output-dev-evaluator-function

Evaluators for Output SDK workflows are created by adding functions to evaluators.ts, which centralize quality checks, content validation, and confidence assignment for workflow outputs.

Deterministic activity isolation in workflow validation ensures evaluator functions execute predictably without side effects, enforcing strict import rules and consistent result types across Output SDK projects.

Yes, workflow evaluators support both flat single-file setups and large folder-based structures, enabling modular reuse of evaluation components while maintaining strict import rules and compatibility.

Yes, the Output SDK supports optional LLM-powered evaluators alongside deterministic ones, allowing you to enforce strict validation rules while leveraging AI for content assessment within evaluators.ts.

Centralizing evaluation logic in evaluators.ts provides a structured way to assess quality, validate outputs, and assign confidence to decisions, ensuring consistent result types and modular reuse across projects.

Deterministic evaluators in Output workflows require strict import rules and deterministic activity isolation, meaning they cannot rely on non-deterministic operations unless explicitly using the optional LLM-powered evaluator support.