ADK Observability Setup Guide

This skill guides you through setting up observability for agents built with the Google ADK (Agent Development Kit) and deployed to the cloud. It kicks in on requests like "set up tracing," "monitor my ADK agent," "configure logging," "debug production traffic."

What it does

It guides you through setting up a three-layer monitoring architecture:

• Cloud Trace — distributed tracing with OpenTelemetry spans (latency, error flow), always on.
• Prompt-Response Logging — exports GenAI interactions to GCS and BigQuery; privacy-preserving default (NO_CONTENT).
• BigQuery Agent Analytics — writes structured agent events (LLM calls, tool usage) to BigQuery.

It also covers third-party platform integrations like AgentOps, Phoenix, and MLflow, the span hierarchy, the critical setup ordering (running infra single-project before deploy), and troubleshooting steps.

When to use it

after taking an ADK agent to production, when you need to observe traffic, audit LLM interactions, debug latency/errors, or collect logs for compliance (KVKK).

Output

A step-by-step Markdown setup guide: a recommended-layer table, a runnable agents-cli command block, a verification checklist, and a troubleshooting table.

Note: For deployment setup use google-agents-cli-deploy, and for API code patterns use google-agents-cli-adk-code.