https://livekit.io logo
Join Slack
Powered by
what are all the prometheus metrics exposed by the...
# ask-ai
c
what are all the prometheus metrics exposed by the livekit workers? Currently i only see very basic ones. where can I find the exhaustive list of metrics and their explanation?
t
Thanks for reaching out, Bharath. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
Thanks — you're asking for the exhaustive list and explanations of Prometheus metrics exposed by LiveKit workers. I can help. Short answer: • The full metric names and labels are exposed on each worker’s Prometheus 
/metrics
endpoint (e.g., http://worker-host:port/metrics). Inspecting that endpoint on a running worker gives the authoritative, exhaustive list. • LiveKit emits a rich set of observability metrics across latency, STT/LLM/TTS timings, token usage, queue depth, per-turn/per-session metrics, and plugin/component metrics (TTFT, TTFB, token counts, end_of_turn detection, etc.). The docs list metric concepts and per-turn metrics and show how metrics are reported in events and dashboards. Where to find details: • Open the worker’s 
/metrics
endpoint to see exact metric names and labels for your deployment. • Use the LiveKit observability pages for explanations and examples: ◦ Metrics reference and per-turn latency: https://docs.livekit.io/deploy/observability/data/#metrics-reference ◦ Agent observability overview: https://livekit.com/products/agent-observability ◦ Realtime model metrics: https://docs.livekit.io/recipes/metrics_realtime If you want, I can: • Check a sample 
/metrics
output if you paste it here and explain specific metric names/labels you’re seeing. • Help locate the 
/metrics
endpoint for your environment (self-hosted vs LiveKit Cloud). Which are you running? Sources: Server-side | livekit/livekit_composite | DeepWiki | LiveKit Agent Observability | Debug and optimize voice agents | LiveKit | Understand and Improve Agent Latency | LiveKit | Build Your First AI Voice Agent in Python: Complete Tutorial | LiveKit | Agents (Python) | livekit/livekit_composite | DeepWiki, with interactive elements
c
I'm checking about the 
prometheus_multiproc_dir
. What child job process metrics are captured in prometheus? just want to know how useful they are
t
Thanks — you're asking which child job/process metrics are captured when LiveKit workers are configured with 
prometheus_multiproc_dir
, and how useful those metrics are. Quick answer: • I don't have an enumerated list of exact metric names for LiveKit's multiprocess setup. The best way to see which per-child metrics are exposed in your deployment is to inspect the 
/metrics
output from your workers (the multiproc exporter aggregates metrics from all child processes into that endpoint). • Typical process-level metrics you should expect from a Prometheus multiprocess setup are: process CPU usage, memory (RSS/virtual), file descriptor counts, thread counts, and process start time. These are useful for spotting resource pressure, memory leaks, and abnormal child restarts/creations. What I can help with next: • If you paste a snippet of your 
/metrics
output (or the contents of your 
prometheus_multiproc_dir
files), I’ll identify which metrics are per-child and explain their usefulness. • Tell me whether you’re self-hosted or on LiveKit Cloud and which worker/process type you mean (main server, audio worker, realtime/agent worker, or custom job). That helps me give focused guidance., with interactive elements
3 Views
Open in Slack
PreviousNext
Powered by