Voice mode

How it works

Click the mic in the dock to open the hands-free voice panel. The loop is: listen (the orb reacts to your voice) → think → speak the answer → listen again.

Speech-to-text

• Groq / Whisper (via the weeagent-transcribe edge function) when GROQ_API_KEY is set — recommended, accurate, all browsers.
• Browser Web Speech as a fallback (Chrome/Edge).

Text-to-speech

• ElevenLabs (via weeagent-tts) when ELEVENLABS_API_KEY is set — reliable audio, real amplitude drives the orb.
• Browser synthesis as a fallback (a reliable local voice is auto-picked).

Daily greeting

On the first open of the day, on the user's first gesture (autoplay policy), WeeAgent greets them by name. Disable with dailyGreeting={false} on the provider; append ?greet to the URL to force a replay while testing.