Hey <@U08CKJH7C12> , we have a somewhat odd behav...
# agents
g
Hey @refined-appointment-81829 , we have a somewhat odd behavior with our agent workers , the agent won't join the room and this happens super randomly , what steps can we take to make this better and avoid this?
r
Make sure you are using unique room names each time.
g
Thanks , can you give me more hints about the origin of this issue? how does things work behind the scenes ?
r
Yes, agent dispatch generally happens on room creation. If a room already exists (like when a room name is reused) that agent will not be dispatched.
✅ 1
g
@refined-appointment-81829 Room creation process , generates a unique names per sessions , still this problem occurs , where else is to look?
r
Do you have a session ID where agent did not join? It will start with
RM_
or if this is a SIP call you can send the call id
SCL_
g
It is not a call , we've integrated livekit into our frontend in order to use it as a comprehensive chatbot , our logic create rooms , so in this case i should look for
RM_
in the dashboard , right? What should i look for exactly in this case?
r
I am looking for a session ID that had the issue you are seeing so I can look in the logs. A LiveKit session ID starts with
RM_
or I can probably look up by room name if you are struggling to find the session id
g
Thanks , I'll share the room ID tomorrow
@refined-toddler-89382 Could you share an example of one session where the worker stops?
r
@gorgeous-gpu-30432 It appeared on this room ID for example:
RM_NVRwZFxwGF5u
g
Thanks for checking @refined-appointment-81829 , could you please check this session ID for us and possibly tell use why the agent is not being connected to the room? Session ID:
RM_NVRwZFxwGF5u
r
For that room we see
Copy code
no worker available to handle job
So most likely you did not have an Agent with capacity to handle the job. This one reported it did not have capacity for the job
AW_gVdRbzwmcNnz
g
@refined-appointment-81829 I see , we are hosting our livekit agent application (livekit-agents python) on a linux server with the following hardware config and livekit packages:
livekit==1.0.12
livekit-agents==1.2.3
livekit-api==1.0.5
livekit-blingfire==1.0.0
livekit-plugins-anthropic==1.2.3
livekit-plugins-deepgram==1.2.3
livekit-plugins-elevenlabs==1.2.3
livekit-plugins-google==1.2.3
livekit-plugins-noise-cancellation==0.2.5
livekit-plugins-openai==1.2.3
livekit-plugins-rag==1.0.13
livekit-plugins-silero==1.2.3
livekit-plugins-turn-detector==1.2.3
livekit-protocol==1.0.4
We are planning to scale to around 1000+ users parallel sessions. which options should we look into to make this possible on our end and do we need to subscribe to a plan on livekit or we need to scale on our side? Would #C08LT757332 be a suitable option for multi-agentic use case with RAG, web search , sophisticated function tools and frontend RPC calls for us or should we scale our existing structure? Please let me know how we should proceed. I'll appreciate extra hints for scalability either from the docs or external sources for our project. Thank you!
r
✅ 1