Add a /voice slash command (and keybinding) to the TUI that records audio push-to-talk style, shows a recording indicator in the prompt area, transcribes the audio, and inserts the transcript into the prompt input without auto-submitting.
Requirements:
- Record via an external binary with graceful degradation: prefer
sox (rec), fall back to arecord (Linux) or ffmpeg; if none is installed, show a toast explaining what to install.
- Transcription via a new server endpoint that uses the OpenAI Whisper API (
whisper-1) with the user's configured OpenAI credentials; clear error toast when no OpenAI credential is configured.
- Keep transcription pluggable (single function) so other providers can be added later.
- Do not auto-submit the transcribed text.
Add a
/voiceslash command (and keybinding) to the TUI that records audio push-to-talk style, shows a recording indicator in the prompt area, transcribes the audio, and inserts the transcript into the prompt input without auto-submitting.Requirements:
sox(rec), fall back toarecord(Linux) orffmpeg; if none is installed, show a toast explaining what to install.whisper-1) with the user's configured OpenAI credentials; clear error toast when no OpenAI credential is configured.