npurag turns a directory into a continuously updated semantic index that you can query by meaning or ask questions against, all running on your laptop's NPU or a compatible backend. It combines embeddings with BM25 for hybrid retrieval and can serve results via CLI, HTTP, or the Model Context Protocol. Designed for developers and power users who need fast, offline access to their own documents without sending data to the cloud. Its on‑device focus makes it energy‑efficient and privacy‑preserving compared to cloud‑based alternatives.
View on GitHub →antumbra-ai/npurag