qmd

Local search for markdown notes is achieved by indexing documents into collections and generating embeddings. The system applies a hybrid search engine combining BM25, vector search, and LLM reranking to retrieve relevant text directly on your device.

Hybrid search for personal documents combines keyword matching with semantic understanding. It uses BM25 for exact term frequency, vector search for semantic similarity, and LLM reranking to refine and order the most accurate results from your knowledge base.

Yes, local document search supports Microcontroller Protocol (MCP) integration. This allows you to seamlessly query your indexed markdown notes and text files through external systems alongside the standard command-line interface.

To set up local knowledge base search for meeting transcripts, add your transcript files to document collections and generate embeddings. You can then execute queries using the command-line interface to find specific information across your text files.

On-device document search operates locally without requiring an external internet connection. By generating local embeddings and utilizing BM25 with vector search, your personal knowledge base queries and document retrieval remain entirely private.

The best way to retrieve specific information from large text file collections is using a hybrid search engine. It indexes documents locally and applies BM25, vector search, and LLM reranking to ensure comprehensive and accurate query results.