QMD brings hybrid search to markdown

QMD is compelling because it treats note search like a real retrieval stack, not a thin UX layer over keyword matching. For developers building agent workflows, the MCP angle makes it more interesting than yet another “second brain” app.

• –Its hybrid pipeline mirrors modern RAG practice: lexical search for exact matches, vector search for semantics, then reranking to clean up the final results.
• –Running locally via GGUF models and `node-llama-cpp` is a real differentiator for privacy-conscious teams that want searchable docs without cloud lock-in.
• –The CLI, JSON output, and MCP server make QMD usable as infrastructure for coding agents and local knowledge workflows, not just as a personal search box.
• –Because it indexes plain markdown collections instead of a proprietary workspace, it fits existing docs-as-files and Obsidian-style setups with less migration pain.
• –The tradeoff is complexity: local embeddings, model downloads, and reranking are more powerful than simple note search, but they also raise setup and performance expectations.