Local search engine that indexes your docs, notes, and meetings with hybrid BM25+vector search and LLM reranking - perfect for AI agents and knowledge workers who want Google-quality search without the cloud

Finally, a search engine that treats your personal knowledge base like it deserves. QMD solves the eternal problem of finding that one crucial detail buried in months of meeting notes, documentation, and random markdown files. While most people resort to grep or pray their note-taking app’s search doesn’t suck, QMD brings enterprise-grade search to your local filesystem.

What makes this special is the triple threat: blazing-fast BM25 keyword search for exact matches, semantic vector search for concept-based queries, and LLM reranking that actually understands context. All running locally with GGUF models via node-llama-cpp. The MCP (Model Context Protocol) integration is chef’s kiss - your AI agents can now actually find relevant context instead of hallucinating. With 7K+ stars and growing, the developer community clearly needed this.

Perfect for anyone drowning in documentation, researchers with extensive note collections, or teams building AI workflows that need reliable knowledge retrieval. The setup is refreshingly simple - point it at your folders, run qmd embed, and suddenly you have semantic search that rivals commercial solutions. No servers, no subscriptions, just pure local knowledge discovery.

⭐ Stars: 7342
💻 Language: TypeScript
🔗 Repository: tobi/qmd