A local-first, private search engine for your Markdown knowledge base (benjaminetienne.com)

PMS (Personal Markdown Search) is a newly launched, lightweight, local-first hybrid search engine and web clipper tailored for personal Markdown knowledge bases. Developed in Python and built on SQLite with FTS5 and ONNX for fast embeddings, PMS aims to provide an efficient alternative to heavier tools like qmd. Unlike its counterparts, PMS favors a minimalist design, enabling offline functionality and maintaining user data privacy by keeping sensitive information stored locally. Its architecture supports light-weight model execution, achieving rapid search results without the complexities of larger, dependency-heavy frameworks. Significantly, PMS introduces structure-aware Markdown chunking, allowing users to efficiently search through their documents while preserving context. By utilizing a hybrid retrieval approach that combines keyword and semantic search, it employs a reciprocal rank fusion score for effective result ranking. The modular and hackable nature of PMS makes it adaptable for various use cases, encouraging developers to customize its functionality as needed. These features position PMS as a formidable tool for users seeking a flexible, privacy-centric solution for managing their Markdown-based knowledge.