qmd

Install qmd via npm, add your notes directory with `qmd collection add`, run `qmd embed` to generate vector embeddings, then use `qmd vsearch` for semantic queries or `qmd query` for hybrid search with reranking. Everything runs locally with no cloud services.

`qmd search` is BM25 keyword search that runs in ~0.2s with no models loaded. `qmd vsearch` is semantic vector search using one embedding model. `qmd query` is hybrid search combining query expansion, parallel BM25 and vector retrieval, and LLM reranking for the best quality.

Yes, qmd runs entirely on-device. It auto-downloads three local GGUF models (~2GB total) for embeddings, reranking, and query expansion on first run, and stores its index in a local SQLite database at ~/.cache/qmd/index.sqlite.

Run `qmd mcp` for stdio mode or `qmd mcp --http --daemon` for a persistent HTTP server, then add the server to your agent's MCP configuration. This registers tools like mcp_qmd_search, mcp_qmd_deep_search, and mcp_qmd_get automatically.

Cold start takes ~19s because the GGUF models must load into memory. Use the HTTP daemon mode (`qmd mcp --http --daemon`) to keep models warm, or use `qmd search` for BM25-only queries that need no models.

qmd requires Node.js 22 or later and is supported on macOS and Linux. On macOS you must also install SQLite via Homebrew because the system SQLite lacks extension loading support.