raise RuntimeError(“This generation handle does not allow interruptions”) 5|Universal_Agent | RuntimeError: This generation handle does not allow interruptions 5|Universal_Agent | 072157.355 ERROR asyncio Task exception was never retrieved 5|Universal_Agent | future: <Task finished 5|Universal_Agent | name=‘Task-213’ 5|Universal_Agent | coro=<RoomIO._on_user_text_input.<… 5|Universal_Agent | done, defined at 5|Universal_Agent | /home/ubuntu/CMS_VA_Agent_Python/.… 5|Universal_Agent | exception=RuntimeError(‘This 5|Universal_Agent | generation handle does not allow 5|Universal_Agent | interruptions’)> 5|Universal_Agent | Traceback (most recent call last): 5|Universal_Agent | File 5|Universal_Agent | “/home/ubuntu/CMS_VA_Agent_Python/… 5|Universal_Agent | line 408, in _read_text 5|Universal_Agent | text_input_result = 5|Universal_Agent | text_input_cb( 5|Universal_Agent | ^^^^^^^^^^… 5|Universal_Agent | File 5|Universal_Agent | “/home/ubuntu/CMS_VA_Agent_Python/… 5|Universal_Agent | line 53, in _default_text_input_cb 5|Universal_Agent | sess.interrupt() 5|Universal_Agent | File 5|Universal_Agent | “/home/ubuntu/CMS_VA_Agent_Python/… 5|Universal_Agent | line 971, in interrupt 5|Universal_Agent | return 5|Universal_Agent | self._activity.interrupt(force=for… 5|Universal_Agent | ^^^^^^^^^^^^^^^^^^^^^^^… 5|Universal_Agent | File 5|Universal_Agent | “/home/ubuntu/CMS_VA_Agent_Python/… 5|Universal_Agent | line 948, in interrupt 5|Universal_Agent | self._current_speech.interrupt… 5|Universal_Agent | File 5|Universal_Agent | “/home/ubuntu/CMS_VA_Agent_Python/… 5|Universal_Agent | line 123, in interrupt 5|Universal_Agent | raise RuntimeError(“This 5|Universal_Agent | generation handle does not allow 5|Universal_Agent | interruptions”) 5|Universal_Agent | RuntimeError: This generation 5|Universal_Agent | handle does not allow interruptions I am getting this error

Thanks for reaching out, rajan. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.

I am creating agent session like this : session = AgentSession( stt=config[“stt”], llm=config[“llm”], tts=config[“tts”], turn_detection=MultilingualModel(), vad=ctx.proc.userdata[“vad”], preemptive_generation=True, user_away_timeout=None, # Disabled - using custom timer-based follow-up system mcp_servers=mcp_servers, )

Thanks — I see you shared your AgentSession creation. You're using

preemptive_generation=True

,

MultilingualModel()

for turn detection, and a custom VAD. That helps. Brief diagnosis • The RuntimeError you saw means an active speech/generation was marked uninterruptible (started with

allow_interruptions=False

) and then

session.interrupt()

was called. • With your config, two likely causes are: ◦ A speech handle was created as uninterruptible (explicitly or by default in some code path). ◦ A race between preemptive generation and your interrupt logic causes you to call interrupt when the current speech disallows interruptions or when no speech is active. Quick checks to try • Search your code for

say()

/

generate_reply()

calls and confirm whether

allow_interruptions

is set. If you want barge-in, set:

await session.say("...", allow_interruptions=True)
# or
session.generate_reply(..., allow_interruptions=True)

• When interrupting, use force or guard the call:

# safer interrupt
if session.current_speech:
    await session.current_speech.interrupt()
# or force
session.interrupt(force=True)

• Add logging around speech start/stop and interrupt calls to catch races:

<http://logger.info|logger.info>("starting speech")
handle = await session.say(...)
<http://logger.info|logger.info>("interrupting")
await session.interrupt(force=True)

If you prefer, share the snippets where you: 1) call

say()

/

generate_reply()

and 2) call

session.interrupt()

(or any place you trigger interrupt), and I’ll point out the exact change to avoid the RuntimeError. Sources: Turns overview | LiveKit docs | Agent session | LiveKit docs | Agent speech and audio | LiveKit docs | LiveKit turn detector plugin | LiveKit docs