google-agents-cli-eval

You evaluate ADK agent performance by configuring evalsets and custom metrics to measure tool call accuracy, response quality, and safety compliance. This process allows development teams to iterate until achieving desired thresholds before deployment.

Evalsets are customizable schemas and criteria used for thorough ADK agent testing. They enable automated evaluation loops with detailed scoring and analysis to measure an agent's tool usage accuracy and response quality.

You set up custom metrics for automated agent evaluation by configuring specific criteria within evalsets. This guides the evaluation loop to generate detailed scores analyzing tool usage and response accuracy.

Yes, the agents-cli is a required dependency to run ADK evaluation tests. It provides the necessary environment to execute automated evals and analyze resulting scores to improve agent performance.

Yes, you can measure safety compliance and response accuracy for ADK agents. The evaluation process uses customizable schemas and metrics to assess these aspects alongside tool call accuracy during testing.