ScoreModelGrader class

Uses an AI model to assign numeric scores.

The model evaluates inputs and returns a score, which is compared against the passThreshold.

Example

final grader = EvalGrader.scoreModel(
  name: 'quality_score',
  model: 'gpt-4o-mini',
  input: [
    ScoreModelInput.system('Rate the quality of this response from 0-10.'),
    ScoreModelInput.user('Response: {{sample.output_text}}'),
  ],
  passThreshold: 7.0,
);
Inheritance
Annotations
  • @immutable

Constructors

ScoreModelGrader({required String name, required String model, required List<ScoreModelInput> input, double? passThreshold, String? samplingParams})
Creates a ScoreModelGrader.
const
ScoreModelGrader.fromJson(Map<String, dynamic> json)
Creates a ScoreModelGrader from JSON.
factory

Properties

hashCode → int
The hash code for this object.
no setteroverride
input → List<ScoreModelInput>
The input messages for the model.
final
model → String
The model to use for scoring.
final
name → String
The name of this grader.
final
passThreshold → double?
The minimum score required to pass.
final
runtimeType → Type
A representation of the runtime type of the object.
no setterinherited
samplingParams → String?
Optional sampling parameters as JSON string.
final
type → String
The type of grader.
no setteroverride

Methods

noSuchMethod(Invocation invocation) → dynamic
Invoked when a nonexistent method or property is accessed.
inherited
toJson() → Map<String, dynamic>
Converts this to JSON.
override
toString() → String
A string representation of this object.
override

Operators

operator ==(Object other) → bool
The equality operator.
override