FluxyChat

Features

Voice stages

Discord Stages-style speaker and listener roles with active-speaker detection in voice rooms.

Voice stages

Split voice participants into speakers and listeners. Only speakers publish audio by default. Listeners request to speak or are promoted by moderators.

Console / in-room UI

Voice AI (/voice-ai) and the in-room Voice stage panel (VoiceStagePanel):

  • Toggle speaker / listener role
  • Request to speak queue for listeners
  • Moderator promote / demote
  • Active speaker highlight (Silero VAD)

Active speaker

useVoiceStageVad plus Silero VAD highlights who is speaking. Same pipeline as voice AI.

API

GET  /rooms/:roomId/voice-stage
POST /rooms/:roomId/voice-stage/join
POST /rooms/:roomId/voice-stage/role
POST /rooms/:roomId/voice-stage/request-speak

Stage state also syncs via Room DO WebSocket presence (speaker | listener).

Environment

Voice pipeline env vars (AI_*, TURN) apply when WebRTC is used alongside stages.

Ideal for AMAs, town halls, and large voice rooms without everyone unmuted.

Offline sync, federation, and merge conflicts

How offline queues, Matrix bridges, and merge-conflict UI fit together in FluxyChat.

Empathy layer (voice)

Opt-in prosody detection adapts agent tone and escalation without telling the user they seem stressed.

On this page

Voice stagesConsole / in-room UIActive speakerAPIEnvironmentRelated