semtools

Why we included this project

Semtools is a Rust-based command line tool for anyone who sits on a pile of documents and wants to actually search inside them. Its `search` command indexes plain text files with multilingual embeddings and ranks matches by cosine similarity, so a query turns up relevant lines even when the exact wording doesn't line up. The `parse` command converts PDF, DOCX, and similar formats to markdown with caching so re-runs don't redo the work, while the `ask` agent chains search and reading to answer questions over a whole collection and can point at any OpenAI-compatible API. The Unix-friendly design with proper stdin/stdout handling makes it easy to slot into existing shell pipelines rather than standing up a web service.