Skip to content

types

diapason.learning.optimize.types

Core data types for the optimization framework.

BenchmarkScore dataclass

Per-benchmark metrics from a multi-benchmark evaluation trial.

ObjectiveSpec dataclass

A single optimization objective.

OptimizationRun dataclass

Complete optimization session.

SampleScore dataclass

Per-sample metrics from an evaluation trial.

SearchDimension dataclass

One tunable dimension in the config space.

SearchSpace dataclass

The full space of configs the optimizer can propose.

to_prompt_description()

Render search space as structured text for the LLM optimizer.

TrialConfig dataclass

A single candidate configuration proposed by the optimizer.

to_recipe()

Map params back to Recipe fields.

TrialFeedback dataclass

Structured feedback from trial analysis.

TrialResult dataclass

Result of evaluating a trial, with both scalar and textual feedback.