cjpais/Handy — reverse-engineered prompt

Build me a simple desktop app called Handy that lets me do fully offline voice transcription on my own computer. I want to press a keyboard shortcut, talk, then have the transcribed text show up in whatever text box I was using. It should feel privacy focused and easy to use, with no cloud requirement. Please make it work on Windows, macOS, and Linux, with a small settings screen where I can choose shortcuts, pick a transcription model, choose language options, and switch between normal tap to start and push to talk.

I also want it to filter silence automatically, support local Whisper style models plus a faster CPU friendly option with automatic language detection, and use GPU acceleration when available. Add useful desktop app touches like a tray icon, start hidden support, basic transcript history, and simple command line controls to start or stop recording from outside the app. If there are platform quirks, especially on Linux, handle them as gracefully as possible and document setup clearly. Look up current docs online if you need to.