https://livekit.io logo
Join Slack
Powered by
How can i disable interruptions between ai speakin...
# ask-ai
w
How can i disable interruptions between ai speaking :-
Copy code
session = AgentSession(
        llm=google.beta.realtime.RealtimeModel(
            model=model,
            voice=voice,
            temperature=temperature,
            instructions=instructions,
            language="en-IN"
        ),
        min_endpointing_delay=5,
    )

 await session.start(
        agent=assistant,
        room=ctx.room,
        room_input_options=RoomInputOptions(
            # LiveKit Cloud enhanced noise cancellation
            # - If self-hosting, omit this parameter
            # - For telephony applications, use `BVCTelephony` for best results
            noise_cancellation=noise_cancellation.BVC(),
            # Don't close session when participants disconnect - allow multiple users
            close_on_disconnect=False,
        ),
    )
t
Thanks for reaching out, Rahul. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
To help you disable interruptions between AI speaking in your LiveKit session, you can use the 
allow_interruptions
parameter. Let me walk you through two simple ways to do this: • Disable Interruptions at the Agent Level: You can set 
allow_interruptions=False
when you initialize the agent. This way, all speech generated by the agent won’t be interrupted.
Copy code
agent = Agent(
    stt=...,  # your STT instance
    llm=...,  # your LLM instance
    tts=...,  # your TTS instance
    allow_interruptions=False  # Disables interruptions for all agent speech
)
• Disable Interruptions for Specific Speech: If you want to prevent interruptions for just certain utterances, you can use 
allow_interruptions=False
in methods like 
session.say()
.
Copy code
await session.say(
    "This message cannot be interrupted.",
    allow_interruptions=False
)
These options should help make your AI speech flow more smoothly without interruptions during your sessions. If you have any questions or need a hand with anything else, please don’t hesitate to ask — I’m here to help! Sources: pipeline-llm/interrupt_user.py | packages/basics/src/uninterruptable.ts | Agent speech and audio | LiveKit Docs | Agent speech and audio | LiveKit Docs
w
I set 
allow_interruptions
false in AgentSession and got this error
Copy code
0|livekit-agent  | ERROR:livekit.agents:unhandled exception while running the job task
0|livekit-agent  | Traceback (most recent call last):
0|livekit-agent  |   File "/home/ubuntu/ai-tutor/py-agent/.venv/lib/python3.12/site-packages/opentelemetry/util/_decorator.py", line 71, in async_wrapper
0|livekit-agent  |     return await func(*args, **kwargs)  # type: ignore
0|livekit-agent  |            ^^^^^^^^^^^^^^^^^^^^^^^^^^^
0|livekit-agent  |   File "/home/ubuntu/ai-tutor/py-agent/.venv/lib/python3.12/site-packages/livekit/agents/ipc/job_proc_lazy_main.py", line 240, in _traceable_entrypoint
0|livekit-agent  |     await self._job_entrypoint_fnc(job_ctx)
0|livekit-agent  |   File "/home/ubuntu/ai-tutor/py-agent/src/agent.py", line 1164, in entrypoint
0|livekit-agent  |     await session.start(
0|livekit-agent  |   File "/home/ubuntu/ai-tutor/py-agent/.venv/lib/python3.12/site-packages/opentelemetry/util/_decorator.py", line 71, in async_wrapper
0|livekit-agent  |     return await func(*args, **kwargs)  # type: ignore
0|livekit-agent  |            ^^^^^^^^^^^^^^^^^^^^^^^^^^^
0|livekit-agent  |   File "/home/ubuntu/ai-tutor/py-agent/.venv/lib/python3.12/site-packages/livekit/agents/voice/agent_session.py", line 502, in start
0|livekit-agent  |     await asyncio.gather(*tasks)
0|livekit-agent  |   File "/home/ubuntu/ai-tutor/py-agent/.venv/lib/python3.12/site-packages/livekit/agents/voice/agent_session.py", line 822, in _update_activity
0|livekit-agent  |     self._next_activity = AgentActivity(agent, self)
0|livekit-agent  |                           ^^^^^^^^^^^^^^^^^^^^^^^^^^
0|livekit-agent  |   File "/home/ubuntu/ai-tutor/py-agent/.venv/lib/python3.12/site-packages/livekit/agents/voice/agent_activity.py", line 131, in __init__
0|livekit-agent  |     raise ValueError(
0|livekit-agent  | ValueError: the RealtimeModel uses a server-side turn detection, allow_interruptions cannot be False, disable turn_detection in the RealtimeModel and use VAD on the AgentSession instead
t
Thanks for reaching out! The error you're seeing happens because the 
RealtimeModel
uses server-side turn detection, which doesn't quite work when you set 
allow_interruptions=False
. No worries though, I’m here to help you through it! You have two great options to fix this: • Option 1: Disable Server-Side Turn Detection You can turn off the Realtime model's turn detection and instead use client-side VAD for managing interruptions. This gives you full control, though it might add a bit of latency.
Copy code
session = AgentSession(
    allow_interruptions=False,
    vad=silero.VAD.load(),
    llm=openai.realtime.RealtimeModel(
        turn_detection=None,
        voice="alloy"
    )
)
• Option 2: Use Server-Side Turn Detection Alternatively, you can keep 
allow_interruptions=True
and rely on the Realtime model’s built-in turn detection, which is optimized for performance. You can even tweak its behavior with settings like 
threshold
and 
silence_duration_ms
.
Copy code
session = AgentSession(
    llm=openai.realtime.RealtimeModel(
        turn_detection=openai.realtime.ServerVadOptions(
            threshold=0.6,
            silence_duration_ms=500
        )
    )
)
For simplicity and smooth performance, I usually recommend Option 2. But if you have any questions or want help setting this up, just let me know—I’m happy to assist! Sources: Turn detection and interruptions | LiveKit Docs | plugins/openai/src/realtime/realtime_model.ts | livekit-agents/livekit/agents/voice/events.py | Voice activity detection (VAD) parameters | LiveKit Docs
2 Views
Open in Slack
PreviousNext
Powered by