fidx released — local semantic search in a single SQLite file

🛠 fidx released — local semantic search in a single SQLite file

The tool combines hybrid search (BM25 + vector search via sqlite-vec) into a single SQLite file, running entirely on the CPU without the need to call an LLM during the query. It supports markdown, text files, chat exports, and code.

🌍 A new approach to client-side RAG (Retrieval-Augmented Generation) implementation has been introduced, which radically reduces latency by excluding the LLM from the query execution path and utilizing optimized SQLite extensions.

👤 You can now organize instant and private search across your local knowledge base or chat history without sending data to the cloud or wasting resources on heavy models for every search.

Source 1: https://github.com/williamliu-ai/fidx