evaluation

Reusable evaluation and verification capabilities for FlossWare.

Evaluation answers: Did the result satisfy the intended outcome, and how strong is the supporting evidence?

Boundary

Intent + result + evidence
          |
          v
       evaluation
          |
          +-- deterministic tests and rules
          +-- model-based judges
          +-- adversarial verification
          +-- human feedback
          +-- external evaluators
          |
          v
EvaluationResult

Evaluation is independent of model invocation. It may use model-gateway when a model-based judge is appropriate, but it does not own provider selection or model routing.

Relationship to Loom

Loom defines execution and evidence semantics. evaluation supplies reusable evaluation behavior and result evidence that Loom Workers, Arbiters, or other compatible callers can consume. It does not replace the Arbiter or become an orchestration layer.

Evaluation results should distinguish measured facts, evaluator judgments, confidence, and provenance. A successful transport call is not proof that the intended outcome was achieved.

Packaging note

The repository is named evaluation. The current distribution/import identifiers remain evaluation-ai and evaluation_ai; repository naming does not automatically rename a published package or Python imports. Change those identifiers only through an explicit compatibility and release decision.