Package com.tnsai.evaluation.evaluators.agent

Class TaskCompletionEvaluator

java.lang.Object

com.tnsai.evaluation.evaluators.AbstractJudgeEvaluator

com.tnsai.evaluation.evaluators.agent.TaskCompletionEvaluator

All Implemented Interfaces:

Evaluator

public final class TaskCompletionEvaluator
extends AbstractJudgeEvaluator

Agent-tier metric: did the agent's final state actually solve the task?

The judge looks at the user's request and the declared goal (whichever of expectedOutput or instructions is populated — at least one must be) and scores how completely the agent's final response satisfied it on a 1–5 scale.

Distinct from PlanAdherenceEvaluator: adherence asks "did you follow the plan?", completion asks "did you reach the destination?" An agent can score perfect adherence on a wrong plan and still fail completion; conversely it can ignore the plan and still nail the goal.

Distinct from G-Eval: G-Eval is criteria-driven (the caller writes the criteria description prompt). TaskCompletion has a fixed contract: goal-vs-response. That makes it cheaper to plug in uniformly across an evaluation suite.

Since:

0.10.1 (TNS-373)

Nested Class Summary

Nested classes/interfaces inherited from interface com.tnsai.evaluation.spi.Evaluator

Evaluator.EvaluationInput

Constructor Summary

Constructors

Constructor	Description
TaskCompletionEvaluator(LLMJudge judge)

Method Summary

Methods inherited from class com.tnsai.evaluation.evaluators.AbstractJudgeEvaluator

evaluate, name

Methods inherited from class java.lang.Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Details

TaskCompletionEvaluator

public TaskCompletionEvaluator(LLMJudge judge)