Skip to content

Conversation

bfioca-openai
Copy link
Contributor

Summary

  • Dependencies: Bumped @openai/agents to ^0.1.9 and openai to ^6.2.0 (plus transitives) so the app targets the GA realtime stack.
  • Server: Replaced the deprecated /v1/realtime/sessions POST with openai.realtime.clientSecrets.create, supplying the GA session payload (model, audio formats, voice) and returning the expected shape.
  • Client:
    • Send GA-compliant session.update payloads (type: "realtime") and adopt the new audio.* shape while listening for GA transcript events (response.output_audio_transcript.*).
    • Fixed push-to-talk drift by clearing the buffer on release and toggling mute on press/release; the same mute logic now covers PTT and playback-disable flows.
    • Emit structured audio format descriptors from audioFormatForCodec.
  • Prompts: Switched all agent configs to the GA marin voice to match the updated transport.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant