openllmetry

Why we included this project

When you run LLM features in production, you often can't see what's happening inside a model call. How many tokens did that request use? Which provider responded slowly? Did an automatic retry double your bill? OpenLLMetry fixes that by adding OpenTelemetry instrumentation to the SDKs you already have, so telemetry for prompts, token counts, latency, and provider calls shows up without you writing exporters. Because it outputs standard OpenTelemetry data, it feeds into Datadog, Grafana, Honeycomb, or Sentry instead of forcing a new dashboard on you. That's handy for small teams that want to keep their monitoring stack and still see what live model calls are doing, and for engineers tracking down slow responses or unexpected token usage.