GH Repository · StarTrail-org
LEANN
[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.
- stars
- 12,942
- 30-day movement
- starts with the next reading
- Related entries
- 60
- Connections
- 3
agent-frameworkPythongpt-ossvectorspython/uvvector-searchfaissvector-databaseragoffline-firstllama-indexailocalstorageollamaretrieval-augmented-generationlangchainprivacyllmpython
LEANN is a Python-based RAG framework from the Star Trail org that aims for large storage savings (claimed 97%) while running retrieval-augmented generation fully locally and privately. It integrates with vector search tooling such as FAISS and LLM stacks including Ollama, LangChain, and Llama Index, per its topics.
You want private, offline RAG on your own device without large vector-index storage overhead.
Use it to
- Build a fully local, private RAG application
- Run retrieval-augmented generation offline on a personal device
- Index personal files with a low-storage vector search pipeline
- Experiment with compact vector search using FAISS or Llama Index
For Developers building local, privacy-focused RAG applications
- Role
- agent-framework
- Language
- Python
- Licence
- MIT
- Forks
- 1,169
- Open issues
- 37
- Last push
- 2026-09-05
- Latest release
- v0.1.1 · 2025-07-24
topicsragvector-searchprivacylocal-storagepythonllm