qmd

To search markdown notes locally without cloud dependencies, you can use an on-device search engine that indexes personal documentation and meeting transcripts directly on your machine.

Hybrid search with LLM reranking combines traditional keyword matching and semantic vector understanding, then uses a large language model to reorder results for higher accuracy in a local knowledge base.

Yes, you can run semantic search on meeting transcripts using command-line tools by executing a search command like 'qmd search' after indexing your local knowledge base.

Yes, you need Node.js and SQLite to set up this local markdown search engine, as it requires the Node.js runtime, SQLite database, and the @tobilu/qmd package for on-device indexing.

The best way to integrate local document search with an automated agent is using MCP integration, which allows automated searches of your indexed markdown notes without manual CLI input.

On-device vector search works for large collections of personal documentation by utilizing local SQLite indexing to manage notes and transcripts efficiently without requiring external cloud services.