Voice commands

The microphone button in the chat and the Speak a request command open the “What do you want?” panel. You speak, review the transcript, and confirm.

• Dictation understands spoken punctuation, such as “comma” and “period”.
• The IDE works out whether the request is a task for the Maestro, a one-off conversation, or a message, as well as the profile, repository, and done criteria.
• In the IDE, transcription uses an OpenAI-compatible API: OpenAI, Groq, or a local server. Set it up in Set up voice. The audio stays in memory only.
• On the phone, the app uses the browser’s speech recognition.