What problem does it solve? Deployed AI agents are hard to debug and audit without telemetry. This Skill guides you through setting up distributed tracing, prompt-response logging, and analytics for agents built with Google's agents-cli, so you can monitor production traffic, audit LLM interactions, and troubleshoot failures. ## Core Features & Use Cases - Cloud Trace Setup: Distributed tracing via OpenTelemetry spans across agent invocations, LLM calls, and tool executions, with setup guidance per deployment target (Agent Runtime, Cloud Run, GKE, local). - Prompt-Response Logging: Export full GenAI prompts and responses to GCS and BigQuery using Terraform-provisioned log sinks, external tables, and completions views. - BigQuery Agent Analytics & Third-Party Integrations: Enable the ADK BigQuery analytics plugin for conversational analytics and LLM-as-judge evals, or connect platforms like AgentOps, Phoenix, MLflow, and Weave. - Use Case: After deploying an ADK agent to Agent Runtime, use this Skill to verify traces appear in Cloud Trace, confirm prompt-response data lands in BigQuery, and add a user-feedback endpoint that routes ratings into the telemetry dataset. ## Quick Start Ask the assistant to set up tracing and prompt-response logging for your deployed agents-cli agent and verify the telemetry data is flowing.
