qmd

Search local Markdown notes and project docs offline using hybrid retrieval that combines BM25 keyword search and vector search to deliver relevant results without cloud dependency. The system indexes files entirely on your device and returns matching documents directly through a CLI interface.

Hybrid retrieval for knowledge bases works by combining BM25 keyword search, vector search, and LLM reranking to identify and rank relevant documents. This multi-stage approach ensures accurate results by matching exact keywords first, then refining semantic relevance using vector embeddings and LLM-based reranking.

Yes, you can use MCP integration to search personal knowledge bases. The system provides access through Rawclaw MCP, allowing you to retrieve relevant information from local text collections and meeting transcripts directly within your MCP-connected workflow.

Yes, local knowledge search works without an internet connection by running entirely offline on your device. It indexes and retrieves information from scattered personal notes and project docs across local Markdown and text collections with no cloud dependency required.

The best way to find specific information across scattered meeting transcripts is using a CLI interface with hybrid ranking. By applying BM25 and vector search followed by LLM reranking, you can quickly locate accurate, relevant passages stored locally in your text collections.