Skip to content

Add knowledge-rag — Local RAG system for Claude Code #3530

Description

@lyonzin

Project: knowledge-rag

GitHub: https://github.com/lyonzin/knowledge-rag
PyPI: https://pypi.org/project/knowledge-rag/
Author: Lyon.
License: MIT

Description

Local RAG (Retrieval-Augmented Generation) system that integrates with Claude Code via MCP. Hybrid search combining semantic embeddings + BM25 keyword search with cross-encoder reranking. Zero external servers — everything runs in-process via ONNX Runtime.

Key Features

  • Hybrid Search: Semantic + BM25 with Reciprocal Rank Fusion (RRF)
  • Cross-Encoder Reranking: Xenova/ms-marco-MiniLM-L-6-v2 for precision boost
  • Markdown-Aware Chunking: Splits .md files by section headers
  • Query Expansion: 54 security/technical term synonyms
  • 12 MCP Tools: search, get, reindex, list, stats, add, update, remove, add_from_url, search_similar, evaluate_retrieval
  • Zero Dependencies: No Ollama, no Docker, no API keys. Pure ONNX in-process via FastEmbed
  • Install: pip install knowledge-rag

Suggested Category

Knowledge & Memory

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions