RAG・知識ベース
autoflow
Graph RAG conversational knowledge base built on TiDB Serverless vector storage
★ 3.0k スターApache-2.0
ガイドを見る →LEANNは、StarTrail-orgが公開しているオープンソースのRAG・知識ベースプロジェクトです（GitHub: StarTrail-org/LEANN）。GitHubスターは13,012、フォークは1,179で、MITライセンスの下で公開されています。GitHubでの最終更新は2026年9月29日です。
The smallest vector index in the world: private RAG over everything on your laptop with 97% less storage.
Pick a source: documents, email, browser history, chat exports, a codebase, or live data from platforms such as Slack via MCP servers.
Install uv if needed, clone the repo for the examples, and install the leann package from PyPI into a fresh virtual environment.
Use the builder API or the leann build CLI to index your data - the graph index stays dramatically smaller than a conventional embedding store.
Query with LeannSearcher, chat with LeannChat using a local or API model, or plug LEANN into Claude Code as its semantic-search MCP service.
Graph-based selective recomputation with high-degree-preserving pruning avoids storing every embedding, while the project reports no accuracy loss versus heavyweight solutions.
Ready-made support for files (PDF, TXT, Markdown), Apple Mail, browser history, WeChat and iMessage, ChatGPT and Claude conversations, codebases, and live data via MCP servers.
Your data never leaves your laptop: no cloud, no OpenAI requirement - local Hugging Face or Ollama models cover embedding and generation.
Build, search and chat in a few lines with LeannBuilder, LeannSearcher and LeannChat, plus a full CLI for scripting.
Acts as a drop-in semantic search MCP service for Claude Code; the project's ContextBench comparison reported 2.1x the initial relevant-code recall of BM25 on 30 SWE-Bench Pro tasks.
The tiny index makes it cheap to copy your entire personal knowledge base between devices or share it with others.
Clone LEANN (the README clones the original yichuan-w repository into a folder named leann) to access all examples and apps.
git clone https://github.com/StarTrail-org/LEANN.git
cd LEANNCreate a uv virtual environment and install the leann package into it.
uv venv
uv pip install leannFollow the README quick start: build an index with LeannBuilder, then query it with LeannSearcher or chat with LeannChat.
python demo.py # adapt the README quick-start snippet - see README for the exact codeLEANN is maintained by StarTrail-org and contributors; the underlying research paper (arXiv 2506.08276) received a best paper award at MLSys 2026, per the project.
Yes. It is MIT-licensed and runs on your own machine with zero cloud costs.
No: embedding and generation can run fully locally (Hugging Face models, Ollama, LM Studio and others). API providers such as OpenAI are optional and use your own key.
Those store full embedding vectors for every chunk; LEANN recomputes embeddings on demand from a pruned graph, trading some compute for roughly 97% less storage at personal scale.
Install uv, clone the repo, uv pip install leann, and run the quick-start snippet to build and search your first index.
Graph RAG conversational knowledge base built on TiDB Serverless vector storage
RAGApp: deploy a RAG app over your documents with Docker — configure it like a custom GPT
Microsoft's graph-based RAG: index a corpus into a knowledge graph, then query it with an LLM
A blazing-fast vector index in Rust, built on TurboQuant - with Python bindings
DeepSearcher — open-source deep research over your private data, with reasoning models and Milvus retrieval.
Free rag & knowledge tool that enterprise-ready vector database toolkit for building s.
LibreAIカタログに掲載されたRAG・知識ベースプロジェクトの中で最もスターの多いもの — LEANNが合わなかった場合に検討したい、現実的な候補リストです。
Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or…
RAGFlow is a leading open-source Retrieval-Augmented Generation (RAG) engine that fuses cutting-edge RAG with Agent capabilities to create a superior context…
Lightweight OCR toolkit turning PDFs and images into structured, LLM-ready data in 100+ languages
MinerU — turns PDFs and Office documents into LLM-ready Markdown and JSON for agents and RAG pipelines, Apache-2.0 plus terms
Stop renting your intelligence. Own it with AnythingLLM. Everything you need for a powerful local-first agent experience
カタログの中でLEANNとスター数が近いRAG・知識ベースプロジェクト — 前後の棚もそのまま見て回れます。
Local knowledge-base Q&A over PDFs, Office files, images and web pages - offline and CPU-friendly.
All-in-one AI framework for semantic search, LLM orchestration and language model workflows.
Unified framework for enterprise RAG pipelines with small, specialized models - local, private and secure.
Incremental engine that keeps RAG indexes and agent context fresh by processing only the delta.
Lightweight, lightning-fast in-process vector database from Alibaba that embeds in your app.
Polyglot document intelligence in Rust — extract text, tables and structured data from 106 formats.