Vibe Coding DiscoverLEANN
View on GitHub[MLsys2026 Best Paper]: https://arxiv.org/abs/2506.08276. RAG on Everything with LEANN. Enjoy 97% storage savings while running a fast, accurate, and 100% private RAG application on your personal device.
LEANN is a local, privacy-first vector database for RAG that uses graph-based selective recomputation to cut embedding storage by ~97% versus traditional vector DBs, with no accuracy loss. It indexes documents, emails, browser history, chat logs, and code, and ships an MCP server for agent use.
Use Cases
personal document and file-system RAGemail search over Apple Mailbrowser history semantic searchWeChat and iMessage chat history searchChatGPT/Claude conversation archive searchsemantic code search for coding agentsMCP server for live data retrieval in Claude Codeprivate offline RAG on a laptopagent memory storage and retrievalSlack and Twitter bookmark searchmillion-scale local knowledge base indexingdrop-in semantic search layer for coding agents
Built With
- Language
- Python
- Frameworks
- llama-index · langchain · ollama · sglang · sentence-transformers · transformers · faiss · hnsw · diskann · pytorch · mlx · tree-sitter · typer · mcp
Tags
rag · vector-search · vector-database · hnsw · graph-index · embeddings · semantic-search · on-device · privacy · offline-first · local-first · storage-efficient · personal-ai · mcp · faiss · diskann