Thanks — I see the error you’re hitting:

twirp_code=resource_exhausted twirp_status=429

. That indicates the request was rejected due to resource limits (rate limiting/concurrency caps). I can help troubleshoot — to give the most useful next steps, could you share a few details? • Which API/service call produced the 429 (for example: Egress, SIP, Agent, inference, WebRTC connect)? • Your LiveKit plan/tier (self‑serve tier or Scale/Enterprise)? • Is this a short burst of traffic or sustained higher traffic? • Approximate traffic numbers when it happens (requests/sec or concurrent sessions)? • Any recent changes in traffic pattern or deployment? Meanwhile, immediate things you can try: • Implement exponential backoff retries for 429 responses to smooth bursts. • Check your LiveKit dashboard quota widgets to see current usage vs limits. • If this is inference concurrency hitting a hard cap, increasing the plan is required to raise that limit. For other limits, a plan increase or quota uplift can help. Share the details above and I’ll recommend the precise next steps for your situation. Sources: How much can LiveKit scale? | LiveKit | server-sdk-go | livekit/livekit_composite | DeepWiki