StaticDeriveDerive the plain-language trust verdict for a model from its metrics. Holdout metrics preferred. Classification grades on AUC (then Accuracy); regression on R². An unmeasured model is gated OFF.
the model's metric JSON + problem type.
the grade, plain copy, and the canAct gate (false for Poor or unmeasured models).
StaticTrustThe 0–5 "filled dots/stars" count for a grade, for compact visuals.
StaticTrustHuman-readable "checked against N members it had never seen" evidence line. Returns a generic phrasing when no count is available (never fabricates a number).
Optionalopts: TrustEvidenceOptions
Static entry points for the plain-language trust verdict and evidence helpers.