Class SessionEvaluator

Index

Constructors

Methods

compareSessions evaluate

Constructors

constructor

new SessionEvaluator(config?: Partial<SessionEvalConfig>): SessionEvaluator
Parameters
- Optionalconfig: Partial<SessionEvalConfig>
Returns SessionEvaluator
- Defined in src/agents/sessionEval.ts:218

Methods

compareSessions

compareSessions(
    sessionA: SessionEvalResult,
    sessionB: SessionEvalResult,
): {
    dimensionDeltas: Record<string, number>;
    improvements: string[];
    overallDelta: number;
    regressions: string[];
    winner: "A" | "tie" | "B";
}
Compare two sessions (e.g. A/B test of agent versions)
Parameters
- sessionA: SessionEvalResult
- sessionB: SessionEvalResult
Returns {
    dimensionDeltas: Record<string, number>;
    improvements: string[];
    overallDelta: number;
    regressions: string[];
    winner: "A" | "tie" | "B";
}
- Defined in src/agents/sessionEval.ts:531

evaluate

evaluate(
    turns: SessionTurn[],
    goals?: SessionGoal[],
    metadata?: Record<string, unknown>,
): SessionEvalResult
Evaluate a complete session
Parameters
- turns: SessionTurn[]
- goals: SessionGoal[] = []
- Optionalmetadata: Record<string, unknown>
Returns SessionEvalResult
- Defined in src/agents/sessionEval.ts:227