Before you press play
What you’ll find in the video
- The console supports supplied categories, yes-or-no criteria, and ordered rating scales over explicit state.
- The routing example scores incoming task complexity before choosing a smaller or larger generative model.
- The guardrail example evaluates a draft against a disclosure rule; passing these few examples does not establish protection from prompt leaks.
Worth knowing
Auto-generated English captions reviewed with Gemini. The speed, low cost, and guardrail decisions shown are subjective ad-hoc UI tests, not formal safety evaluations or controlled latency benchmarks.