Google's Agent Development Kit (ADK) now supports native live evaluation for voice agents. Developers can use a simulated audio user to test spoken conversations, score replies with natural-language rubrics, and integrate evaluations into CI/CD pipelines via AgentEvaluator or run them interactively in ADK Web with transcript and audio playback. Test cases support both improvised conversation scenarios with configurable personas and fixed verbatim conversations, with max_allowed_invocations setting a hard turn limit on dynamic cases.
Read the full story on developers.googleblog.com
What to doInstall the ADK eval extras with uv pip install -e ".[eval]" and configure API credentials for both the Live API and Gemini TTS to enable live voice agent evaluation.
- Provider