#1010th of 12 in Vector databases
leann
Vector index that recomputes embeddings on demand to cut storage
Live demo ↗
(opens in a new tab)Documentation ↗
(opens in a new tab)Repository on GitHub ↗
(opens in a new tab)
- Stars
- 13k
- License
- MIT
- Last commit
- Oct 2026
- Last release
- Sep 2026
- Language
- Python
Missed the latest GitHub check; these facts may be older.
Overview
LEANN is a Python vector database that stores a pruned graph and recomputes embeddings at query time instead of storing them. The README claims 97% less storage, for example 60M chunks in 6GB versus 201GB. It ships HNSW and DiskANN backends, a CLI, RAG examples for files, mail and chat history, and an MCP server for Claude Code.
Who it is for: Developers wanting a low-storage local vector index for personal data RAG
Strengths
- Graph-based selective recomputation avoids storing embeddings; README claims 97% less storage
- HNSW and DiskANN backends, selectable with --backend-name
- Works with Ollama, LM Studio, vLLM, llama.cpp and any OpenAI-compatible endpoint
- Includes ready RAG loaders for PDFs, Apple Mail, WeChat, iMessage, ChatGPT and Claude history
Weaknesses
- Recomputing embeddings at query time adds latency compared with stored vectors
- DiskANN needs a native toolchain (Boost, protobuf, MKL) when built from source
- GPU acceleration is only a survey item for a future version
- Linux uv hardlink installs can fail with a nltk pathsec error without a workaround
What it needs
- GPU optional
- Needs uv, Ollama or OpenAI-compatible API (optional, for generation)
- Models: HuggingFace, Ollama, OpenAI-compatible APIs, Anthropic, sentence-transformers
Also in Vector databases
See all 12| Rank | Project | Score |
|---|---|---|
| 1 | MilvusDistributed vector database with dense, sparse and hybrid search at scale | 82 out of 100 |
| 2 | MeilisearchRust search engine API with full-text, vector and hybrid search | 75 out of 100 |
| 3 | QdrantRust vector database with payload filtering, REST and gRPC | 69 out of 100 |
| #4 | faissC++ library for similarity search and clustering of dense vectors | 66 out of 100 |
| #5 | ChromaEmbedding database with a four-function API for Python and JavaScript | 62 out of 100 |