Skip to content

lyonzin/knowledge-rag

MCP Developer Tools

Local RAG system for Claude Code with hybrid search (BM25 + semantic), cross-encoder reranking, markdown-aware chunking, query expansion, and 12 MCP tools. Runs entirely offline with zero external servers.

Python Latest v3.9.0 · 24d ago Security brief →

Features

  • Local, zero‑cloud retrieval-augmented generation (RAG) for documents
  • Hybrid search combining BM25, semantic vectors, and cross‑encoder reranking
  • Supports 20+ file formats (PDF, markdown, code, notebooks, etc.)
  • Runs entirely on the machine via ONNX – no Docker or Ollama required
  • Optional NVIDIA GPU acceleration for faster embedding

Recent releases

View all 22 releases →
No immediate action
v3.9.0 Maintenance

Routine maintenance and dependency updates.

Upgrade now
v3.8.1 Bug fix

Embedding error handling

No immediate action
v3.8.0 New feature

Lazy FastEmbed loading

No immediate action
v3.7.0 Bug fix

Resilience fixes

Monitor
v3.6.2 Security relevant
Dependencies

SLSA provenance attestation

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

About

Stars
88
Forks
17
Languages
Python Shell PowerShell
Downloads/week
102 ↑2%
NPM Maintainers
1 Single npm maintainer
Contributors
4

Install & Platforms

Install via
pip
Platforms
linux macos windows

Beta — feedback welcome: [email protected]