Module dynamic 

Dynamic-question evaluation (spec 019, 3.7): offline reproduction of a retained outcome with zero backend calls, and a dedicated dataset, evaluator configuration and report for one dynamic task family.

Nothing here calls a backend or reads a clock: there is no backend parameter anywhere in this module. A changed request, or a retained outcome without a raw response, is incomparable, never permission to ask again. Dynamic evidence never becomes a registered-plan replay bundle, and evidence for one dynamic task family, question schema, label order, template or artifact does not transfer to another.

Reports state coverage and abstention first, then accepted error and class-appropriate metrics, each with its denominator; an empty denominator is unknown, never a pass. Distributions stay uncalibrated: this module fits no calibration and reports none. A report on synthetic data measures mechanics only, and every gate over it is unknown.

Structs

DynamicCase

DynamicDataset

A dynamic dataset, rustev.dynamic-dataset/1 (3.7).

DynamicEvalReport

A dynamic evaluation report, rustev.dynamic-eval-report/1 (3.7).

DynamicEvaluatorConfig

A dynamic evaluator configuration, rustev.dynamic-evaluator-config/1 (3.7): what a report’s numbers mean anything under.

DynamicGate

GateResult

IncomparableCase

QuestionSchema

The exact question and label schema a dataset and a configuration bind: results never transfer to another question, label order or output kind.

Enums

CalibrationStatus

DynamicDatasetError

Why a dataset was refused.

DynamicEvalError

Why a report could not be produced.

DynamicIncomparable

Why a retained outcome cannot be compared (3.7).

DynamicReproduction

The result of reproducing one retained outcome.

GateDirection

GateStatus

QualityClaim

Constants

DYNAMIC_DATASET

DYNAMIC_EVALUATOR_CONFIG

DYNAMIC_REPORT

Functions

evaluate

Evaluate retained outcomes of one split against a dynamic dataset (3.7). outcomes maps case ids to retained outcomes; a missing outcome or one that does not reproduce is listed as incomparable and left out of every denominator. Nothing is dispatched.

reproduce

Reproduce a retained outcome from the request, the capability and the retained raw response, with zero backend calls (3.7).