Built-in Judges

Every built-in judge returns the normalized Judgment model. Status is required, while a stored score and label are present only when that judge completed a measurement or classification.

Core file-system judges

Artifact: agent-judge-core.

Judge	Purpose
`FileExistsJudge`	Check that a workspace-relative path exists
`FileContentJudge`	Compare file content with exact, contains, or regular-expression matching
`SupersetDiffJudge`	Require output to contain a byte-exact reference project plus any allowed additions

These judges produce Boolean outcomes and therefore normally store no score. The executable starting point is tutorial module 01.

Command and coverage judges

Artifact: agent-judge-exec.

Judge	Purpose
`BuildSuccessJudge`	Run Maven, Gradle, or a configured build command and judge its exit status
`CommandJudge`	Execute an arbitrary command through Agent Sandbox and compare the exit code
`ClassVersionJudge`	Validate Java class-file bytecode versions
`CoveragePreservationJudge`	Reject a JaCoCo coverage regression beyond the configured tolerance
`CoverageImprovementJudge`	Measure normalized coverage improvement with an optional floor

Build and command judges run real processes in the context workspace. Use the project wrapper when present and choose timeouts appropriate to untrusted work. Tutorial module 02 executes the Maven build path.

Semantic file judges

Artifact: agent-judge-file.

Judge	Purpose
`FileComparisonJudge`	Dispatch files to the appropriate semantic comparator
`JavaSemanticJudge`	Compare Java syntax trees rather than formatting
`MavenSemanticJudge`	Compare important Maven model structure
`XmlSemanticJudge`	Compare XML document structure
`TextFileJudge`	Compare whitespace-normalized text

These judges expect reference and actual file locations in the evaluation context as documented in the aggregate Javadoc.

LLM judge

Artifact: agent-judge-llm.

Judge	Purpose
`CorrectnessJudge`	Ask a Spring AI-backed model whether output satisfies the goal
`LLMJudge`	Base template for specialized Spring AI LLM judges

For a framework-neutral composition, use ModelBackedJudge from agent-judge-ai-core with a JudgePromptTemplate, JudgeModel, and JudgmentClassifier. Tutorial module 08 runs that complete path with no credentials.

RAG judges

Artifact: agent-judge-rag.

Judge	Question answered
`FaithfulnessJudge`	Is the generated answer supported by the retrieved context?
`ContextualRelevanceJudge`	Is the retrieved context relevant to the question?
`HallucinationJudge`	Does the answer introduce unsupported claims?

RagContext defines the rag.question, rag.context, and rag.answer metadata keys and the framework fallback conventions. When required evidence is absent, these judges abstain rather than inventing a score.

Framework evaluators

Evaluators are bridges rather than judges. They execute or adapt framework output into JudgmentContext, then apply an ordinary Judge or Jury.

Artifact	Evaluator	Evaluated value
`agent-judge-spring-ai`	`SpringAiEvaluator`	Spring AI `ChatResponse`
`agent-judge-langchain4j`	`LangChain4jEvaluator`	LangChain4j `Result<T>`
`agent-judge-koog`	`KoogEvaluator`	Koog `AIAgent` output
`agent-judge-agent-client`	`AgentClientEvaluator`	AgentClient responses and workspaces

The credential-free bridge samples are Koog module 09 and LangChain4j module 10.

Projects

AgentWorks

Supporting Projects

Migration

Core file-system judges

Command and coverage judges

Semantic file judges

LLM judge

RAG judges

Framework evaluators

​Core file-system judges

​Command and coverage judges

​Semantic file judges

​LLM judge

​RAG judges

​Framework evaluators

​Related

Core file-system judges

Command and coverage judges

Semantic file judges

LLM judge

RAG judges

Framework evaluators

Related