semquery
A local, offline-ready Rust command-line tool for retrieval-augmented document search and question answering, storing everything in a single SQLite index.
semquery indexes a personal document collection and lets a user search passages or ask questions with cited answers, entirely on their own machine: indexes, models and queries never leave the device. It combines BM25 keyword search, dense vector search, reciprocal rank fusion and cross-encoder reranking, with Chinese-optimized sentence-level chunking and jieba tokenization for BM25.
Everything lives in one SQLite database using sqlite-vec and FTS5, and its ask command returns natural-language answers with inline citations pointing back to source files. Supported document formats include Markdown, plain text, PDF and Microsoft Word, with PDF and DOCX support enabled by default through optional feature flags. Its core traits live in a separate library crate so heavier backends stay isolated behind feature flags.
It installs from source with Cargo, and commands include init, add, index, search and ask.