Essential - Recent changes [en]

Main Page

Tcepo — Sat, 16 May 2026 01:37:09 GMT

← Older revision		Revision as of 03:37, 16 May 2026
(5 intermediate revisions by the same user not shown)
Line 86:		Line 86:
	\| API \|\| [https://api-audio2txt.ailab.infocepo.com/docs STT] \|\| Transcription audio		\| API \|\| [https://api-audio2txt.ailab.infocepo.com/docs STT] \|\| Transcription audio
	\|-		\|-
	\| API \|\| [https://api-~~txt2audio~~.ailab.infocepo.com/docs TTS] \|\| Synthèse vocale		\| API \|\| [https://api-tts-omnivoice.ailab.infocepo.com/docs TTS] \|\| Synthèse vocale
	\|-		\|-
	\| API \|\| [https://github.com/ynotopec/api-realtime-ai realtime-ai] \|\| Temps réel WebSocket / WebRTC		\| API \|\| [https://github.com/ynotopec/api-realtime-ai realtime-ai] \|\| Temps réel WebSocket / WebRTC
Line 122:		Line 122:

	== Nouveautés 26/04/2026 ==		== Nouveautés 26/04/2026 ==
			* compatibilité Open WebUI avec Agentic RAG
			* upgrade [https://translate-rt.ailab.infocepo.com '''realtime translation'''] : réduction significative des hallucinations lors des silences, diminution de la latence et ajout de la plupart des langues en TTS
	* ajout de [https://api-tts-omnivoice.ailab.infocepo.com '''TTS Omnivoice'''] : Qualité TTS augmenté et ajout plus global des langues (600)		* ajout de [https://api-tts-omnivoice.ailab.infocepo.com '''TTS Omnivoice'''] : Qualité TTS augmenté et ajout plus global des langues (600)
	* ajout de [https://api-lightrag.ailab.infocepo.com '''lightRAG'''] : LightRAG est un framework RAG avancé et léger qui combine graphs de connaissances et recherche vectorielle pour une analyse contextuelle profonde et efficace.		* ajout de [https://api-lightrag.ailab.infocepo.com '''lightRAG'''] : LightRAG est un framework RAG avancé et léger qui combine graphs de connaissances et recherche vectorielle pour une analyse contextuelle profonde et efficace.
Line 156:		Line 158:

	== Backlog / veille ==		== Backlog / veille ==
	* OPENRAG > implement / evaluate / add OIDC
	* short audio transcription
	* translation latency > [https://github.com/ynotopec/api-realtime-ai api-realtime-ai]
	* RAG sur PDF avec images		* RAG sur PDF avec images
	* compatibilité Open WebUI avec Agentic RAG
	* scalability		* scalability
	* security > [https://github.com/ynotopec/ai-security ai-security] / [https://github.com/NVIDIA-NeMo/Guardrails NeMo Guardrails]
	* [https://github.com/openclaw/openclaw openclaw]		* [https://github.com/openclaw/openclaw openclaw]
	* faster-whisper mutualisé		* faster-whisper mutualisé
Line 174:		Line 171:
	* [https://github.com/vllm-project/semantic-router semantic-router]		* [https://github.com/vllm-project/semantic-router semantic-router]
	* [https://github.com/KeygraphHQ/shannon Shannon]		* [https://github.com/KeygraphHQ/shannon Shannon]
	* [https://huggingface.co/Qwen/Qwen3-ASR-1.7B Qwen3-ASR-1.7B]
	* [https://huggingface.co/tencent/Youtu-VL-4B-Instruct Youtu-VL-4B-Instruct]
	* [https://huggingface.co/stepfun-ai/Step3-VL-10B Step3-VL-10B]
	* [https://huggingface.co/Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice Qwen3-TTS-12Hz-1.7B-CustomVoice]
	* [https://github.com/resemble-ai/chatterbox chatterbox]		* [https://github.com/resemble-ai/chatterbox chatterbox]
	* deepset-ai/haystack		* deepset-ai/haystack
Line 188:		Line 181:
	* N8N		* N8N
	* API Compressor		* API Compressor
	* LightRAG
	* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]		* [https://huggingface.co/Qwen/Qwen3-Omni-30B-A3B-Instruct Qwen3-Omni-30B-A3B-Instruct]
	* Metabase		* Metabase
Line 473:		Line 465:
	= API TTS =		= API TTS =

	* URL : <code>https://api-~~txt2audio~~.ailab.infocepo.com/v1</code>		* URL : <code>https://api-tts-omnivoice.ailab.infocepo.com/v1</code>
	* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code>		* Clé : <code>OPENAI_API_KEY=sk-XXXXX</code>
	* Documentation : [https://tts.ailab.infocepo.com:wait-2026-06/docs API TTS docs]		* Documentation : [https://tts.ailab.infocepo.com:wait-2026-06/docs API TTS docs]
Line 481:		Line 473:
	export OPENAI_API_KEY=sk-XXXXX		export OPENAI_API_KEY=sk-XXXXX

	curl https://api-~~txt2audio~~.ailab.infocepo.com/v1/audio/speech \		curl https://api-tts-omnivoice.ailab.infocepo.com/v1/audio/speech \
	-H "Authorization: Bearer $OPENAI_API_KEY" \		-H "Authorization: Bearer $OPENAI_API_KEY" \
	-H "Content-Type: application/json" \		-H "Content-Type: application/json" \

Човешки нужди

Tcepo — Sun, 03 May 2026 17:15:27 GMT

Created page with "= Основни обекти на цивилизацията и материални зависимости = == Въведение == Тази статия представя структуриран поглед върху фундаменталните обекти, необходими за една модерна цивилизация, техните материални зависимости, както и научните и технологични основ..."

New page

= Основни обекти на цивилизацията и материални зависимости =

== Въведение ==
Тази статия представя структуриран поглед върху фундаменталните обекти, необходими за една модерна цивилизация, техните материални зависимости, както и научните и технологични основи, които ги правят възможни.
Тя свързва всекидневните нужди (жилище, хигиена, транспорт, инструменти), информационните системи (обработка, съхранение, предаване), материалните ограничения (изобилни срещу редки елементи) и незаменимите научни открития.

== 1. Фундаментални обекти на цивилизацията (g1) ==

=== 1.1 Физически и материални нужди ===
* '''[https://en.wikipedia.org/wiki/Shelter Жилище]'''
* '''Структурирани елементарни частици'''
* '''[https://en.wikipedia.org/wiki/Clothes Облекло]'''
* '''[https://en.wikipedia.org/wiki/Toilet Тоалетна]'''
* '''[https://en.wikipedia.org/wiki/Cleaner Чистящи средства]'''
* '''[https://en.wikipedia.org/wiki/Transportation Транспорт]'''
* '''[https://en.wikipedia.org/wiki/Defense Защита/Отбрана]'''
* '''[https://en.wikipedia.org/wiki/Help Помощ]'''
* '''[https://en.wikipedia.org/wiki/Cutlery Бърдо]'''
* '''[https://en.wikipedia.org/wiki/Mattress Матрак]'''
* '''[https://en.wikipedia.org/wiki/Table_(furniture) Маса (меки мебели)]'''
* '''[https://en.wikipedia.org/wiki/Chair stol]'''

=== 1.2 Отдих и обучение/създаване ===
* '''[https://en.wikipedia.org/wiki/Leisure Отдих]'''
* '''[https://en.wikipedia.org/wiki/Training Обучение]'''
* '''[https://en.wikipedia.org/wiki/Stimulants Стимуланти]'''

== 2. Информационна верига (функционални обекти) ==

=== 2.1 Преобразуване на информация ===
* '''[https://en.wikipedia.org/wiki/Transducer Информациян трансдуктор]'''
* '''[https://en.wikipedia.org/wiki/Information_processing Обработка на информация]'''

=== 2.2 Съхранение, търсене и предаване ===
* '''[https://en.wikipedia.org/wiki/Data_storage Съхранение на данни]'''
* '''[https://en.wikipedia.org/wiki/Search_engine Търсачка]'''
* '''[https://en.wikipedia.org/wiki/Transmitter Информациян предавател]'''

== 3. Материални зависимости: Изобилни (AU) срещу Стратегически редки (RS) ==

=== 3.1 Легенда ===
* '''AU''' = Изобилни/Универсални елементи
(CHON + Si, Al, Fe, Ca, Mg, Na, K + S, P, Cl + Cu, Zn)
* '''RS''' = Редки/Стратегически елементи
(Cr, Ni, Ti, Mn, B, F, Li + Nd/Pr/Dy, Co, W, Ag, Au, Ga, As…)
* Оценка: 0–5 (5 = критично важно, 0 = незначително)

=== 3.2 Таблица на зависимостите по обекти (g1) ===
{| class="wikitable sortable"
! Елемент (g1)
! AU (елементи + оценка)
! RS (елементи + оценка)
! Бележки
|-
| Жилище
| Si,O,Al,Ca,Fe,Mg,C,H (5)
| Cr,Ni,Ti,B (3)
| Бетон, стъкло, керамика, стомана; сплави, ако са налични
|-
| Структурирани елементарни частици
| C,H,O,N,Si,Fe,Al (5)
| Ti,Cr,Ni,B (2)
| Здрави материални структури без високотехнологично производство
|-
| Облекло
| C,H,O,N,S (5)
| F,Ti (2)
| Влакна и обработки; F за мембрани
|-
| Тоалетна
| Si,O,Al,Ca,Fe,C,H (5)
| Cr,Ni,Cu (3)
| Керамика и водопровод; предпочитат се неръждаема стомана
|-
| Чистящи средства
| H,O,Na,Cl,C,S (5)
| F,P (2)
| Основи, соли, разтворители; P за напреднали перилни препарати
|-
| Транспорт
| Fe,Al,Si,C,O,Cu (5)
| Ni,Cr,Ti,Mn,Li (4)
| Сплави, окабеляване; Li за електрически системи
|-
| Защита/Отбрана
| Fe,C,Al,Si (4)
| Ti,Cr,Ni,W (4)
| W за висока плътност
|-
| Помощ
| Fe,C,H,O,N,Ca,Na,Cl (5)
| Cu,Zn,Ag (2)
| Хигиена и базови инструменти
|-
| Информациян трансдуктор
| Si,O,Al,Fe,C,Cu (4)
| B,Ga (4)
| Сензорите и актуаторите се нуждаят от легиращи елементи (dopants)
|-
| Обработка на информация
| Si,O,Al,C,Cu (4)
| B,P,As,Ga (5)
| Изчисленията зависят от легирането и фините процеси
|-
| Съхранение на данни
| C,H,O (хартия), Fe (магнитни), Si,O (стъкло) (4)
| Co,Ni,Nd (4)
| Високотехнологичното съхранение изисква Co/Ni и магнити
|-
| Търсачка
| Si,Al,Cu,C (4)
| Li,Ni,Co,Nd (4)
| Изчисления, енергия, връзки
|-
| Информациян предавател
| Cu,Al,Si,O (5)
| Ag,Au (2)
| Мед/алуминий + влакно са достатъчни
|-
| Стимуланти
| C,H,O,N,S,P (5)
| — (0)
| Органична химия и хранене
|-
| Отдих
| C,H,O,N,Si,Fe,Al (4)
| Cu,Li,Nd (2)
| Електрониката е по желание
|-
| Обучение
| C,H,O,N,Si (4)
| Cu,Li (2)
| Медии и инструменти
|-
| Бърдо
| Fe,C (5)
| Cr,Ni (4)
| Подобрение с неръждаема стомана
|-
| Матрак
| C,H,O,N (5)
| Si (1)
| Тери, латекс, влакна
|-
| Маса (мебели)
| C,H,O или Fe/Al или Si,O (стъкло) (5)
| Cr,Ni,Ti (1)
| Лесно се изгражда с изобилни елементи
|-
| stol
| C,H,O или Fe/Al (5)
| Ti,Cr (1)
| Изобилните елементи са достатъчни
|}

== 4. Редки/Стратегически елементи (RS) и източници ==

=== 4.1 Основни RS елементи и производящи страни ===
{| class="wikitable sortable"
! RS елемент
! Основна употреба
! Основни държави-източници
|-
| Cr (Хром)
| Неръждаема стомана, корозионна устойчивост
| Южна Африка; Казахстан; Индия; Турция
|-
| Ni (Никел)
| Неръждаема стомана, батерии
| Индонезия; Филипини; Русия; Канада; Австралия; Нова Каледония; Бразилия
|-
| Ti (Титан)
| Леги и здрави сплави
| Китай; Мозамбик; Южна Африка; Австралия; Канада; Индия
|-
| Mn (Манган)
| Сталеделане, батерии
| Южна Африка; Габон; Австралия; Китай; Индия
|-
| B (Бор)
| Модификатор за стъкло и стомана
| Турция; САЩ; Аржентина; Чили; Русия; Китай
|-
| F (Флуорит)
| Флуорохимикалии, батерии
| Китай; Мексико; Монголия; Южна Африка; Виетнам
|-
| Li (Литий)
| Батерии, съхранение на енергия
| Австралия; Чили; Аржентина; Китай; Зимбабве; Канада; Бразилия
|-
| Co (Кобалт)
| Батерии, сплави
| ДР Конго; Индонезия; Русия; Австралия; Филипини; Куба
|-
| W (Волфрам)
| Твърдост, плътност
| Китай; Виетнам; Русия; Руанда; Испания; Австрия; Боливия; Португалия
|-
| Ga (Галий)
| ВЧ и силова електроника
| Китай; Япония; Южна Корея; Русия
|-
| Nd/Pr/Dy
| Постоянни магнити
| Китай; Австралия; САЩ; Мианмар
|}

== 5. Минимални подмножества от RS елементи по тип цивилизация ==
* '''Издръжлива ежедневна цивилизация''' : Cr, Ni, Ti, Mn, B
* '''Електронико-тежка цивилизация''' : Ga, Nd/Pr/Dy, Li, Co, Cu
* '''Екстремна плътност и твърдост''' : W

== 6. Примери за глобална информационна инфраструктура ==

=== 6.1 Най-посещавани уебсайтове (функционална класификация) ===
{| class="wikitable"
! Уебсайт !! Тип
|-
| [https://www.google.com Google] || Търсачка
|-
| [https://www.youtube.com YouTube] || Споделяне на видеа
|-
| [https://www.facebook.com Facebook] || Социални мрежи
|-
| [https://www.chatgpt.com ChatGPT] || Чатбот
|-
| [https://www.whatsapp.com WhatsApp] || Мигновено съобщение
|-
| [https://www.wikipedia.org Wikipedia] || Енциклопедия
|-
| [https://www.yahoo.co.jp Yahoo! Japan] || Новини
|-
| [https://www.amazon.com Amazon] || Пазарно място
|-
| [https://bet.br BET.br] || Хазарт
|-
| [https://www.office.com Microsoft 365] || Софтуер
|-
| [https://www.netflix.com Netflix] || Стрийминг
|-
| [https://www.pornhub.com Pornhub] || Съдържание за възрастни
|-
| [https://www.live.com Live] || Имейл
|-
| [https://www.twitch.tv Twitch] || На живо излъчване
|-
| [https://www.samsung.com Samsung] || Потребителска електроника
|-
| [https://www.weather.com Weather] || Времето
|-
| [https://www.fandom.com Fandom] || Хостинг за уики
|-
| [https://www.stripchat.com Stripchat] || Порно камера
|-
| [https://www.zoom.com Zoom] || Видеоконференции
|-
| [https://www.nytimes.com New York Times] || Медии с новини
|-
| [https://www.espn.com ESPN] || Спорт
|-
| [https://www.roblox.com Roblox] || Игрална платформа
|}

== 7. Тренд за отворен код (примери) ==
* [https://github.com/Stability-AI/StableCascade StableCascade] — модел за генериране на изображения

== 8. Научни открития, които все още са незаменими (след 1700 г.) ==

=== 18-ти век ===
* 1712–1781 — Парна машина
* 1796 — Ваксинация
* Късен 18-ти век — Модерна химия

=== 19-ти век ===
* 1800 — Електрическа батерия
* 1824–1870 — Термодинамика
* 1831–1860 — Електромагнетизъм
* 1846–1847 — Модерна анестезия
* 1850–1880 — Гериална теория и асептика
* 1865 — Генетика
* 1869 — Периодична таблица
* 1895 — Рентгенови лъчи
* 1897 — Електрон

=== 20-ти век ===
* 1900–1930 — Квантова механика
* 1905–1915 — Релативност
* 1909–1913 — Процес на Хaber-Бош
* 1928 — Антибиотици
* 1947 — Транзистор
* 1948 — Теория на информацията
* 1953 — Структура на ДНК
* 1958–1959 — Интегрална схема
* 1960 — Лазер
* 1960-те–1980-те — Интернет
* 1970-те — Модерна криптография
* 1983 — PCR (Полимеразна верижна реакция)

=== 21-ви век ===
* 2012 — CRISPR
* 2010-те — Дълбоко обучение (Deep learning)

== 9. Бележки и ограничения ==
* Рафинирането (пречистването) често е по-критично от самото извличане
* Някои редки елементи могат да бъдат заменени за сметка на производителността
* Рециклирането частично намалява зависимостта от редки елементи, но не я елиминира напълно

Main Page

Tcepo — Fri, 01 May 2026 20:07:23 GMT

← Older revision		Revision as of 22:07, 1 May 2026
(12 intermediate revisions by the same user not shown)
Line 94:		Line 94:
	\| API \|\| [https://api-summary.ailab.infocepo.com:wait-2026-06/docs summary] \|\| Résumé de textes longs		\| API \|\| [https://api-summary.ailab.infocepo.com:wait-2026-06/docs summary] \|\| Résumé de textes longs
	\|-		\|-
	\| API \|\| [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs ~~text2mbeddings~~] \|\| Embeddings pour RAG		\| API \|\| [https://text-embeddings.ailab.infocepo.com:wait-2026-06/docs text2embeddings] \|\| Embeddings pour RAG
	\|-		\|-
	\| API \|\| [https://chromadb.ailab.infocepo.com:wait-2026-06 ChromaDB] \|\| Base vecteur		\| API \|\| [https://chromadb.ailab.infocepo.com:wait-2026-06 ChromaDB] \|\| Base vecteur
Line 121:		Line 121:
	= Nouveautés =		= Nouveautés =

	== Nouveautés 18/04/2026 ==		== Nouveautés 26/04/2026 ==
			* ajout de [https://api-tts-omnivoice.ailab.infocepo.com '''TTS Omnivoice'''] : Qualité TTS augmenté et ajout plus global des langues (600)
			* ajout de [https://api-lightrag.ailab.infocepo.com '''lightRAG'''] : LightRAG est un framework RAG avancé et léger qui combine graphs de connaissances et recherche vectorielle pour une analyse contextuelle profonde et efficace.
			* ajout de [https://api-reranker.ailab.infocepo.com '''API reranker''']
			* ajout de [https://api-embedding.ailab.infocepo.com '''API embedding''']
			* [https://huggingface.co/openai/privacy-filter '''privacy-filter'''] : filtrage données personnelles
	* Un seul fichier [https://github.com/multica-ai/andrej-karpathy-skills '''CLAUDE.md'''] inspiré d’Andrej Karpathy pour transformer Claude en un vrai ingénieur logiciel.		* Un seul fichier [https://github.com/multica-ai/andrej-karpathy-skills '''CLAUDE.md'''] inspiré d’Andrej Karpathy pour transformer Claude en un vrai ingénieur logiciel.
	* Ajout de '''qwen3.6''' : Qwen3.6 delivers substantial upgrades in agentic coding and thinking preservation than previous Qwen models.		* Ajout de '''qwen3.6''' : Qwen3.6 delivers substantial upgrades in agentic coding and thinking preservation than previous Qwen models.
	* [https://github.com/NousResearch/hermes-agent '''Hermes Agent'''] : l’agent qui s’améliore et grandit avec toi.		* [https://github.com/NousResearch/hermes-agent '''Hermes Agent'''] : l’agent qui s’améliore et grandit avec toi.
	* ~~Ajout de~~ [https://github.com/ynotopec/api-audio2txt-gemma4 '''gemma4 STT'''] : API de transcription compatible OpenAI. La qualité est très bonne. Il faut comparer avec Whisper3-turbo. Il est plus gourmant en mémoire.		* [https://github.com/ynotopec/api-audio2txt-gemma4 '''gemma4 STT'''] : API de transcription compatible OpenAI. La qualité est très bonne. Il faut comparer avec Whisper3-turbo. Il est plus gourmant en mémoire. Il ne retourne pas de "timestamp" "sentence".
	* ~~Ajout de~~ [https://github.com/ynotopec/api-audio2txt-qwen3 '''qwen3 STT'''] : API de transcription compatible OpenAI. La qualité est moins bonne en français que Whisper3-turbo. Mais il faudrait tester avec d'autres langues. Il peut théoriquement prendre beaucoup de charge avec le backend actuel vLLM.		* [https://github.com/ynotopec/api-audio2txt-qwen3 '''qwen3 STT'''] : API de transcription compatible OpenAI. La qualité est moins bonne en français que Whisper3-turbo. Mais il faudrait tester avec d'autres langues. Il peut théoriquement prendre beaucoup de charge avec le backend actuel vLLM.
	* '''cohere STT''' : premiers tests non convainquants. Certainement pertinent dans la transcription monolangue, mais non adapté au multilangue. Il faut définir la langue avant transcription. Il ne retourne pas de "timestamp" "sentence".		* '''cohere STT''' : premiers tests non convainquants. Certainement pertinent dans la transcription monolangue, mais non adapté au multilangue. Il faut définir la langue avant transcription. Il ne retourne pas de "timestamp" "sentence".
	* 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.
	* [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.		* [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.
	* DGX Spark : architecture CPU ARM.		* DGX Spark : architecture CPU ARM.
Line 146:		Line 150:
	== Top tasks ==		== Top tasks ==
	* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.		* Ajouter [https://github.com/microsoft/presidio '''Presidio'''] : anonymisation / masquage PII, socle RGPD.
	* Ajouter [https://github.com/sgl-project/sglang '''SGLang'''] : serving LLM haute performance.
	* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.		* Ajouter [https://github.com/llm-d/llm-d '''llm-d'''] : blueprints + charts Kubernetes pour industrialiser les déploiements.
	* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.		* Ajouter [https://github.com/ai-dynamo/dynamo '''Dynamo'''] : orchestration inférence multi-nœuds.
Line 216:		Line 219:

	; '''Découverte de modèles'''		; '''Découverte de modèles'''
	* [https://ollama.com/library LLM Trending]
	* [https://huggingface.co/models Models Trending]		* [https://huggingface.co/models Models Trending]

Line 227:		Line 229:

	== Matériel IA & GPU ==		== Matériel IA & GPU ==
	* ~~[https://www.nvidia.com/en-us/data-center/h100/~~ NVIDIA ~~H100]~~		* NVIDIA GH200
	* DGX Spark		* DGX Spark
	* NVIDIA 5080
	* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]		* [https://www.mouser.fr/ProductDetail/BittWare/RS-GQ-GC1-0109?qs=ST9lo4GX8V2eGrFMeVQmFw%3D%3D GROQ LLM accelerator]

Line 282:		Line 283:
	! Endpoint !! Description / usage principal		! Endpoint !! Description / usage principal
	\|-		\|-
	\| '''ai-~~multilingue~~''' \|\| '''qwen3.6 fp8''' en mode '''nothink''' – ~~multilingue~~		\| '''ai-multilingual''' \|\| '''qwen3.6 fp8''' en mode '''nothink''' – multilingual
	\|-		\|-
	\| '''ai-tools''' \|\| '''qwen3.6 fp8''' – tâches agentiques et outils		\| '''ai-tools''' \|\| '''qwen3.6 fp8''' – tâches agentiques et outils
Line 292:		Line 293:
	\| '''ai-embedding''' \|\| '''bge-m3''' – recherche sémantique		\| '''ai-embedding''' \|\| '''bge-m3''' – recherche sémantique
	\|-		\|-
	\| '''ai-stt''' \|\| '''whisper3-turbo''' – transcription vocale ~~multilingue~~		\| '''ai-stt''' \|\| '''whisper3-turbo''' – transcription vocale multilingual
	\|-		\|-
	\| '''ai-tts''' \|\| '''Kokoro-82M''' – TTS ~~multilingue~~ limité ''(actuel, internal dev)''		\| '''ai-tts''' \|\| '''Kokoro-82M''' – TTS multilingual limité ''(actuel, internal dev)''
	\|-		\|-
	\| '''ai-tts-next''' \|\| '''OmniVoice''' – TTS ~~multilingue~~ en évaluation		\| '''ai-tts-next''' \|\| '''OmniVoice''' – TTS multilingual en évaluation
	\|-		\|-
	\| '''ai-image''' \|\| '''OpenDalle''' – image génération		\| '''ai-image''' \|\| '''OpenDalle''' – image génération
Line 324:		Line 325:

	'''DEV (au choix)'''		'''DEV (au choix)'''
	* '''A.''' <code>LiteLLM → vLLM</code> : tests perf / compatibilité		* '''A.''' <code>LiteLLM → vLLM/SgLang</code> : tests perf / compatibilité
	* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer		* '''B.''' <code>LiteLLM → Ollama</code> : simple, rapide à itérer
	* '''C.''' <code>Ollama</code> direct : POC ultra-léger		* '''C.''' <code>Ollama</code> direct : POC ultra-léger
Line 332:		Line 333:

	'''PROD'''		'''PROD'''
	* '''Standard :''' <code>LiteLLM → vLLM</code>		* '''Standard :''' <code>LiteLLM → vLLM/SgLang</code>
	* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM</code>		* '''Pont DEV→PROD :''' <code>LiteLLM (DEV) → LiteLLM (PROD) → vLLM/SgLang</code>

	'''Notes :'''		'''Notes :'''
	* '''LiteLLM''' = passerelle unique (clés, quotas, logs)		* '''LiteLLM''' = passerelle unique (clés, quotas, logs)
	* '''vLLM''' = performance / stabilité en charge		* '''vLLM/SgLang''' = performance / stabilité en charge
	* '''Ollama''' = simplicité de prototypage		* '''Ollama''' = simplicité de prototypage

AI/ML LAB

Tcepo — Thu, 23 Apr 2026 10:41:10 GMT

Tcepo deleted page AI/ML LAB content was: "{| class="wikitable" |- ! Factor !! Approximate Weight (%) |- | Diverse training language || 30 |- | Model size || 15 |- | Fine-tuning || 15 |- | Regularization || 10 |- | Optimizer and learning rate || 10 |- | Context size || 10 |- | Libraries || 5 |- | Model architecture || 3 |- | Model initialization || 2 |} == Implement Vicuna (ChatGPT) == * Download LLAMA model (LLM model) :...", and the only contributor was "Tcepo" (talk)

Main Page

Tcepo — Mon, 20 Apr 2026 15:27:35 GMT

← Older revision		Revision as of 17:27, 20 April 2026
(15 intermediate revisions by the same user not shown)
Line 121:		Line 121:
	= Nouveautés =		= Nouveautés =

	== Nouveautés 10/04/2026 ==		== Nouveautés 18/04/2026 ==
			* Un seul fichier [https://github.com/multica-ai/andrej-karpathy-skills '''CLAUDE.md'''] inspiré d’Andrej Karpathy pour transformer Claude en un vrai ingénieur logiciel.
			* Ajout de '''qwen3.6''' : Qwen3.6 delivers substantial upgrades in agentic coding and thinking preservation than previous Qwen models.
			* [https://github.com/NousResearch/hermes-agent '''Hermes Agent'''] : l’agent qui s’améliore et grandit avec toi.
	* Ajout de [https://github.com/ynotopec/api-audio2txt-gemma4 '''gemma4 STT'''] : API de transcription compatible OpenAI. La qualité est très bonne. Il faut comparer avec Whisper3-turbo. Il est plus gourmant en mémoire.		* Ajout de [https://github.com/ynotopec/api-audio2txt-gemma4 '''gemma4 STT'''] : API de transcription compatible OpenAI. La qualité est très bonne. Il faut comparer avec Whisper3-turbo. Il est plus gourmant en mémoire.
	* Ajout de [https://github.com/ynotopec/api-audio2txt-qwen3 '''qwen3 STT'''] : API de transcription compatible OpenAI. La qualité est moins bonne en français que Whisper3-turbo. Mais il faudrait tester avec d'autres langues. Il peut théoriquement prendre beaucoup de charge avec le backend actuel vLLM.		* Ajout de [https://github.com/ynotopec/api-audio2txt-qwen3 '''qwen3 STT'''] : API de transcription compatible OpenAI. La qualité est moins bonne en français que Whisper3-turbo. Mais il faudrait tester avec d'autres langues. Il peut théoriquement prendre beaucoup de charge avec le backend actuel vLLM.
	* '''cohere STT''' : premiers tests non convainquants. Certainement pertinent dans la transcription monolangue, mais non adapté au multilangue. Il faut définir la langue avant transcription. Il ne retourne pas de "timestamp" "sentence".		* '''cohere STT''' : premiers tests non convainquants. Certainement pertinent dans la transcription monolangue, mais non adapté au multilangue. Il faut définir la langue avant transcription. Il ne retourne pas de "timestamp" "sentence".
	* 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.		* 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.
	* ~~Ajout de '''nemotron-cascade-2''' : modèle open 30B MoE NVIDIA orienté raisonnement et tâches agentiques.~~		* [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.
	* Ajout de '''nemotron-3-super''' : NVIDIA Nemotron 3 Super is a 120B open MoE model activating just 12B parameters to deliver maximum compute efficiency and accuracy for complex multi-agent applications.
	* Ajout de [https://github.com/sst/opencode '''opencode'''] : CLI coder à comparer avec Aider / OpenHands.
	* DGX Spark : architecture CPU ARM.		* DGX Spark : architecture CPU ARM.
	* Ajout de '''qwen3.5''' : famille de modèles open source multimodaux.
	* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.		* Ajout de [https://github.com/ynotopec/api-convert2md '''api-convert2md'''] : extraction de tableaux pour RAG compatible Open WebUI.
	* Mise à jour des paramètres '''RAG optimisation'''.		* Mise à jour des paramètres '''RAG optimisation'''.
Line 138:		Line 138:
	* Ajout de [https://langextract.ailab.infocepo.com '''langextract'''] : démo extraction d’entités.		* Ajout de [https://langextract.ailab.infocepo.com '''langextract'''] : démo extraction d’entités.
	* Ajout de [https://sam-audio.c1.ailab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.		* Ajout de [https://sam-audio.c1.ailab.infocepo.com:wait-2026-06 '''sam-audio'''] : séparation audio sémantique.
	* Ajout de '''glm-4.7-flash''' : modèle 30B léger orienté performance / efficacité.
	* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.		* Ajout de '''API Realtime''' : WebRTC / WebSocket bidirectionnel basse latence.

Line 219:		Line 218:
	* [https://ollama.com/library LLM Trending]		* [https://ollama.com/library LLM Trending]
	* [https://huggingface.co/models Models Trending]		* [https://huggingface.co/models Models Trending]
	* [https://huggingface.co/models?pipeline_tag=image-text-to-text&sort=trending Img2txt Trending]
	* [https://arena.ai/leaderboard/text-to-image Txt2img Evaluation]

	; '''Évaluation & benchmarks'''		; '''Évaluation & benchmarks'''
	* [https://~~lmarena~~.ai/leaderboard ~~ChatBot~~ Evaluation]		* [https://arena.ai/leaderboard/code Agentic Evaluation]
	* [https://huggingface.co/spaces/mteb/leaderboard Embedding Leaderboard]
	* [https://ann-benchmarks.com Vectors DB Ranking]
	* [https://top500.org/lists/green500 HPC Efficiency]

	; '''Outils de développement & fine-tuning'''		; '''Outils de développement & fine-tuning'''
	* [https://github.com/~~search~~?q=~~stars%3A%3E22222+forks%3A%3E222+AI+OR+LLM&type=repositories&s=updated&o=desc~~ Project Trending]		* [https://github.com/trending?since=weekly Project Trending]
	* [https://grok.com News search]		* [https://grok.com News search]

Line 281:		Line 275:
	== Modèles ouverts & endpoints internes ==		== Modèles ouverts & endpoints internes ==

	''Dernière mise à jour : 2026-04-15''		''Dernière mise à jour : 2026-04-20''

	Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.		Les modèles ci-dessous correspondent à des '''endpoints logiques''' exposés derrière une passerelle.
Line 288:		Line 282:
	! Endpoint !! Description / usage principal		! Endpoint !! Description / usage principal
	\|-		\|-
	\| '''ai-~~chat~~''' \|\| '''qwen3.~~5:122b~~ fp8''' en mode '''nothink''' – ~~chat~~ multilingue		\| '''ai-multilingue''' \|\| '''qwen3.6 fp8''' en mode '''nothink''' – multilingue
	\|-		\|-
	\| '''ai-tools''' \|\| '''qwen3.~~5:122b~~ fp8''' – tâches agentiques et outils		\| '''ai-tools''' \|\| '''qwen3.6 fp8''' – tâches agentiques et outils
	\|-		\|-
	\| '''ai-~~ocr~~''' \|\| '''qwen3.~~5:122b~~ fp8~~''' en mode '''nothink~~''' – ~~OCR~~		\| '''ai-thinking''' \|\| '''qwen3.6 fp8''' – thinking
	\|-		\|-
	\| '''ai-~~summary~~''' \|\| '''qwen3.~~5:122b~~ fp8''' en mode '''nothink''' – ~~résumé~~		\| '''ai-vision''' \|\| '''qwen3.6 fp8''' en mode '''nothink''' – vision/OCR
	\|-
	~~\| '''ai-rag''' \|\| '''qwen3.5:122b fp8''' en mode '''nothink''' – RAG multilingue~~
	\|-
	~~\| '''ai-translate''' \|\| '''qwen3.5:122b fp8''' en mode '''nothink''' – traduction multilingue~~
	\|-		\|-
	\| '''ai-embedding''' \|\| '''bge-m3''' – recherche sémantique		\| '''ai-embedding''' \|\| '''bge-m3''' – recherche sémantique
Line 307:		Line 297:
	\|-		\|-
	\| '''ai-tts-next''' \|\| '''OmniVoice''' – TTS multilingue en évaluation		\| '''ai-tts-next''' \|\| '''OmniVoice''' – TTS multilingue en évaluation
			\|-
			\| '''ai-image''' \|\| '''OpenDalle''' – image génération
	\|}		\|}

Line 353:		Line 345:

	* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.		* Utilise l’API LLM avec un endpoint adapté à l’OCR / VLM.
	* Modèle recommandé : <code>ai-~~ocr~~</code>		* Modèle recommandé : <code>ai-vision</code>

	== Exemple bash ==		== Exemple bash ==
Line 363:		Line 355:
	jq -n --rawfile img img.b64 \		jq -n --rawfile img img.b64 \
	'{		'{
	model: "ai-~~ocr~~",		model: "ai-vision",
	messages: [		messages: [
	{		{
Line 392:		Line 384:

	API_KEY = os.getenv("OPENAI_API_KEY")		API_KEY = os.getenv("OPENAI_API_KEY")
	MODEL = "ai-~~ocr~~"		MODEL = "ai-vision"
	IMG_PATH = "/path/to/image.png"		IMG_PATH = "/path/to/image.png"
	API_URL = "https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions"		API_URL = "https://api.ailab.infocepo.com:wait-2026-06/v1/chat/completions"
Line 686:		Line 678:
	* <code>chunk_size=1200</code>		* <code>chunk_size=1200</code>
	* <code>chunk_overlap=100</code>		* <code>chunk_overlap=100</code>
	* LLM : <code>qwen3</code>		* LLM : <code>qwen3.6</code>
	* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.		* Pour les PDF mixtes : '''PDF → image → OCR / VLM''' peut améliorer les résultats.