Voice

Voice is the interface the system was designed around. You speak an intent, it routes, and the answer comes back spoken. The desktop application is a client of that path, not the product.

How it fits together

A desktop application with a Python backend handles the speech connection and the canvas, and passes resolved intents on. Work is handed out over event streams per space, so a slow task does not block you from speaking the next one. The voice layer executes nothing itself.

There are several routes from voice into the rest of the system, tried in order. The two routes that go through the Brain are switched off by default.

Language

The voice paths were built German-first. Broader coverage across German and English is active work. Other European languages are planned, not present.