JSON & evaluation

1 JSON model evaluates content and returns structured answers.

Model	Provider	Purpose
typesafe-evaluate — Jev	TypeSafe	Evaluate named yes/no, choice, and score questions

Execution

Use the SDK's ai.run(Models.TypesafeEvaluate, inputs) method. JSON models are rejected by generate(), generateText(), and the SDK's async lifecycle methods. The generic execution result contains { result, status, usage, id }; it does not extract media URLs or automatically save to Drive.

See the TypeSafe reference for a complete example and required inputs. CLI, MCP, and Playground support must be checked separately.