Talk and realtime voice migration
Realtime voice, telephony, meeting, and browser Talk code shares one Talk session controller exported byopenclaw/plugin-sdk/realtime-voice. The
controller owns the common Talk event envelope, active turn state, capture
state, output-audio state, recent event history, and stale-turn rejection.
Provider plugins own vendor-specific realtime sessions. Browser-meeting plugins
use openclaw/plugin-sdk/meeting-runtime for session, browser, audio, node-host,
agent-consult, and voice-call mechanics, then implement MeetingPlatformAdapter
for URL rules, DOM scripts, manual-action mapping, captions, creation, and dial-in
plans. Platform REST APIs, OAuth, artifacts, selectors, and wire names remain in
the plugin. Browser permission plans receive the requested meeting URL so each
platform can grant only its exact supported origins. Session runtimes must also
normalize platform-specific live health after confirmed browser departure;
historical transcript fields may remain, but caption and audio readiness must
not stay active after leave.
All bundled surfaces run on the shared controller: browser relay,
managed-room handoff, voice-call realtime, voice-call streaming STT, Google
Meet realtime, and native push-to-talk. Gateway advertises one live Talk event
channel in hello-ok.features.events: talk.event.
New code should not call createTalkEventSequencer(...) directly unless
implementing a low-level adapter or test fixture. Use the shared controller so
turn-scoped events cannot be emitted without a turn id, stale turnEnd /
turnCancel calls cannot clear a newer active turn, and output-audio
lifecycle events stay consistent across telephony, meetings, browser relay,
managed-room handoff, and native Talk clients.
The public API shape:
talk.client.create,
because the browser owns provider negotiation and media transport while the
Gateway owns credentials, instructions, and tool policy. talk.session.* is
the common Gateway-managed surface for gateway-relay realtime, gateway-relay
transcription, and managed-room native STT/TTS sessions.
Legacy configs that place realtime selectors beside talk.provider /
talk.providers should be repaired with openclaw doctor --fix; runtime Talk
does not reinterpret speech/TTS provider config as realtime provider config.
The supported talk.session.create combinations are intentionally small:
Method map for readers migrating from the older
talk.realtime.* /
talk.transcription.* / talk.handoff.* families (all removed):
The unified control vocabulary is also deliberately narrow:
Do not introduce provider or platform special cases in core to make this work.
Core owns Talk session semantics. Provider plugins own vendor session setup.
Voice-call and Google Meet own telephony/meeting adapters. Browser and native
apps own device capture/playback UX.