LEANN

Active

Description

RAG on Everything with LEANN. Enjoy 97% storage savings while running a fast, accurate, and 100% private RAG application on your personal device. Published at MLsys 2026.

Key Features

97% storage reduction vs traditional vector databases with no accuracy loss
Graph-based selective recomputation — embeddings computed on demand, not stored
100% local and private — data never leaves your laptop
Native MCP integration for Claude Code and other AI assistants
Handles 60M+ documents in 6GB instead of 201GB on consumer hardware
Cross-platform: macOS, Linux, Windows (WSL), with Python 3.9-3.13

Use Cases

💡 Personal RAG over emails, browser history, chat logs, and local files

💡 Semantic search for Claude Code replacing basic grep-style keyword search

💡 Indexing and searching WeChat, iMessage, Slack, and Twitter conversations

💡 Privacy-sensitive document retrieval without cloud dependency

💡 Building personal AI assistants with full local knowledge bases

Strengths & Limitations

✅ Strengths

• Actively maintained, recent updates
• High community interest (12.9k stars)
• Permissive open-source license (MIT)
• Responsive to issues, low backlog

Quick Start

curl -LsSf https://astral.sh/uv/install.sh | sh (install uv). git clone https://github.com/yichuan-w/LEANN.git && cd leann. uv venv && source .venv/bin/activate && uv pip install leann. Then follow the example notebooks in the repo.

Visit GitHub

Related Projects

DeepLake

9.2k · C++

Stale B

AI Data Runtime for Agents. Provides serverless Postgres with a multimodal datalake, enabling scalable retrieval and training. Unifies vector storage, dataset management, and streaming data loading for AI agent workflows.

vector-databasedata-processingpython +3

· AI data runtime for agents providing serverless Postgres with multimodal datalake architecture
· Unified vector storage, dataset management, and streaming data loading for LLM apps and deep learning training
· Native multi-cloud support with a single API to operate S3, GCP, Azure, and local storage

View full details → View on GitHub

ColiVara

1.5k · Python

Normal B

ColiVara is a suite of services for storing, searching, and retrieving documents based on visual embeddings. It uses vision models instead of chunking and text-processing, achieving state-of-the-art retrieval on both text and visual documents without OCR.

ragvector-databaseembedding +2

· Vision-based retrieval — uses ColPali vision language models for document embeddings instead of text chunking or OCR
· 100+ format support — handles PDF, DOCX, PPTX, and 100+ other file formats with automatic image conversion
· Metadata filtering — filter searches by collection and document metadata fields (author, year, tags, etc.)

View on GitHub

RAGatouille

4.0k · Python

Stale C

Easily use and train state of the art late-interaction retrieval methods (ColBERT) in any RAG pipeline. Designed for modularity and ease-of-use, backed by research.

ragpythonembedding +1

· ColBERT late-interaction retrieval: wraps advanced ColBERT models into simple APIs, enabling late-interaction retrieval methods in RAG pipelines without deep IR research knowledge
· End-to-end training and fine-tuning: built-in RAGTrainer and TrainingDataProcessor with automatic deduplication, positive/negative pairing and hard negative mining, supporting training from scratch or fine-tuning pretrained models
· Modular composable architecture: DataProcessor, NegativeMiner and other components can be used independently, with support for custom NegativeMiner integration into training pipelines

View on GitHub

Quivr

39.5k · Python

Active A+

Opinionated RAG framework for integrating GenAI into your apps. Works with any LLM, any vectorstore, any files — so you can focus on your product instead of building RAG pipelines.

ragpythonvector-database +3

· Opinionated RAG framework — Ready-to-use RAG pipeline, no need to build from scratch
· Any file support — PDF, TXT, Markdown and more with custom parser support
· Multi-LLM compatible — OpenAI, Anthropic, Mistral, Gemma and Ollama local models