```Traceback (most recent call last): File "/app...
# agents
r
Copy code
Traceback (most recent call last):
  File "/app/venv/lib/python3.13/site-packages/livekit/rtc/audio_stream.py", line 276, in _run
    if self._processor is not None and self._processor.enabled:
       ^^^^^^^^^^^^^^^
AttributeError: 'AudioStream' object has no attribute '_processor'
Team we are running into a an issue with the TTS. Any ideas what might be going on? Out agent isn't responding in any of our calls. The error isn't very helpful either.
d
@rapid-painting-51133 when does this take place? have you changed any package dependencies?
r
It just started happening, about 30 mins ago
No updates on lk version, using latest 1.3.6
d
can you share an example session?
has there been redeploys or anything?
r
yes we had deployed a small change, removed the English turn detection model and just keeping Multilinguial as English is on deprecation path
tried reverting but still same issue
d
ok, that probably pulled in a later version of the
livekit
package
for now, could you do:
Copy code
livekit==1.0.20
r
we are using agents, do you want me to add this as a new dependecy?
to override?
Copy code
dependencies = [
    "coloredlogs>=15.0.1",
    "livekit-agents[cartesia,deepgram,elevenlabs,openai,silero,turn-detector]==1.3.6",
    "livekit-plugins-noise-cancellation~=0.2.5",
    "pydantic>=2.11.5",
    "pydantic-settings>=2.9.1",
    "tenacity>=9.1.2",
    "python-dotenv>=1.1.0",
    "azure-monitor-opentelemetry>=1.7.0",
    "opentelemetry-instrumentation-httpx>=0.46b0",
    "opentelemetry-sdk>=1.34.1,<1.40.0",
    "opentelemetry-api>=1.34.1,<1.40.0",
]
Our current dependencies
d
add an explicit one there
r
got it adding
it's deploying
okay service back up
Thanks David! 🙏
How can I prevent something like this from happening again? All we did was a simple change which crashed all of our calls. Reverts didn't really work either.
d
glad to hear it.. yeah.. this is an issue on our side. the bug was in the livekit SDK, which got released a few hours ago. redeploys would've picked up that change due to the >= version specifier in the agent framework
Cloud agent rollbacks should have worked though.. but we need to do a better job of preventing these from happening in the future.
r
Got it, all good. The cloud agent rollback is different then just reverting code change and deploying right? We have a CI/CD pipeline that auto deploys, but we never use the rollback command.
d
yes. that's different.. it uses the exact image that was built
in case future incidents with deploys occur, it should be the quickest way to get back up and running
r
Got it, good to know. Will keep that for future reference. Thank you!!!