# StarTrail-org/LEANN

[MLsys2026]: RAG on Everything with LEANN. Enjoy 97% storage savings while running a fast, accurate, and 100% private RAG application on your personal device.

Repository: https://github.com/StarTrail-org/LEANN
Canonical: https://ross.abutalabs.com/products/leann
Homepage: https://arxiv.org/abs/2506.08276
Language: Python
License: MIT
License Family: permissive
Topics: ai, faiss, langchain, llama-index, llm, localstorage, offline-first, ollama, privacy, python, rag, retrieval-augmented-generation, vector-database, vector-search, vectors, gpt-oss
Last push: 2026-08-25T18:44:24+00:00

## Health v2 (maintenance only)
Score: 77/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 99, release rhythm 74, longevity 32
- inputs: {"age_days": 450, "days_push": 8, "days_rel": 178, "gap_med": 0, "n_releases_24m": 28}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 12836, forks 1152 (observed 2026-08-28T04:10:59.895370+00:00)

## What it is
LEANN is a low-storage vector index that enables retrieval-augmented generation (RAG) on personal devices by recomputing embeddings on the fly instead of storing them, achieving up to 97% (50x) storage savings with no accuracy loss. It ships as a Python library and CLI with integrations for Ollama, LangChain, LlamaIndex, and MCP, letting users semantically search files, emails, browser history, and documents fully offline and privately.

## Use cases
- run a private RAG assistant on my laptop over my own documents
- semantic search across PDFs, emails, and files without uploading data to the cloud
- build a vector index that doesn't blow up my disk with embeddings
- chat with my local files using Ollama models offline
- index millions of documents on a personal device with minimal storage
- add retrieval-augmented generation to a LangChain or LlamaIndex app
- query my codebase or browser history locally with an MCP-connected agent

## When to choose
- you need 100% private, offline RAG on a laptop or personal device
- storage overhead of traditional vector indices (FAISS, etc.) is a blocker
- you want local semantic search over heterogeneous personal data (files, email, browser history)
- you use Ollama, LangChain, LlamaIndex, or MCP and want drop-in RAG

## When to avoid
- you need a distributed, multi-node vector database for server-scale deployments
- your workload demands the lowest possible query latency and can afford large storage
- you need GPU-accelerated or heavily optimized production serving today
- you require a managed service with built-in replication and sharding

## Facets
- artifact type: library
- maturity: active
- function: vector-database, rag, search-engine, llm-inference, mcp
- domain: large-language-models, databases, privacy, self-hosted
- platform: windows, python, cli, self-hosted
- tags: local-rag, vector-index, storage-efficient, semantic-search, ollama, personal-ai, offline-first, embeddings, retrieval-augmented-generation, search, macos, linux

## Member repositories
- StarTrail-org/LEANN (main) score 77

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:59.895370+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:13:45.408770+00:00, confidence not recorded.
  - readme: https://github.com/StarTrail-org/LEANN (fetched 2026-08-28T04:10:59.895370+00:00, sha bd83287695ae)
  - homepage: https://arxiv.org/abs/2506.08276 (fetched 2026-08-29T08:09:51.381924+00:00, sha 937419d25e95)
  - site_page: https://info.arxiv.org/about/donate.html (fetched 2026-08-29T08:09:51.391242+00:00, sha cca9c3a11c56)
  - site_page: https://info.arxiv.org/about/ourmembers.html (fetched 2026-08-29T08:09:51.395931+00:00, sha 47cbc55ff1de)
  - site_page: https://info.arxiv.org/about (fetched 2026-08-29T08:09:51.398893+00:00, sha a1f16f915a9a)
  - site_page: https://info.arxiv.org/labs/index.html (fetched 2026-08-29T08:09:51.393399+00:00, sha b14a8d05a0ec)
- Data as of 2026-08-30T08:39:29.467469+00:00.