Thanks — you're asking how to ensure traffic is distributed between two (or more) EC2 instances running the Python LiveKit agent. I’ll address that specifically. Short answer • You do not put an ELB or standard HTTP load balancer in front of agent processes. Run the same agent worker on each EC2 instance and LiveKit’s server will register and dispatch jobs to those workers, distributing rooms/sessions across instances automatically. What to do • Run the Python agent worker on each EC2 instance (same code/config). Each worker opens an outbound connection to the LiveKit server and registers itself. • Let LiveKit perform job dispatch (round‑robin / single assignment) so each room/session is assigned to one worker. • Ensure instances have outbound network access to the LiveKit server (no inbound ports required). • Use the agent’s built‑in load control (e.g.,

load_fnc

and

load_threshold

) so unhealthy or overloaded instances stop receiving new jobs. • For autoscaling, add/remove instances running the worker; new instances auto‑register and begin receiving jobs. Scale based on CPU or other metrics (scale up before hitting the agent’s

load_threshold

). Optional • If you need geographic affinity, place EC2 instances near your users or use LiveKit Cloud features for region routing. • Store secrets (API keys) in environment variables or a secrets manager on each instance. If you’d like, tell me: • whether you’re using LiveKit Cloud or a self‑hosted LiveKit server, and • how you plan to autoscale (manual, AWS ASG), so I can give precise examples or sample

Worker

configuration. Sources: Custom agent deployments | LiveKit docs