Create Evaluation
Create an evaluation (a bundle of metrics + judge config).
Headers
Organization ID for user PATs (pat_ prefix tokens)
Body
Schema for API requests to create a new evaluation (metric bundle).
No input/judge_config: the judge is always Gemini Pro and the content
to resolve is derived from the bundle's metrics' modalities.
Response
Successful Response
Evaluation entity — a bundle of metrics (+ optional shared framing).
Renamed from Scorecard: "scorecard" connotes a score (a rollup), which is deferred to the derived layer (PRD §13.1). Many evaluations may run on one conversation (compliance vs quality vs sales bundles). The judge is always Gemini Pro and the content needed is derived from the metrics' modalities — so there's no per-evaluation input/judge_config.

