📦LEANN
StarTrail-org/LEANN · homepage ↗
[MLsys2026]: RAG on Everything with LEANN. Enjoy 97% storage savings while running a fast, accurate, and 100% private RAG application on your personal device.
⭐ Very popular: 13k stars, gaining about 17 a week
View on GitHub ↗repo profile
momentum
durability
bus factor = how many people it takes to cover more than half the commits (6 months). 1 is a solo project; higher means the work is spread across a team. top-author share is the single busiest author's slice of those commits.
since we covered it
why it's a big deal
- Lets people run semantic search and RAG over personal data on their own device, so nothing leaves the machine.
- Cuts index storage by 97 percent, for example 6GB instead of 201GB for 60M Wikipedia documents, making large local corpora practical on a laptop.
- Indexes personal sources including Apple Mail, iMessage, Chrome history, WeChat, and ChatGPT or Claude exports, plus live Slack and Twitter data over MCP.
under the hood
- Written in Python; instead of storing every embedding it keeps a pruned graph and recomputes embeddings on demand only for nodes on the search path.
- Offers two backends, HNSW by default for maximum storage savings and DiskANN for faster PQ-based traversal with reranking, using CSR format for graph storage.
- Generation is pluggable across local backends such as Ollama, LM Studio, and vLLM or cloud providers, while the index and retrieval stay local, and it exposes an MCP server for Claude Code code search.
Radar summary, generated from the project's public sources
star history
- PR#16 9k 2025-09-03
- now 13k + 3k since first covered
curve is sampled from GitHub's star history, plus our own daily readings since we covered it; the dashed stretch is before we first covered it, the solid line since. figures at coverage are the numbers we printed then (approx.), current count is live.
understory
Better known than its recent output, coasting a little on attention.
- output, commits & releases
- clout, star velocity
output = commits & releases; clout = star velocity, both 0 to 100 monthly indices; the gap where output runs above clout is the understory. The understory →
covered in
-
Lightweight Vector Index for Personal RAG
similar projects
compare these →- 📑 PageIndex
2.8× the stars
📑 PageIndex: Document Index for Vectorless, Reasoning-based RAG
35k ACTIVE - 🗜️ headroom
5.2× the stars
Compress tool outputs, logs, files, and RAG chunks before they reach the LLM. 20% fewer tokens for coding agents, 60-95% fewer tokens for JSON, same answers. Library, proxy, MCP server.
66k ACTIVE - 🧠 Mem0
5× the stars
Universal memory layer for AI Agents
63k ACTIVE
comments
Sign in with GitHub to add your blip on LEANN.