BeaconFind
Press Ctrl+K and search inside your files by meaning, not filename. Every model runs on your own machine, so files, queries and the index never leave it. Solo project: Rust and Tauri backend, React and TypeScript front end, SQLite with encryption at rest, and a hybrid search engine that merges keyword and semantic results. Ask questions about a document, search photos by what is in them, find duplicates; more than 70 file types and over 50 languages. Version 1.3 on Windows and macOS, sold at beaconfind.app and listed on the Microsoft Store.
Why it exists
Windows searches file names. That works right up to the moment you cannot remember what you called something, which for the files that actually matter is most of the time. BeaconFind searches what is inside them instead: press Ctrl+K, describe the thing, get the document back.
How it is built
The engine is hybrid, because neither half is enough on its own. A keyword index catches the things you can quote exactly – a part number, an error code, a name – and a semantic index catches the things you can only describe. The two result sets are scored and merged rather than run one after the other. Models run locally through Candle, the index lives in SQLite and is encrypted at rest, and files, queries and the index itself never leave the machine. On top of the search sits a small local language model that answers questions about what it found and shows the sources it used, and an image model that captions photos so they can be found by description. Rust and Tauri underneath, React and TypeScript on top.
Built with
- Rust
- Tauri
- React
- TypeScript
- SQLite
- Candle
Role Solo: architecture, backend, front end, design, docs, launch
Questions or problems with BeaconFind: [email protected]