We have AI that can write complex code, but if I type a slightly misspelled keyword into my OS search bar, my computer acts like the file doesn't exist.
I was wasting so much time hunting down old PDFs, code snippets, and design assets that I finally snapped and decided to build my own desktop search engine.
Meet Vexil.
The Tech Stack
I wanted this thing to be blazingly fast and run in the background without draining battery, so Electron was immediately off the table.
- Backend: Rust π¦
- Framework: Tauri (the resulting binaries are tiny and the memory footprint is practically nothing).
- Frontend: React + TypeScript.
- Database: Local SQLite for lightning-fast indexing.
- Intelligence: Local embedding models so it understands semantic context without sending your private files to an OpenAI server.
The hardest part?
Getting cross-platform file system crawling to be performant. Rust's ignore and notify crates were lifesavers here, but managing background thread indexing without locking up the UI thread took some serious architecture rewrites.
Itβs completely local, runs offline, and you can trigger it anywhere with a global hotkey.
I just pushed the first release binaries for macOS, Windows, and Linux to GitHub.
Iβd love for the community to try it out. If you're curious about Tauri vs Electron, or how local SQLite holds up for full-text search, ask me anything in the comments!
π Download it here:
Top comments (0)