What is the recommended resource amount for deploy...
# ask-ais
straight-painting-40361
08/15/2025, 2:46 PMWhat is the recommended resource amount for deploying a voice-service - I'm still using Livekits managed service but am just deploying the Agent on my python app
t
tall-airline-16756
08/15/2025, 2:47 PMThanks for reaching out, Amir. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
tall-airline-16756
08/15/2025, 2:47 PMHey there! If you're looking to deploy a voice service using LiveKit's managed service with the Agent on your Python app, here’s a friendly rundown of the resource recommendations:
• CPU and Memory: A good starting point is 4 CPU cores and 8GB of memory per worker. This setup typically handles around 10-25 concurrent sessions, depending on the plugins you’re using.
• Number of Workers: For about 25 concurrent sessions, 1 worker should do the trick. If you’re aiming for 100 sessions, you might want to scale up to 4-8 workers.
• Networking: Your workers just need outbound WebSocket connections to LiveKit servers. You can also optionally set up a health check endpoint at
<HEALTH_CHECK_ENDPOINT>LIVEKIT_URLLIVEKIT_API_KEYLIVEKIT_API_SECRETOPENAI_API_KEYs
straight-painting-40361
08/15/2025, 3:47 PMTell me more about the HEALTH_CHECK_ENDPOINT
t
tall-airline-16756
08/15/2025, 3:48 PMThe HEALTH_CHECK_ENDPOINT for LiveKit worker deployment helps keep an eye on the health and availability of your worker instances. Here’s a quick rundown:
• Default Endpoint: It listens on
<HEALTH_CHECK_ENDPOINT>portWorkerOptionss
straight-painting-40361
08/15/2025, 3:54 PMCan you show me code for setting it up
t
tall-airline-16756
08/15/2025, 3:54 PMHere’s a quick example to help you set up the HEALTH_CHECK_ENDPOINT for a LiveKit worker in Python:
Copy code
from livekit import agent
from livekit.agents import Worker, WorkerOptions
# Custom worker configuration with a different health check port
options = WorkerOptions(
entrypoint_fnc=async def(job_context):
# Your job entrypoint logic here
pass,
port=8082, # Change the health check port to 8082
)
# Initialize and run the worker
worker = Worker(options)
worker.run()portWorkerOptions<http://localhost:<port>/>s
straight-painting-40361
08/16/2025, 9:55 AMDoes the livekit agent need to be open to requests from the public? Which service sends the request to the Agent ?
t
tall-airline-16756
08/16/2025, 9:55 AMHey there! Just so you know, the LiveKit agent doesn’t need to be open to public requests. It chats with the LiveKit server through a WebSocket connection, so there’s no need to expose any inbound ports.
When a user joins a room, the server sends the request to an available worker, which then runs the agent code. So, all the communication flows smoothly behind the scenes.
If you have any questions or want me to explain anything further, just give me a shout—I’m here to help!
Sources: Deployment and scaling | LiveKit Docs | Worker lifecycle | LiveKit Docs
s
straight-painting-40361
08/16/2025, 9:56 AMDo I need to open a port for the livekit server to be able to contact my Agent service?
t
tall-airline-16756
08/16/2025, 9:56 AMGood news! You don’t need to open any ports for the LiveKit server to reach your Agent service. The Agent service (worker) actually makes an outbound WebSocket connection to the LiveKit server, so there’s no need to worry about inbound ports on your end.
If you have any more questions or want me to explain anything else, just let me know—I’m here to help!
Sources: Deployment and scaling | LiveKit Docs | Deploying to production | LiveKit Docs | Ports and firewall | LiveKit Docs