feat: realtime voice demo (fastrtc + Cartesia) alongside legacy Gradio#1

Open

deepmind11 wants to merge 10 commits intomainfrom

feat/realtime-voice

Commits on Apr 15, 2026

feat(primary): widen history window from 6 to 20 turns
deepmind11
committed
docs: realtime voice architecture plan
deepmind11
committed
feat: realtime voice demo via fastrtc + Cartesia streaming TTS
deepmind11
committed
fix(realtime): restore rich UI, correct TTS language hint, warn on Groq free tier
deepmind11
committed
fix(realtime): size WebRTC component so Record button is visible
deepmind11
committed
fix(realtime): accept fastrtc's placeholder arg in voice handler
deepmind11
committed
fix(tts): native voice per language, Spanish stops sounding American
deepmind11
committed
fix(primary): detect current-utterance language and inject directive
deepmind11
committed
fix(realtime): drop gr.State from AdditionalOutputs so UI updates every turn
deepmind11
committed
fix(realtime): restore module-level _state binding
deepmind11
committed