C’est le groupe le plus large : interfaces de chat, SDK et frameworks, outils de prompts et d’évaluation, et applications reposant sur des grands modèles de langage. Beaucoup fonctionnent avec les modèles de plusieurs fournisseurs, et certains peuvent faire tourner des modèles sur votre propre matériel.
Pour tout ce que vous comptez mettre en production, vérifiez d’abord trois faits : la licence, car certains projets populaires restreignent l’usage commercial ; la prise en charge du fournisseur de modèles que vous utilisez déjà ; et la maintenance du projet. L’encadré « En bref » de chaque page projet répond aux questions de licence et de maintenance à partir des données GitHub.
Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.
Fully Open Framework for Democratized Multimodal Training
WebAssembly binding for llama.cpp - Enabling on-browser LLM inference
🥤 RAGLite is a Python toolkit for Retrieval-Augmented Generation (RAG) with DuckDB or PostgreSQL
[JMLR 2026] "UQLM: A Python Package for Uncertainty Quantification in Large Language Models"
Production-ready AI chat. Start here and make it your own. Formerly Sparka AI
Neural Network Compression Framework for enhanced OpenVINO™ inference
AI辅助长篇小说创作，卡片式创作，支持基于 JSON Schema的结构化 AI 生成与上下文引用，可扩展性强。
Instant SECURE Full Stack Apps and AI Agents
Evaluate and improve models and agents using environments
AI-powered code assistant for Vim. OpenAI and ChatGPT plugin for Vim and Neovim.
Agentic SOC Platform: A powerful, flexible, open-source, and agent-centric automated security operations platform (AI SOC)
A practical lab for building, testing, and evaluating apps with Apple's Foundation Models framework.
Train speculative decoding models effortlessly and port them smoothly to SGLang serving.
Simple go utility to download HuggingFace Models and Datasets
AI gateway / AI control plane / AI proxy written in Go. Unified OpenAI-compatible and Anthropic-compatible API for OpenAI, Anthropic, Gemini, Groq, xAI, Ollama, vLLM and more. A LiteLLM alternative with observability, guardrails, streaming, cost tracking, intelligent routing, sticky sessions, failover, real-time logs and usage tracking. Prod ready.
A desktop MCP client designed as a tool unitary utility integration, accelerating AI adoption through the Model Context Protocol (MCP) and enabling cross-vendor LLM API orchestration.
[ICLR 2026] LightMem: Lightweight and Efficient Memory-Augmented Generation
An open-source, PyTorch-like runtime for dynamic multi-agent and multi-session workflows.
Établi de recherche en IA de bureau open source et local pour le calcul scientifique avec Python/R, les outils bioinformatiques MCP, les environnements d'exécution SSH/WSL/GPU et les modèles OpenAI/Anthropic.
Open-source, self-hosted enterprise & site search server built on OpenSearch. Crawls web / file / DB / cloud sources, 20+ languages, REST API, and AI/RAG & semantic search. Apache-2.0.
🚀 MassGen is an open-source multi-agent scaling system that runs in your terminal, autonomously orchestrating frontier models and agents to collaborate, reason, and produce high-quality results. | Join us on Discord: discord.massgen.ai
Unify Efficient Fine-tuning of RAG Retrieval, including Embedding, ColBERT, ReRanker.
Your AI forgets. This remembers. Spec-driven coding harness for vibecoders, product owners, CEOs and real builders — self-improving context memory, 15 agents, 33 skills working with /goal, agent-team, & workflow on autopilot loops with 0 need for human gate. Kills context rot, ships features, not spaghetti. Claude Code & Codex. Any stack