.stripchat.com Stripchat] || Adult camming<br />
|-<br />
| [https://www.zoom.com Zoom Communications] || Computers Electronics and Technology - Other<br />
|-<br />
| [https://www.nytimes.com The New York Times] || News & Media Publishers<br />
|-<br />
| [https://www.espn.com ESPN] || Sports<br />
|-<br />
| [https://www.roblox.com Roblox] || Gaming and game development platform<br />
|}<br />
<br />
== Open Source trend ==<br />
* [https://github.com/Stability-AI/StableCascade StableCascade] : Image generative<br />
<br />
ynotopec at gmail.com (mail me)</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Enigme&diff=1901
Enigme
2025-12-06T21:14:25Z
<p>Tcepo: </p>
<hr />
<div>'''1. Quel est l'âge de Paul ?<br />
<br />
Paul s'adresse à Pierre :<br />
<br />
"J'ai le double de l'âge que tu avais quand j'avais ton âge.<br />
Quand tu auras mon âge, ensemble nous aurons 72 ans."<br />
<br />
Quel âge a Paul ?<br />
Quel âge a Pierre ?<br />
<br />
* English :<br />
Paul addresses Peter:<br />
<br />
"I am twice as old as you were when I was your age. When you are my age, together we will be 72 years old."<br />
<br />
How old is Paul? How old is Peter?<br />
<br />
<br />
<br />
* Avec somme future 81 : Paul a 36 ans, Pierre a 27 ans.<br />
<br />
<br />
'''2. Un lac étonnant!'''<br />
<br />
Un lac double de volume chaque jour.<br />
Au bout de 30 jours il est plein.<br />
<br />
Le lac est donc plein à la moitié au bout de combien de jours ?<br />
<br />
'''3. Les 3 prisionniers<br />
<br />
Trois prisonniers sont ensemble. Chacun portant une pancarte au dos tiré au hasard parmi 3 pancartes blanches et 2 noirs. <br />
<br />
Chaque prisonnier ne peut pas voir sa propre pancarte, mais peut voir celles des autres.<br />
<br />
Dans cette situation, les trois prisonniers reçoivent chacun une pancarte blanche.<br />
<br />
Ceux qui devinent la couleur de leur propre pancarte annoncent à tout le monde qu'ils ont trouvé et sont libérés.<br />
<br />
Comment est-il possible de deviner la couleur de sa propre pancarte dans cette situation ?<br />
<br />
* English :<br />
Three prisoners are together. Each of them wears a sign on their back, randomly drawn from a set of 3 white signs and 2 black signs.<br />
<br />
Each prisoner cannot see the color of their own sign but can see the signs of the others.<br />
<br />
In this situation, all three prisoners receive a white sign.<br />
<br />
The first prisoner who correctly guesses the color of their own sign raises their hand and is set free.<br />
<br />
How is it possible to determine the color of their own sign in this scenario?</div>
Tcepo
https://infocepo.com/wiki/index.php?title=File:Infocepo-picture.png&diff=1900
File:Infocepo-picture.png
2025-11-24T14:22:40Z
<p>Tcepo: Tcepo uploaded a new version of File:Infocepo-picture.png</p>
<hr />
<div></div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1899
Main Page
2025-11-24T14:13:19Z
<p>Tcepo: </p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Découvrir le cloud et l’IA sur infocepo.com]]<br />
<br />
= infocepo.com – Cloud, IA & Labs =<br />
<br />
Bienvenue sur le portail '''infocepo.com'''.<br />
<br />
Ce wiki s’adresse aux administrateurs système, ingénieurs cloud, développeurs, étudiants et passionnés qui veulent :<br />
<br />
* Comprendre les architectures modernes (Kubernetes, OpenStack, bare-metal, HPC…)<br />
* Mettre en place des assistants IA privés et des outils de productivité<br />
* Construire des labs concrets pour apprendre en pratiquant<br />
* Préparer des audits, migrations et automatisations à grande échelle<br />
<br />
L’objectif : transformer la théorie en '''scripts, diagrammes et architectures réutilisables'''.<br />
<br />
__TOC__<br />
<br />
----<br />
<br />
= Démarrer rapidement =<br />
<br />
== Parcours recommandés ==<br />
<br />
; 1. Monter un assistant IA privé<br />
* Déployer un stack type : '''Open WebUI + Ollama + GPU''' (H100 ou GPU grand public)<br />
* Ajouter un modèle de chat + un modèle de résumé<br />
* Intégrer des données internes (RAG, embeddings)<br />
<br />
; 2. Lancer un lab Cloud<br />
* Créer un petit cluster (Kubernetes, OpenStack ou bare-metal)<br />
* Mettre en place un pipeline de déploiement (Helm, Ansible, Terraform…)<br />
* Ajouter un service IA (transcription, résumé, chatbot…)<br />
<br />
; 3. Préparer un audit/migration<br />
* Inventorier les serveurs avec '''ServerDiff.sh'''<br />
* Dessiner l’architecture cible (diagrammes cloud)<br />
* Automatiser la migration avec des scripts reproductibles<br />
<br />
== Résumé des contenus ==<br />
<br />
* '''Guides & outils IA''' : assistants, modèles, évaluations, GPU<br />
* '''Cloud & infra''' : HA, HPC, web-scale, DevSecOps<br />
* '''Labs & scripts''' : audit, migration, automation<br />
* '''Tableaux comparatifs''' : Kubernetes vs OpenStack vs AWS vs bare-metal, etc.<br />
<br />
----<br />
<br />
= Assistants IA & Outils Cloud =<br />
<br />
== Assistants IA ==<br />
<br />
; '''ChatGPT'''<br />
* [https://chat.openai.com ChatGPT] – Assistant conversationnel public, adapté à l’exploration, à la rédaction et à l’expérimentation rapide.<br />
<br />
; '''Assistants IA auto-hébergés'''<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] <br />
: Stack type pour assistants privés, LLMs auto-hébergés et API compatibles OpenAI.<br />
* [https://github.com/ynotopec/summarize Private summary] – Résumeur local, rapide et hors-ligne pour vos propres données.<br />
<br />
== Développement, modèles & suivi ==<br />
<br />
; '''Découvrir et suivre les modèles'''<br />
* [https://ollama.com/library LLM Trending] – Bibliothèque de modèles (chat, code, RAG…) pour déploiement local.<br />
* [https://huggingface.co/models Models Trending] – Marketplace de modèles, filtrage par tâche, taille, licence.<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Modèles vision-langage (image → texte).<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Comparaison de modèles de génération d’images.<br />
<br />
; '''Évaluation & benchmarks'''<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Classement de chatbots (modèles open-source et propriétaires).<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Benchmark des modèles d’embedding pour RAG et recherche sémantique.<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Comparatif de bases vectorielles (latence, mémoire, fonctionnalités).<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Classement des supercalculateurs les plus efficaces énergétiquement.<br />
<br />
; '''Outils de développement & fine-tuning'''<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Grands projets open source récents, triés par popularité et activité.<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Framework avancé pour fine-tuning de LLMs (instruction-tuning, LoRA, etc.).<br />
* [https://www.perplexity.ai Perplexity AI] – Recherche avancée et synthèse orientée “copilote de recherche”.<br />
<br />
== Matériel IA & GPU ==<br />
<br />
; '''GPU & accélérateurs'''<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – GPU datacenter pour clusters Kubernetes et workloads IA intensifs.<br />
* NVIDIA 5080 – GPU grand public pour déploiements privés de LLMs à moindre coût.<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Accélérateur matériel dédié à l’inférence LLM.<br />
<br />
----<br />
<br />
= Modèles ouverts & Endpoints internes =<br />
<br />
''(Dernière mise à jour : 26/07/2025)''<br />
<br />
Les modèles ci-dessous correspondent à des '''endpoints logiques''' (par exemple via un proxy ou une gateway), choisis pour des usages précis.<br />
<br />
{| class="wikitable"<br />
! Endpoint !! Description / Cas d’usage principal<br />
|-<br />
| '''ai-chat''' || Basé sur '''gemma3n''' – Chat généraliste, bon compromis coût / qualité.<br />
|-<br />
| '''ai-translate''' || gemma3n, température = 0 – Traduction déterministe, reproductible (FR, EN, autres langues).<br />
|-<br />
| '''ai-summary''' || qwen2.5 – Modèle optimisé pour la synthèse de textes longs (rapports, docs, transcriptions).<br />
|-<br />
| '''ai-code''' || mistral-small – Raisonnement sur le code, explication et refactorisation.<br />
|-<br />
| '''ai-code-completion''' || granite-2b – Complétion de code rapide, pensée pour l’auto-complétion IDE.<br />
|-<br />
| '''ai-parse''' || gemma2-simpo – Extraction structurée, parsing de logs / JSON / tableaux.<br />
|-<br />
| '''ai-RAG-FR''' || qwen2.5 – Usage RAG en français (connaissance métier, FAQ internes).<br />
|-<br />
| '''mannix/gemma2-9b-simpo''' || Intégration avec '''OllamaFunctions''' (outils, API, automation).<br />
|}<br />
<br />
Idée d’usage : chaque endpoint est associé à un ou plusieurs labs (chat, résumé, parsing, RAG, etc.) dans la section Cloud Lab.<br />
<br />
----<br />
<br />
= Actualités & Tendances =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Sélection vidéo des actus IA.<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Exemple de transcription temps réel avec détection de locuteurs.<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Extension / client moderne pour traduction assistée par LLM.<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Recherche conversationnelle basée sur LLMs et OpenSearch.<br />
<br />
----<br />
<br />
= Formation & Apprentissage =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Introduction aux Transformers, architecture clé des LLMs.<br />
* Labs, scripts et retours d’expérience concrets dans le projet [[LAB project|CLOUD LAB]] ci-dessous.<br />
<br />
----<br />
<br />
= Cloud Lab & Projets d’Audit =<br />
<br />
[[File:Infocepo.drawio.png|400px|Diagramme de référence Cloud Lab]]<br />
<br />
Le '''Cloud Lab''' propose des scénarios reproductibles : audit d’infrastructure, migration vers le cloud, automatisation, haute disponibilité.<br />
<br />
== Projet d’audit – Cloud Audit ==<br />
<br />
; '''[[ServerDiff.sh]]'''<br />
Script Bash d’audit de serveurs pour :<br />
<br />
* détecter les dérives de configuration,<br />
* comparer plusieurs environnements,<br />
* préparer une migration ou un plan de remédiation.<br />
<br />
== Exemple de migration Cloud ==<br />
<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Diagramme de migration Cloud]]<br />
<br />
Exemple : migration d’environnements virtuels vers un cloud modernisé, avec audit, design d’architecture et automatisation.<br />
<br />
{| class="wikitable"<br />
! Tâche !! Description !! Durée (jours)<br />
|-<br />
| Audit de l’infrastructure || 82 services, audit automatisé via '''ServerDiff.sh''' || 1,5<br />
|-<br />
| Diagramme d’architecture cloud || Conception et documentation visuelle || 1,5<br />
|-<br />
| Contrôle de conformité || 2 clouds, 6 hyperviseurs, 6 To de RAM || 1,5<br />
|-<br />
| Installation des plateformes cloud || Déploiement des environnements cibles principaux || 1,0<br />
|-<br />
| Vérification de stabilité || Tests de fonctionnement précoce || 0,5<br />
|-<br />
| Étude d’automatisation || Identification et automatisation des tâches répétitives || 1,5<br />
|-<br />
| Développement de templates || 6 templates, 8 environnements, 2 clouds / OS || 1,5<br />
|-<br />
| Diagramme de migration || Illustration du processus de migration || 1,0<br />
|-<br />
| Écriture du code de migration || 138 lignes (voir '''MigrationApp.sh''') || 1,5<br />
|-<br />
| Stabilisation du process || Validation que la migration est reproductible || 1,5<br />
|-<br />
| Benchmarking cloud || Comparaison de performance vs infrastructure legacy || 1,5<br />
|-<br />
| Calage du downtime || Calcul du temps de coupure par migration || 0,5<br />
|-<br />
| Chargement des VMs || 82 VMs : OS, code, 2 IP par VM || 0,1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 jours·homme<br />
|}<br />
<br />
=== Vérification de stabilité (HA minimale) ===<br />
<br />
{| class="wikitable"<br />
! Action !! Résultat attendu<br />
|-<br />
| Extinction d’un nœud || Tous les services doivent redémarrer automatiquement sur les nœuds restants.<br />
|-<br />
| Extinction / rallumage simultané de tous les nœuds || Tous les services doivent remonter correctement après redémarrage.<br />
|}<br />
<br />
----<br />
<br />
= Architecture Web & Bonnes Pratiques =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Architecture Web de référence]]<br />
<br />
Principes pour concevoir des architectures web scalables et portables :<br />
<br />
* Favoriser une infrastructure '''simple, modulaire et flexible'''.<br />
* Suivre la localisation des clients (GDNS ou équivalent) pour rapprocher les contenus.<br />
* Utiliser des load balancers réseau (LVS, IPVS) pour la montée en charge.<br />
* Comparer systématiquement les coûts, et se méfier du '''vendor lock-in'''.<br />
* TLS :<br />
** HAProxy pour les frontends rapides,<br />
** Envoy pour la compatibilité et les cas avancés (mTLS, HTTP/2/3).<br />
* Cache :<br />
** Varnish, Apache Traffic Server pour les gros volumes de contenu.<br />
* Privilégier les stacks open source et les caches de base de données (ex. Memcached).<br />
* Utiliser des files de messages, buffers et quotas pour lisser les pics de charge.<br />
* Pour des architectures complètes :<br />
** [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture]<br />
** [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Comparatif des grandes plateformes Cloud =<br />
<br />
{| class="wikitable"<br />
! Fonction !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Outils de déploiement''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Méthode de bootstrap''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Contrôle des routeurs''' || Kube-router || API Routeur/Sous-réseau || Route Table / Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Contrôle du pare-feu''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Pare-feu Linux || Pare-feu Linux || Pare-feu Linux || API<br />
|-<br />
| '''Virtualisation réseau''' || VLAN, VxLAN, autres || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Options de stockage''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
Ce tableau sert de point de départ pour choisir la bonne pile selon :<br />
<br />
* Niveau de contrôle souhaité (API vs bare-metal),<br />
* Contexte (on-prem, cloud public, HPC, CRM…),<br />
* Outillage d’automatisation déjà en place.<br />
<br />
----<br />
<br />
= Liens utiles Cloud & IT =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared] – Correspondance des services AWS / Azure / GCP.<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map] – Cartographie globale de l’Internet.<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape] – Panorama des projets cloud-native (CNCF).<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki] – Infrastructure Wikimedia, exemple réel à grande échelle.<br />
* [https://openapm.io OpenAPM – SRE Tools] – Outillage APM / observabilité.<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser] – Recherche de paquets et versions chez Red Hat.<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Baromètre TJM Freelance IT] – Référentiel de tarifs freelances.<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)] – Indications de salaires IT.<br />
<br />
----<br />
<br />
= Avancé : Haute Disponibilité, HPC & DevSecOps =<br />
<br />
== Haute Disponibilité avec Corosync & Pacemaker ==<br />
<br />
[[File:HA-REF.drawio.png|400px|Architecture de cluster HA]]<br />
<br />
Principes de base :<br />
<br />
* Clusters multi-nœuds ou bi-sites pour la redondance.<br />
* Utilisation d’IPMI pour le fencing, provisioning via PXE/NTP/DNS/TFTP.<br />
* Pour un cluster 2 nœuds : <br />
– séquencer le fencing pour éviter les split-brains, <br />
– 3 nœuds ou plus restent recommandés pour la production.<br />
<br />
=== Pattern de ressources courant ===<br />
<br />
* Stockage multipath, LUN, LVM, NFS.<br />
* Ressources utilisateurs et processus applicatifs.<br />
* IP virtuelle, enregistrements DNS, listeners réseau.<br />
<br />
== HPC ==<br />
<br />
[[File:HPC.drawio.png|400px|Vue d’ensemble d’un cluster HPC]]<br />
<br />
* Orchestration de jobs (SLURM ou équivalent).<br />
* Partage de stockage haute performance (GPFS, Lustre…).<br />
* Intégration possible avec des workloads IA (entraînement massif, inference sur GPU).<br />
<br />
== DevSecOps ==<br />
<br />
[[File:DSO-POC-V3.drawio.png|400px|Design de référence DevSecOps]]<br />
<br />
* Pipelines CI/CD avec contrôles de sécurité intégrés (linting, SAST, DAST, SBOM).<br />
* Observabilité (logs, metrics, traces) intégrée dès la conception.<br />
* Automatisation des scans de vulnérabilités, gestion des secrets, policy-as-code.<br />
<br />
----<br />
<br />
= À propos & Contributions =<br />
<br />
Pour plus d’exemples, de scripts, de diagrammes et de retours d’expérience, consulter :<br />
<br />
* [https://infocepo.com infocepo.com]<br />
<br />
Les suggestions de correction, d’amélioration de diagrammes, ou de nouveaux labs sont les bienvenues. <br />
Ce wiki a vocation à rester un '''laboratoire vivant''' pour l’IA, le cloud et l’automatisation.</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1898
Main Page
2025-11-24T10:28:30Z
<p>Tcepo: /* Discover Cloud Computing on infocepo.com */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud computing on infocepo.com]]<br />
<br />
= Discover Cloud Computing on infocepo.com =<br />
<br />
Welcome! This portal is designed for IT professionals, engineers, students, and enthusiasts who want to master cloud infrastructure, explore AI tools, and accelerate their IT skills through hands-on labs and open-source solutions.<br />
<br />
<br />
__TOC__<br />
<br />
== Quick Start ==<br />
* '''Master cloud infrastructure:''' Practical guides and labs<br />
* '''Explore artificial intelligence:''' Trends and hands-on tools<br />
* '''Compare cloud providers:''' Kubernetes, AWS, OpenStack, and more<br />
* '''Develop expertise:''' Training, open-source, and real-world projects<br />
<br />
----<br />
<br />
= AI & Cloud Tools =<br />
<br />
; '''AI Assistants'''<br />
* [https://chat.openai.com ChatGPT] – Public conversational AI with strong learning capabilities<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] – Private assistants and self-hosted LLM APIs<br />
* [https://github.com/ynotopec/summarize Private summary] – Fast, offline summarizer for your data<br />
<br />
; '''Development & Model Tracking'''<br />
* [https://ollama.com/library LLM Trending] – Latest open-source LLMs<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Top trending codebases since 2022<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Compare chatbot performance<br />
* [https://www.perplexity.ai Perplexity AI] – Cutting-edge research and question answering<br />
* [https://huggingface.co/models Models Trending] – Model marketplace<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Advanced training framework<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Ranking for vector search models<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Database speed and feature comparison<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – HPC/AI GPUs for Kubernetes clusters<br />
* NVIDIA 5080 – Consumer GPU for private deployments<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Vision-language models<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Compare generative image models<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Top green supercomputers<br />
<br />
----<br />
<br />
== Notable Open LLMs ==<br />
''(Last updated: 26/07/2025)''<br />
<br />
{| class="wikitable"<br />
! Model !! Description / Notable Features<br />
|-<br />
| '''ai-chat''' || gemma3n, cost efficient<br />
|-<br />
| '''ai-translate''' || gemma3n, temperature=0 (deterministic translation)<br />
|-<br />
| '''ai-summary''' || qwen2.5, optimized for summarization<br />
|-<br />
| '''ai-code''' || mistral-small, advanced code reasoning<br />
|-<br />
| '''ai-code-completion''' || granite-2b, fast code suggestions<br />
|-<br />
| '''ai-parse''' || gemma2-simpo, parsing & extraction<br />
|-<br />
| '''ai-RAG-FR''' || qwen2.5, French RAG applications<br />
|-<br />
| '''mannix/gemma2-9b-simpo''' || OllamaFunctions integration<br />
|}<br />
<br />
----<br />
<br />
= Industry News & Trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Video digest<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Speaker tracking<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Modern open-source translation<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Fast, low-cost inference hardware<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Enhanced search experiences<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Intro to Transformers algorithm<br />
* Hands-on labs and scripts in the [[LAB project|CLOUD LAB]] below<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab Reference Diagram]]<br />
<br />
; '''Lab Project''' <br />
Experiment with high-availability, cloud migration, and audit automation.<br />
<br />
=== Cloud Audit ===<br />
* '''[[ServerDiff.sh]]''' – Bash script for auditing servers, tracking config drift, and checking environment consistency<br />
<br />
=== Cloud Migration Example ===<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud Migration Diagram]]<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Audit infrastructure || 82 services, automated via ServerDiff.sh || 1.5<br />
|-<br />
| Diagram cloud architecture || Visual design || 1.5<br />
|-<br />
| Compliance check || 2 clouds, 6 hypervisors, 6TB RAM || 1.5<br />
|-<br />
| Install cloud platforms || Deploy core cloud environments || 1.0<br />
|-<br />
| Stability check || Early operations || 0.5<br />
|-<br />
| Automation study || Automate deployment/tasks || 1.5<br />
|-<br />
| Develop templates || 6 templates, 8 envs, 2 clouds/OS || 1.5<br />
|-<br />
| Migration diagram || Process illustration || 1.0<br />
|-<br />
| Write migration code || 138 lines (see MigrationApp.sh) || 1.5<br />
|-<br />
| Process stabilization || Ensure repeatable migration || 1.5<br />
|-<br />
| Cloud benchmarking || Performance test vs legacy || 1.5<br />
|-<br />
| Downtime calibration || Per-migration time calculation || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs each || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 man-days<br />
|}<br />
<br />
==== Stability check ====<br />
<br />
{| class="wikitable"<br />
! Action !! Expected Result<br />
|-<br />
| Power off one node || All resources started<br />
|-<br />
| Power off/on all nodes simultaneously || All resources started<br />
|}<br />
<br />
----<br />
<br />
= Web Infrastructure & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Web Architecture Reference]]<br />
<br />
* Favor minimal, flexible infrastructure<br />
* Track customer location via GDNS or similar<br />
* Use network load balancers (LVS, IPVS) for scaling<br />
* Compare prices and beware of vendor lock-in<br />
* For TLS: use HAProxy for fast frontend, Envoy for compatibility<br />
* Caching: Varnish, Apache Traffic Server for large content<br />
* Prefer open-source stacks and database caches (e.g. Memcached)<br />
* Use message queues and buffers for workload smoothing<br />
* For more examples: [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture], [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Major Cloud Platforms: Feature Comparison =<br />
<br />
{| class="wikitable"<br />
! Function !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment Tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap Method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router Control''' || Kube-router || Router/Subnet API || Route Table/Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall Control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux Firewall || Linux Firewall || Linux Firewall || API<br />
|-<br />
| '''Network Virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage Options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
----<br />
<br />
= Useful Cloud & IT Links =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM – SRE Tools]<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Freelance IT Rates]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)]<br />
<br />
----<br />
<br />
= Advanced: High-Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA Cluster Architecture]]<br />
<br />
* Multi-node or dual-room clusters for redundancy<br />
* Use IPMI for fencing, provision via PXE/NTP/DNS/TFTP<br />
* For 2-node clusters: stagger fencing for stability; 3+ nodes recommended<br />
<br />
=== Common Resources Pattern ===<br />
* Multipath storage, LUN, LVM, NFS<br />
* User and process resources<br />
* IP, DNS, Listener management<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|HPC Cluster Overview]]<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps Reference Design]]<br />
<br />
----<br />
<br />
'''For more examples, guides, and scripts, visit [https://infocepo.com infocepo.com]. Contributions and suggestions welcome!'''</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1897
Main Page
2025-11-24T10:28:04Z
<p>Tcepo: /* Discover Cloud Computing on infocepo.com */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud computing on infocepo.com]]<br />
<br />
= Discover Cloud Computing on infocepo.com =<br />
<br />
Welcome! This portal is designed for IT professionals, engineers, students, and enthusiasts who want to master cloud infrastructure, explore AI tools, and accelerate their IT skills through hands-on labs and open-source solutions.<br />
<br />
<br />
__TOC__<br />
<br />
<br />
== Quick Start ==<br />
* '''Master cloud infrastructure:''' Practical guides and labs<br />
* '''Explore artificial intelligence:''' Trends and hands-on tools<br />
* '''Compare cloud providers:''' Kubernetes, AWS, OpenStack, and more<br />
* '''Develop expertise:''' Training, open-source, and real-world projects<br />
<br />
----<br />
<br />
= AI & Cloud Tools =<br />
<br />
; '''AI Assistants'''<br />
* [https://chat.openai.com ChatGPT] – Public conversational AI with strong learning capabilities<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] – Private assistants and self-hosted LLM APIs<br />
* [https://github.com/ynotopec/summarize Private summary] – Fast, offline summarizer for your data<br />
<br />
; '''Development & Model Tracking'''<br />
* [https://ollama.com/library LLM Trending] – Latest open-source LLMs<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Top trending codebases since 2022<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Compare chatbot performance<br />
* [https://www.perplexity.ai Perplexity AI] – Cutting-edge research and question answering<br />
* [https://huggingface.co/models Models Trending] – Model marketplace<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Advanced training framework<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Ranking for vector search models<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Database speed and feature comparison<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – HPC/AI GPUs for Kubernetes clusters<br />
* NVIDIA 5080 – Consumer GPU for private deployments<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Vision-language models<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Compare generative image models<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Top green supercomputers<br />
<br />
----<br />
<br />
== Notable Open LLMs ==<br />
''(Last updated: 26/07/2025)''<br />
<br />
{| class="wikitable"<br />
! Model !! Description / Notable Features<br />
|-<br />
| '''ai-chat''' || gemma3n, cost efficient<br />
|-<br />
| '''ai-translate''' || gemma3n, temperature=0 (deterministic translation)<br />
|-<br />
| '''ai-summary''' || qwen2.5, optimized for summarization<br />
|-<br />
| '''ai-code''' || mistral-small, advanced code reasoning<br />
|-<br />
| '''ai-code-completion''' || granite-2b, fast code suggestions<br />
|-<br />
| '''ai-parse''' || gemma2-simpo, parsing & extraction<br />
|-<br />
| '''ai-RAG-FR''' || qwen2.5, French RAG applications<br />
|-<br />
| '''mannix/gemma2-9b-simpo''' || OllamaFunctions integration<br />
|}<br />
<br />
----<br />
<br />
= Industry News & Trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Video digest<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Speaker tracking<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Modern open-source translation<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Fast, low-cost inference hardware<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Enhanced search experiences<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Intro to Transformers algorithm<br />
* Hands-on labs and scripts in the [[LAB project|CLOUD LAB]] below<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab Reference Diagram]]<br />
<br />
; '''Lab Project''' <br />
Experiment with high-availability, cloud migration, and audit automation.<br />
<br />
=== Cloud Audit ===<br />
* '''[[ServerDiff.sh]]''' – Bash script for auditing servers, tracking config drift, and checking environment consistency<br />
<br />
=== Cloud Migration Example ===<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud Migration Diagram]]<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Audit infrastructure || 82 services, automated via ServerDiff.sh || 1.5<br />
|-<br />
| Diagram cloud architecture || Visual design || 1.5<br />
|-<br />
| Compliance check || 2 clouds, 6 hypervisors, 6TB RAM || 1.5<br />
|-<br />
| Install cloud platforms || Deploy core cloud environments || 1.0<br />
|-<br />
| Stability check || Early operations || 0.5<br />
|-<br />
| Automation study || Automate deployment/tasks || 1.5<br />
|-<br />
| Develop templates || 6 templates, 8 envs, 2 clouds/OS || 1.5<br />
|-<br />
| Migration diagram || Process illustration || 1.0<br />
|-<br />
| Write migration code || 138 lines (see MigrationApp.sh) || 1.5<br />
|-<br />
| Process stabilization || Ensure repeatable migration || 1.5<br />
|-<br />
| Cloud benchmarking || Performance test vs legacy || 1.5<br />
|-<br />
| Downtime calibration || Per-migration time calculation || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs each || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 man-days<br />
|}<br />
<br />
==== Stability check ====<br />
<br />
{| class="wikitable"<br />
! Action !! Expected Result<br />
|-<br />
| Power off one node || All resources started<br />
|-<br />
| Power off/on all nodes simultaneously || All resources started<br />
|}<br />
<br />
----<br />
<br />
= Web Infrastructure & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Web Architecture Reference]]<br />
<br />
* Favor minimal, flexible infrastructure<br />
* Track customer location via GDNS or similar<br />
* Use network load balancers (LVS, IPVS) for scaling<br />
* Compare prices and beware of vendor lock-in<br />
* For TLS: use HAProxy for fast frontend, Envoy for compatibility<br />
* Caching: Varnish, Apache Traffic Server for large content<br />
* Prefer open-source stacks and database caches (e.g. Memcached)<br />
* Use message queues and buffers for workload smoothing<br />
* For more examples: [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture], [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Major Cloud Platforms: Feature Comparison =<br />
<br />
{| class="wikitable"<br />
! Function !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment Tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap Method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router Control''' || Kube-router || Router/Subnet API || Route Table/Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall Control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux Firewall || Linux Firewall || Linux Firewall || API<br />
|-<br />
| '''Network Virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage Options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
----<br />
<br />
= Useful Cloud & IT Links =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM – SRE Tools]<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Freelance IT Rates]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)]<br />
<br />
----<br />
<br />
= Advanced: High-Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA Cluster Architecture]]<br />
<br />
* Multi-node or dual-room clusters for redundancy<br />
* Use IPMI for fencing, provision via PXE/NTP/DNS/TFTP<br />
* For 2-node clusters: stagger fencing for stability; 3+ nodes recommended<br />
<br />
=== Common Resources Pattern ===<br />
* Multipath storage, LUN, LVM, NFS<br />
* User and process resources<br />
* IP, DNS, Listener management<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|HPC Cluster Overview]]<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps Reference Design]]<br />
<br />
----<br />
<br />
'''For more examples, guides, and scripts, visit [https://infocepo.com infocepo.com]. Contributions and suggestions welcome!'''</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1896
Main Page
2025-11-24T10:27:03Z
<p>Tcepo: /* Discover Cloud Computing on infocepo.com */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud computing on infocepo.com]]<br />
<br />
= Discover Cloud Computing on infocepo.com =<br />
<br />
Welcome! This portal is designed for IT professionals, engineers, students, and enthusiasts who want to master cloud infrastructure, explore AI tools, and accelerate their IT skills through hands-on labs and open-source solutions.<br />
<br />
__TOC__<br />
<br />
<br />
== Quick Start ==<br />
* '''Master cloud infrastructure:''' Practical guides and labs<br />
* '''Explore artificial intelligence:''' Trends and hands-on tools<br />
* '''Compare cloud providers:''' Kubernetes, AWS, OpenStack, and more<br />
* '''Develop expertise:''' Training, open-source, and real-world projects<br />
<br />
----<br />
<br />
= AI & Cloud Tools =<br />
<br />
; '''AI Assistants'''<br />
* [https://chat.openai.com ChatGPT] – Public conversational AI with strong learning capabilities<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] – Private assistants and self-hosted LLM APIs<br />
* [https://github.com/ynotopec/summarize Private summary] – Fast, offline summarizer for your data<br />
<br />
; '''Development & Model Tracking'''<br />
* [https://ollama.com/library LLM Trending] – Latest open-source LLMs<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Top trending codebases since 2022<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Compare chatbot performance<br />
* [https://www.perplexity.ai Perplexity AI] – Cutting-edge research and question answering<br />
* [https://huggingface.co/models Models Trending] – Model marketplace<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Advanced training framework<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Ranking for vector search models<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Database speed and feature comparison<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – HPC/AI GPUs for Kubernetes clusters<br />
* NVIDIA 5080 – Consumer GPU for private deployments<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Vision-language models<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Compare generative image models<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Top green supercomputers<br />
<br />
----<br />
<br />
== Notable Open LLMs ==<br />
''(Last updated: 26/07/2025)''<br />
<br />
{| class="wikitable"<br />
! Model !! Description / Notable Features<br />
|-<br />
| '''ai-chat''' || gemma3n, cost efficient<br />
|-<br />
| '''ai-translate''' || gemma3n, temperature=0 (deterministic translation)<br />
|-<br />
| '''ai-summary''' || qwen2.5, optimized for summarization<br />
|-<br />
| '''ai-code''' || mistral-small, advanced code reasoning<br />
|-<br />
| '''ai-code-completion''' || granite-2b, fast code suggestions<br />
|-<br />
| '''ai-parse''' || gemma2-simpo, parsing & extraction<br />
|-<br />
| '''ai-RAG-FR''' || qwen2.5, French RAG applications<br />
|-<br />
| '''mannix/gemma2-9b-simpo''' || OllamaFunctions integration<br />
|}<br />
<br />
----<br />
<br />
= Industry News & Trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Video digest<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Speaker tracking<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Modern open-source translation<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Fast, low-cost inference hardware<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Enhanced search experiences<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Intro to Transformers algorithm<br />
* Hands-on labs and scripts in the [[LAB project|CLOUD LAB]] below<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab Reference Diagram]]<br />
<br />
; '''Lab Project''' <br />
Experiment with high-availability, cloud migration, and audit automation.<br />
<br />
=== Cloud Audit ===<br />
* '''[[ServerDiff.sh]]''' – Bash script for auditing servers, tracking config drift, and checking environment consistency<br />
<br />
=== Cloud Migration Example ===<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud Migration Diagram]]<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Audit infrastructure || 82 services, automated via ServerDiff.sh || 1.5<br />
|-<br />
| Diagram cloud architecture || Visual design || 1.5<br />
|-<br />
| Compliance check || 2 clouds, 6 hypervisors, 6TB RAM || 1.5<br />
|-<br />
| Install cloud platforms || Deploy core cloud environments || 1.0<br />
|-<br />
| Stability check || Early operations || 0.5<br />
|-<br />
| Automation study || Automate deployment/tasks || 1.5<br />
|-<br />
| Develop templates || 6 templates, 8 envs, 2 clouds/OS || 1.5<br />
|-<br />
| Migration diagram || Process illustration || 1.0<br />
|-<br />
| Write migration code || 138 lines (see MigrationApp.sh) || 1.5<br />
|-<br />
| Process stabilization || Ensure repeatable migration || 1.5<br />
|-<br />
| Cloud benchmarking || Performance test vs legacy || 1.5<br />
|-<br />
| Downtime calibration || Per-migration time calculation || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs each || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 man-days<br />
|}<br />
<br />
==== Stability check ====<br />
<br />
{| class="wikitable"<br />
! Action !! Expected Result<br />
|-<br />
| Power off one node || All resources started<br />
|-<br />
| Power off/on all nodes simultaneously || All resources started<br />
|}<br />
<br />
----<br />
<br />
= Web Infrastructure & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Web Architecture Reference]]<br />
<br />
* Favor minimal, flexible infrastructure<br />
* Track customer location via GDNS or similar<br />
* Use network load balancers (LVS, IPVS) for scaling<br />
* Compare prices and beware of vendor lock-in<br />
* For TLS: use HAProxy for fast frontend, Envoy for compatibility<br />
* Caching: Varnish, Apache Traffic Server for large content<br />
* Prefer open-source stacks and database caches (e.g. Memcached)<br />
* Use message queues and buffers for workload smoothing<br />
* For more examples: [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture], [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Major Cloud Platforms: Feature Comparison =<br />
<br />
{| class="wikitable"<br />
! Function !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment Tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap Method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router Control''' || Kube-router || Router/Subnet API || Route Table/Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall Control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux Firewall || Linux Firewall || Linux Firewall || API<br />
|-<br />
| '''Network Virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage Options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
----<br />
<br />
= Useful Cloud & IT Links =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM – SRE Tools]<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Freelance IT Rates]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)]<br />
<br />
----<br />
<br />
= Advanced: High-Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA Cluster Architecture]]<br />
<br />
* Multi-node or dual-room clusters for redundancy<br />
* Use IPMI for fencing, provision via PXE/NTP/DNS/TFTP<br />
* For 2-node clusters: stagger fencing for stability; 3+ nodes recommended<br />
<br />
=== Common Resources Pattern ===<br />
* Multipath storage, LUN, LVM, NFS<br />
* User and process resources<br />
* IP, DNS, Listener management<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|HPC Cluster Overview]]<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps Reference Design]]<br />
<br />
----<br />
<br />
'''For more examples, guides, and scripts, visit [https://infocepo.com infocepo.com]. Contributions and suggestions welcome!'''</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1895
Main Page
2025-10-07T18:13:59Z
<p>Tcepo: /* AI & Cloud Tools */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud computing on infocepo.com]]<br />
<br />
= Discover Cloud Computing on infocepo.com =<br />
<br />
Welcome! This portal is designed for IT professionals, engineers, students, and enthusiasts who want to master cloud infrastructure, explore AI tools, and accelerate their IT skills through hands-on labs and open-source solutions.<br />
<br />
__TOC__<br />
<br />
== Quick Start ==<br />
* '''Master cloud infrastructure:''' Practical guides and labs<br />
* '''Explore artificial intelligence:''' Trends and hands-on tools<br />
* '''Compare cloud providers:''' Kubernetes, AWS, OpenStack, and more<br />
* '''Develop expertise:''' Training, open-source, and real-world projects<br />
<br />
----<br />
<br />
= AI & Cloud Tools =<br />
<br />
; '''AI Assistants'''<br />
* [https://chat.openai.com ChatGPT] – Public conversational AI with strong learning capabilities<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] – Private assistants and self-hosted LLM APIs<br />
* [https://github.com/ynotopec/summarize Private summary] – Fast, offline summarizer for your data<br />
<br />
; '''Development & Model Tracking'''<br />
* [https://ollama.com/library LLM Trending] – Latest open-source LLMs<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Top trending codebases since 2022<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Compare chatbot performance<br />
* [https://www.perplexity.ai Perplexity AI] – Cutting-edge research and question answering<br />
* [https://huggingface.co/models Models Trending] – Model marketplace<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Advanced training framework<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Ranking for vector search models<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Database speed and feature comparison<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – HPC/AI GPUs for Kubernetes clusters<br />
* NVIDIA 5080 – Consumer GPU for private deployments<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Vision-language models<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Compare generative image models<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Top green supercomputers<br />
<br />
----<br />
<br />
== Notable Open LLMs ==<br />
''(Last updated: 26/07/2025)''<br />
<br />
{| class="wikitable"<br />
! Model !! Description / Notable Features<br />
|-<br />
| '''ai-chat''' || gemma3n, cost efficient<br />
|-<br />
| '''ai-translate''' || gemma3n, temperature=0 (deterministic translation)<br />
|-<br />
| '''ai-summary''' || qwen2.5, optimized for summarization<br />
|-<br />
| '''ai-code''' || mistral-small, advanced code reasoning<br />
|-<br />
| '''ai-code-completion''' || granite-2b, fast code suggestions<br />
|-<br />
| '''ai-parse''' || gemma2-simpo, parsing & extraction<br />
|-<br />
| '''ai-RAG-FR''' || qwen2.5, French RAG applications<br />
|-<br />
| '''mannix/gemma2-9b-simpo''' || OllamaFunctions integration<br />
|}<br />
<br />
----<br />
<br />
= Industry News & Trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Video digest<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Speaker tracking<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Modern open-source translation<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Fast, low-cost inference hardware<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Enhanced search experiences<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Intro to Transformers algorithm<br />
* Hands-on labs and scripts in the [[LAB project|CLOUD LAB]] below<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab Reference Diagram]]<br />
<br />
; '''Lab Project''' <br />
Experiment with high-availability, cloud migration, and audit automation.<br />
<br />
=== Cloud Audit ===<br />
* '''[[ServerDiff.sh]]''' – Bash script for auditing servers, tracking config drift, and checking environment consistency<br />
<br />
=== Cloud Migration Example ===<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud Migration Diagram]]<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Audit infrastructure || 82 services, automated via ServerDiff.sh || 1.5<br />
|-<br />
| Diagram cloud architecture || Visual design || 1.5<br />
|-<br />
| Compliance check || 2 clouds, 6 hypervisors, 6TB RAM || 1.5<br />
|-<br />
| Install cloud platforms || Deploy core cloud environments || 1.0<br />
|-<br />
| Stability check || Early operations || 0.5<br />
|-<br />
| Automation study || Automate deployment/tasks || 1.5<br />
|-<br />
| Develop templates || 6 templates, 8 envs, 2 clouds/OS || 1.5<br />
|-<br />
| Migration diagram || Process illustration || 1.0<br />
|-<br />
| Write migration code || 138 lines (see MigrationApp.sh) || 1.5<br />
|-<br />
| Process stabilization || Ensure repeatable migration || 1.5<br />
|-<br />
| Cloud benchmarking || Performance test vs legacy || 1.5<br />
|-<br />
| Downtime calibration || Per-migration time calculation || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs each || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 man-days<br />
|}<br />
<br />
==== Stability check ====<br />
<br />
{| class="wikitable"<br />
! Action !! Expected Result<br />
|-<br />
| Power off one node || All resources started<br />
|-<br />
| Power off/on all nodes simultaneously || All resources started<br />
|}<br />
<br />
----<br />
<br />
= Web Infrastructure & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Web Architecture Reference]]<br />
<br />
* Favor minimal, flexible infrastructure<br />
* Track customer location via GDNS or similar<br />
* Use network load balancers (LVS, IPVS) for scaling<br />
* Compare prices and beware of vendor lock-in<br />
* For TLS: use HAProxy for fast frontend, Envoy for compatibility<br />
* Caching: Varnish, Apache Traffic Server for large content<br />
* Prefer open-source stacks and database caches (e.g. Memcached)<br />
* Use message queues and buffers for workload smoothing<br />
* For more examples: [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture], [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Major Cloud Platforms: Feature Comparison =<br />
<br />
{| class="wikitable"<br />
! Function !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment Tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap Method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router Control''' || Kube-router || Router/Subnet API || Route Table/Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall Control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux Firewall || Linux Firewall || Linux Firewall || API<br />
|-<br />
| '''Network Virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage Options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
----<br />
<br />
= Useful Cloud & IT Links =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM – SRE Tools]<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Freelance IT Rates]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)]<br />
<br />
----<br />
<br />
= Advanced: High-Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA Cluster Architecture]]<br />
<br />
* Multi-node or dual-room clusters for redundancy<br />
* Use IPMI for fencing, provision via PXE/NTP/DNS/TFTP<br />
* For 2-node clusters: stagger fencing for stability; 3+ nodes recommended<br />
<br />
=== Common Resources Pattern ===<br />
* Multipath storage, LUN, LVM, NFS<br />
* User and process resources<br />
* IP, DNS, Listener management<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|HPC Cluster Overview]]<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps Reference Design]]<br />
<br />
----<br />
<br />
'''For more examples, guides, and scripts, visit [https://infocepo.com infocepo.com]. Contributions and suggestions welcome!'''</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Main_Page&diff=1894
Main Page
2025-10-07T13:38:24Z
<p>Tcepo: /* AI & Cloud Tools */</p>
<hr />
<div>[[File:Infocepo-picture.png|thumb|right|Discover cloud computing on infocepo.com]]<br />
<br />
= Discover Cloud Computing on infocepo.com =<br />
<br />
Welcome! This portal is designed for IT professionals, engineers, students, and enthusiasts who want to master cloud infrastructure, explore AI tools, and accelerate their IT skills through hands-on labs and open-source solutions.<br />
<br />
__TOC__<br />
<br />
== Quick Start ==<br />
* '''Master cloud infrastructure:''' Practical guides and labs<br />
* '''Explore artificial intelligence:''' Trends and hands-on tools<br />
* '''Compare cloud providers:''' Kubernetes, AWS, OpenStack, and more<br />
* '''Develop expertise:''' Training, open-source, and real-world projects<br />
<br />
----<br />
<br />
= AI & Cloud Tools =<br />
<br />
; '''AI Assistants'''<br />
* [https://chat.openai.com ChatGPT4] – Public conversational AI with strong learning capabilities<br />
* [https://github.com/open-webui/open-webui Open WebUI] + [https://www.scaleway.com/en/h100-pcie-try-it-now/ GPU H100] + [https://ollama.com Ollama] – Private assistants and self-hosted LLM APIs<br />
* [https://github.com/ynotopec/summarize Private summary] – Fast, offline summarizer for your data<br />
<br />
; '''Development & Model Tracking'''<br />
* [https://ollama.com/library LLM Trending] – Latest open-source LLMs<br />
* [https://github.com/search?q=stars%3A%3E15000+forks%3A%3E1500+created%3A%3E2022-06-01&type=repositories&s=updated&o=desc Project Trending] – Top trending codebases since 2022<br />
* [https://lmarena.ai/leaderboard ChatBot Evaluation] – Compare chatbot performance<br />
* [https://www.perplexity.ai Perplexity AI] – Cutting-edge research and question answering<br />
* [https://huggingface.co/models Models Trending] – Model marketplace<br />
* [https://github.com/hiyouga/LLaMA-Factory LLM Fine Tuning] – Advanced training framework<br />
* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard] – Ranking for vector search models<br />
* [https://ann-benchmarks.com Vectors DB Ranking] – Database speed and feature comparison<br />
* [https://www.nvidia.com/en-us/data-center/h100/ NVIDIA H100] – HPC/AI GPUs for Kubernetes clusters<br />
* NVIDIA 5080 – Consumer GPU for private deployments<br />
* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending] – Vision-language models<br />
* [https://huggingface.co/spaces/TIGER-Lab/GenAI-Arena Txt2img Evaluation] – Compare generative image models<br />
* [https://top500.org/lists/green500/ HPC Efficiency] – Top green supercomputers<br />
<br />
----<br />
<br />
== Notable Open LLMs ==<br />
''(Last updated: 26/07/2025)''<br />
<br />
{| class="wikitable"<br />
! Model !! Description / Notable Features<br />
|-<br />
| '''ai-chat''' || gemma3n, cost efficient<br />
|-<br />
| '''ai-translate''' || gemma3n, temperature=0 (deterministic translation)<br />
|-<br />
| '''ai-summary''' || qwen2.5, optimized for summarization<br />
|-<br />
| '''ai-code''' || mistral-small, advanced code reasoning<br />
|-<br />
| '''ai-code-completion''' || granite-2b, fast code suggestions<br />
|-<br />
| '''ai-parse''' || gemma2-simpo, parsing & extraction<br />
|-<br />
| '''ai-RAG-FR''' || qwen2.5, French RAG applications<br />
|-<br />
| '''mannix/gemma2-9b-simpo''' || OllamaFunctions integration<br />
|}<br />
<br />
----<br />
<br />
= Industry News & Trends =<br />
<br />
* [https://www.youtube.com/@lev-selector/videos Top AI News] – Video digest<br />
* [https://betterprogramming.pub/color-your-captions-streamlining-live-transcriptions-with-diart-and-openais-whisper-6203350234ef Real-time transcription with Diart + Whisper] – Speaker tracking<br />
* [https://github.com/openai-translator/openai-translator OpenAI Translator] – Modern open-source translation<br />
* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator] – Fast, low-cost inference hardware<br />
* [https://opensearch.org/docs/latest/search-plugins/conversational-search Opensearch with LLM] – Enhanced search experiences<br />
<br />
----<br />
<br />
= Training & Learning =<br />
<br />
* [https://www.youtube.com/watch?v=4Bdc55j80l8 Transformers Explained] – Intro to Transformers algorithm<br />
* Hands-on labs and scripts in the [[LAB project|CLOUD LAB]] below<br />
<br />
----<br />
<br />
= Cloud Lab & Audit Projects =<br />
<br />
[[File:Infocepo.drawio.png|400px|Cloud Lab Reference Diagram]]<br />
<br />
; '''Lab Project''' <br />
Experiment with high-availability, cloud migration, and audit automation.<br />
<br />
=== Cloud Audit ===<br />
* '''[[ServerDiff.sh]]''' – Bash script for auditing servers, tracking config drift, and checking environment consistency<br />
<br />
=== Cloud Migration Example ===<br />
[[File:Diagram-migration-ORACLE-KVM-v2.drawio.png|400px|Cloud Migration Diagram]]<br />
<br />
{| class="wikitable"<br />
! Task !! Description !! Duration (days)<br />
|-<br />
| Audit infrastructure || 82 services, automated via ServerDiff.sh || 1.5<br />
|-<br />
| Diagram cloud architecture || Visual design || 1.5<br />
|-<br />
| Compliance check || 2 clouds, 6 hypervisors, 6TB RAM || 1.5<br />
|-<br />
| Install cloud platforms || Deploy core cloud environments || 1.0<br />
|-<br />
| Stability check || Early operations || 0.5<br />
|-<br />
| Automation study || Automate deployment/tasks || 1.5<br />
|-<br />
| Develop templates || 6 templates, 8 envs, 2 clouds/OS || 1.5<br />
|-<br />
| Migration diagram || Process illustration || 1.0<br />
|-<br />
| Write migration code || 138 lines (see MigrationApp.sh) || 1.5<br />
|-<br />
| Process stabilization || Ensure repeatable migration || 1.5<br />
|-<br />
| Cloud benchmarking || Performance test vs legacy || 1.5<br />
|-<br />
| Downtime calibration || Per-migration time calculation || 0.5<br />
|-<br />
| VM loading || 82 VMs: OS, code, 2 IPs each || 0.1<br />
|-<br />
! colspan=2 align="right"| '''Total''' !! 15 man-days<br />
|}<br />
<br />
==== Stability check ====<br />
<br />
{| class="wikitable"<br />
! Action !! Expected Result<br />
|-<br />
| Power off one node || All resources started<br />
|-<br />
| Power off/on all nodes simultaneously || All resources started<br />
|}<br />
<br />
----<br />
<br />
= Web Infrastructure & Best Practices =<br />
<br />
[[File:WebModelDiagram.drawio.png|400px|Web Architecture Reference]]<br />
<br />
* Favor minimal, flexible infrastructure<br />
* Track customer location via GDNS or similar<br />
* Use network load balancers (LVS, IPVS) for scaling<br />
* Compare prices and beware of vendor lock-in<br />
* For TLS: use HAProxy for fast frontend, Envoy for compatibility<br />
* Caching: Varnish, Apache Traffic Server for large content<br />
* Prefer open-source stacks and database caches (e.g. Memcached)<br />
* Use message queues and buffers for workload smoothing<br />
* For more examples: [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Architecture], [https://github.com/systemdesign42/system-design System Design GitHub]<br />
<br />
----<br />
<br />
= Major Cloud Platforms: Feature Comparison =<br />
<br />
{| class="wikitable"<br />
! Function !! Kubernetes !! OpenStack !! AWS !! Bare-metal !! HPC !! CRM !! oVirt<br />
|-<br />
| '''Deployment Tools''' || Helm, YAML, ArgoCD, Juju || Ansible, Terraform, Juju || CloudFormation, Terraform, Juju || Ansible, Shell || xCAT, Clush || Ansible, Shell || Ansible, Python<br />
|-<br />
| '''Bootstrap Method''' || API || API, PXE || API || PXE, IPMI || PXE, IPMI || PXE, IPMI || PXE, API<br />
|-<br />
| '''Router Control''' || Kube-router || Router/Subnet API || Route Table/Subnet API || Linux, OVS || xCAT || Linux || API<br />
|-<br />
| '''Firewall Control''' || Istio, NetworkPolicy || Security Groups API || Security Group API || Linux Firewall || Linux Firewall || Linux Firewall || API<br />
|-<br />
| '''Network Virtualization''' || VLAN, VxLAN, others || VPC || VPC || OVS, Linux || xCAT || Linux || API<br />
|-<br />
| '''DNS''' || CoreDNS || DNS-Nameserver || Route 53 || GDNS || xCAT || Linux || API<br />
|-<br />
| '''Load Balancer''' || Kube-proxy, LVS || LVS || Network Load Balancer || LVS || SLURM || Ldirectord || N/A<br />
|-<br />
| '''Storage Options''' || Local, Cloud, PVC || Swift, Cinder, Nova || S3, EFS, EBS, FSx || Swift, XFS, EXT4, RAID10 || GPFS || SAN || NFS, SAN<br />
|}<br />
<br />
----<br />
<br />
= Useful Cloud & IT Links =<br />
<br />
* [https://cloud.google.com/free/docs/aws-azure-gcp-service-comparison Cloud Providers Compared]<br />
* [https://global-internet-map-2021.telegeography.com/ Global Internet Topology Map]<br />
* [https://landscape.cncf.io/?fullscreen=yes CNCF Official Landscape]<br />
* [https://wikitech.wikimedia.org/wiki/Wikimedia_infrastructure Wikimedia Cloud Wiki]<br />
* [https://openapm.io OpenAPM – SRE Tools]<br />
* [https://access.redhat.com/downloads/content/package-browser RedHat Package Browser]<br />
* [https://www.silkhom.com/barometre-2021-des-tjm-dans-informatique-digital Freelance IT Rates]<br />
* [https://www.glassdoor.fr/salaire/Hays-Salaires-E10166.htm IT Salaries (Glassdoor)]<br />
<br />
----<br />
<br />
= Advanced: High-Availability, HPC & DevSecOps =<br />
<br />
== High Availability with Corosync & Pacemaker ==<br />
[[File:HA-REF.drawio.png|400px|HA Cluster Architecture]]<br />
<br />
* Multi-node or dual-room clusters for redundancy<br />
* Use IPMI for fencing, provision via PXE/NTP/DNS/TFTP<br />
* For 2-node clusters: stagger fencing for stability; 3+ nodes recommended<br />
<br />
=== Common Resources Pattern ===<br />
* Multipath storage, LUN, LVM, NFS<br />
* User and process resources<br />
* IP, DNS, Listener management<br />
<br />
== HPC ==<br />
[[File:HPC.drawio.png|400px|HPC Cluster Overview]]<br />
<br />
== DevSecOps ==<br />
[[File:DSO-POC-V3.drawio.png|400px|DevSecOps Reference Design]]<br />
<br />
----<br />
<br />
'''For more examples, guides, and scripts, visit [https://infocepo.com infocepo.com]. Contributions and suggestions welcome!'''</div>
Tcepo
https://infocepo.com/wiki/index.php?title=Free_time&diff=1893
Free time
2025-09-12T07:07:09Z
<p>Tcepo: Created page with "== Proto-indice “Temps pour réfléchir” (TPR) == Je te propose un **proto-indice “Temps pour réfléchir” (TPR)** en combinant 2 sources OCDE : * **Temps de loisirs & soins personnels** (heures/jour) — plus c’est élevé, mieux c’est. (1) * **Heures travaillées/an** — moins c’est élevé, mieux c'est. (2) === Méthode === 1. Normalisation 0–100 de chaque indicateur par pays. 2. **Score TPR = 60% (l..."</p>
<hr />
<div>== Proto-indice “Temps pour réfléchir” (TPR) ==<br />
<br />
Je te propose un **proto-indice “Temps pour réfléchir” (TPR)** en combinant 2 sources OCDE :<br />
<br />
* **Temps de loisirs & soins personnels** (heures/jour) — plus c’est élevé, mieux c’est. ([[OECD Time Use Database|1]])<br />
* **Heures travaillées/an** — moins c’est élevé, mieux c'est. ([[OECD Hours Worked|2]])<br />
<br />
=== Méthode ===<br />
1. Normalisation 0–100 de chaque indicateur par pays.<br />
2. **Score TPR = 60% (loisirs) + 40% (inverse des heures/an)**.<br />
3. Données les plus récentes accessibles : loisirs OCDE (vagues 2000–2020 selon pays), heures/an 2023–2024. (Les enquêtes “loisirs” ne sont pas annuelles.)<br />
<br />
=== Classement synthétique (indicatif) – grands ordres ===<br />
*(Top = davantage de “temps pour réfléchir” disponible ; Bottom = moins. Basé sur les profils OCDE connus : hauts loisirs + faibles heures/an en tête.)*<br />
<br />
==== Top \~10 ====<br />
# [[Danemark]]<br />
# [[Norvège]]<br />
# [[Pays-Bas]]<br />
# [[Allemagne]]<br />
# [[Suède]]<br />
# [[Finlande]]<br />
# [[France]]<br />
# [[Autriche]]<br />
# [[Belgique]]<br />
# [[Espagne]]<br />
<br />
==== Milieu de tableau (\~bonne moyenne) ====<br />
* [[Royaume-Uni]], [[Canada]], [[Nouvelle-Zélande]], [[Irlande]], [[Suisse]], [[Portugal]], [[Italie]], [[Tchéquie]]<br />
<br />
==== Bas de tableau (TPR plus faible) ====<br />
* [[États-Unis]], [[Japon]], [[Corée du Sud]], [[Turquie]], [[Mexique]]<br />
<br />
> Raison des positions : les pays nordiques et d’Europe occidentale cumulent **beaucoup de temps “loisirs & soins personnels”** et **faibles heures/an**; à l’inverse, **Mexique / Corée / Turquie** ont des **heures/an élevées** et un **temps de loisirs moindre**. ([[List of countries by time devoted to leisure and personal care|3]])<br />
<br />
=== Ce que je peux faire tout de suite pour toi ===<br />
* Te produire un **tableau chiffré** (TPR, heures de loisirs/j, heures travaillées/an) pour **les pays que tu cibles** (ex. UE27 + US/JP/KR/MX), avec les **citations exactes par ligne**.<br />
* Ajouter un **volet “temps de trajet domicile-travail”** (Eurostat/World Bank) si tu veux raffiner l’indice.<br />
<br />
Tu me dis les **pays à inclure** et je te sors le tableau comparatif prêt à l’emploi.<br />
<br />
=== Références ===<br />
# <ref>[https://www.oecd.org/en/data/datasets/time-use-database.html?utm_source=chatgpt.com Time use database]</ref><br />
# <ref>[https://www.oecd.org/en/data/indicators/hours-worked.html?utm_source=chatgpt.com Hours worked]</ref><br />
# <ref>[https://en.wikipedia.org/wiki/List_of_countries_by_time_devoted_to_leisure_and_personal_care?utm_source=chatgpt.com List of countries by time devoted to leisure and personal care]</ref></div>
Tcepo