Wire onSubmit to Gemma for the phone screen and the video interview
Commit
927tanmay2d ago
- VideoEngine runs <AiVoiceAvatar> (Ananya, Kokoro fp32, no gestures under reduced motion) with the same config as the phone engine
- Both engines send each transcript to the app's Gemma worker and speak the streamed reply; barge-in stops generation
- voiceStore carries status, captions, mic errors and controls from the engine to the interview screen
- Start interview opens the mic inside the click; a placeholder interviewer asks the opening question until phase 3
- Hidden engine sits off stage at a real size so the avatar canvas mounts
- Spell it 'practice'; plan the voice-mode orb for step 4.2
Entire-Checkpoint: 01M4184H8F7GT8ZR8PF1HBHYFN