fix(xai): align realtime Voice Agent defaults and API coverage - #6755
Conversation
Default to grok-voice-latest with grok-transcribe, add force_message say(), and handle live caption updates so the LiveKit plugin matches current xAI Realtime behavior (and closes the Pipecat-era default-model gap). Co-authored-by: Cursor <cursoragent@cursor.com>
|
|
![devin-ai-integration[bot]](https://avatars.githubusercontent.com/in/811515?s=60&v=4){kind=small}
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
Co-authored-by: Cursor <cursoragent@cursor.com>
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
Match the OpenAI realtime cancellation path so interrupted force_message calls do not leave has_active_generation stuck or mis-tag later responses. Co-authored-by: Cursor <cursoragent@cursor.com>
Shrink the force_message/response.created race by only tagging replies after the create is on the wire, and support overlapping say() calls. Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
Avoid splitting the plugin module in this PR; behavior is unchanged. Co-authored-by: Cursor <cursoragent@cursor.com>
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
Cancel/timeout previously cleared the pending say id before response.created, so the base discard path never matched and cancelled scripted speech could still play. Co-authored-by: Cursor <cursoragent@cursor.com>
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
A bare cancel before force_message would silence an unrelated in-flight reply. Co-authored-by: Cursor <cursoragent@cursor.com>
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
Address review feedback: one async task owns send/wait/timeout/cancel, kept in _say_tasks so it cannot be GC'd, and cancelled from aclose(). Co-authored-by: Cursor <cursoragent@cursor.com>
This comment was marked as resolved.
This comment was marked as resolved.
Sorry, something went wrong.
aclose() cancelled the task and discarded by id but left the caller's future pending, which could hang AgentActivity awaiting say(). Co-authored-by: Cursor <cursoragent@cursor.com>
3 participants
Summary
grok-voice-latest(resolves togrok-voice-think-fast-2.0) and default input transcription togrok-transcribeforce_messagesupport viaRealtimeSession.say()/supports_say, emit live caption updates fromconversation.item.input_audio_transcription.updated, and forwardreasoning/ transcription / speed optionslivekit-xai-issues.mdTest plan
uv run pytest tests/test_realtime/test_xai_realtime_model.py -q --allow-uncategorized(20 passed)XAI_API_KEY=… uv run pytest tests/test_realtime/test_realtime.py -k xai -q --allow-uncategorized(16 passed)grok-voice-latest,session.update(reasoning / idle_timeout / grok-transcribe), cancel, force_message, deleteMade with Cursor