evaluation

type AsyncMetric interface {
	Metric
	// ScoreAsync evaluates the metric asynchronously.
	ScoreAsync(ctx context.Context, input MetricInput) <-chan *ScoreResult
}

type BaseMetric struct {
	// contains filtered or unexported fields
}

type BatchMetric interface {
	Metric
	// ScoreBatch evaluates multiple inputs and returns results for each.
	ScoreBatch(ctx context.Context, inputs []MetricInput) ScoreResults
}

type CompositeMetric struct {
	BaseMetric
	// contains filtered or unexported fields
}

type ConditionalMetric struct {
	BaseMetric
	// contains filtered or unexported fields
}

type DatasetEvaluator struct {
	// contains filtered or unexported fields
}

type Engine struct {
	// contains filtered or unexported fields
}

type EngineOption func(*Engine)

type EvaluationCallback func(completed, total int, result *EvaluationResult)

type EvaluationResult struct {
	// ItemID is the identifier for the evaluated item.
	ItemID string
	// Input is the input that was evaluated.
	Input MetricInput
	// Scores contains all metric scores for this item.
	Scores ScoreResults
	// Error is set if evaluation failed entirely.
	Error error
}

type EvaluationResults []*EvaluationResult

type Metric interface {
	// Name returns the name of the metric.
	Name() string
	// Score evaluates the metric and returns a score result.
	Score(ctx context.Context, input MetricInput) *ScoreResult
}

type MetricFunc struct {
	BaseMetric
	// contains filtered or unexported fields
}

type MetricInput struct {
	// Input is the input to the LLM/model.
	Input string
	// Output is the output from the LLM/model.
	Output string
	// Expected is the expected/reference output (for comparison metrics).
	Expected string
	// Context is additional context provided to the model.
	Context string
	// Metadata contains additional key-value pairs.
	Metadata map[string]any
}

type ScoreResult struct {
	// Name is the name of the metric.
	Name string `json:"name"`
	// Value is the numeric score value (typically 0.0 to 1.0).
	Value float64 `json:"value"`
	// Reason is an optional explanation for the score.
	Reason string `json:"reason,omitempty"`
	// Metadata contains additional information about the score.
	Metadata map[string]any `json:"metadata,omitempty"`
	// Error is set if the metric evaluation failed.
	Error error `json:"error,omitempty"`
}

type ScoreResults []*ScoreResult

type WeightedMetric struct {
	// contains filtered or unexported fields
}