QMD – Quick Markdown Search (github.com)

QMD, a new on-device search engine, has been launched to streamline the retrieval of markdown notes, meeting transcripts, and documentation by utilizing a combination of BM25 full-text search, vector semantic search, and LLM re-ranking, all executed locally through node-llama-cpp with GGUF models. Users can easily create collections for their documents and enhance search capabilities with added context, allowing for successful queries using either keywords or natural language. Notably, the tool facilitates fast and hybrid searching, enabling users to find relevant documents swiftly, integrate with automatic workflows, and output search results in structured formats suitable for further processing. The significance of QMD lies in its ability to enhance productivity for users who manage extensive knowledge bases. By running entirely on-device, it addresses privacy concerns while maximizing performance through advanced search algorithms and real-time indexing. The introduced hybrid search method combines multiple retrieval techniques, ensuring users receive the most relevant information based on their queries. This tool could revolutionize how developers, researchers, and any knowledge workers interact with their notes and documents, making information access more efficient and context-aware.