Course Wiki

OpenTelemetry

Instrumenting an AI-built app with traces, metrics, and logs flowing into Prometheus, Loki, Tempo, and Grafana.

Module 4 starts from a deployment truth: a shipped app can still fail silently. OpenTelemetry instruments the application once, with vendor-neutral APIs, and emits three signal types — metrics (counts and latencies), logs (structured events), and traces (request journeys across services). The course wires those into a self-hosted stack: Prometheus for metrics, Loki for logs, Tempo for traces, all browsable in Grafana.

The observability stack feeds the module’s incident workflow: an actionable alert fires, a human (or the bounded agent responder) investigates using traces to find where the request went wrong. Observability is what makes the AI-assisted app operable after the agents stop writing code.

Taught in

• LLM Zoomcamp

  • Module 5: Monitoring
    • Next Steps

• AI Dev Tools Zoomcamp

  • Module 4: DevOps and Observability for AI-Built Apps
    • DevOps and Observability for AI-Built Apps ## Related concepts

• Prometheus and Grafana

• Playwright

• CI/CD

• Model Monitoring

• Coding Agents