Struct LlmJudge 

pub struct LlmJudge { /* private fields */ }

Available on crate feature eval only.

LLM-based judge for semantic evaluation

Implementations

impl LlmJudge

pub fn new(model: Arc<dyn Llm>) -> LlmJudge

Create a new LLM judge with the given model

pub fn with_config(model: Arc<dyn Llm>, config: LlmJudgeConfig) -> LlmJudge

Create with custom config

pub async fn semantic_match( &self, expected: &str, actual: &str, config: Option<&SemanticMatchConfig>, ) -> Result<SemanticMatchResult, EvalError>

Judge semantic similarity between expected and actual responses

Returns a score from 0.0 to 1.0 indicating semantic equivalence.

pub async fn evaluate_rubrics( &self, response: &str, context: &str, config: &RubricConfig, ) -> Result<RubricEvaluationResult, EvalError>

Evaluate response against rubrics

Returns scores for each rubric and an overall weighted score.

pub async fn evaluate_safety( &self, response: &str, ) -> Result<SafetyResult, EvalError>

Evaluate safety of a response

pub async fn detect_hallucinations( &self, response: &str, context: &str, ground_truth: Option<&str>, ) -> Result<HallucinationResult, EvalError>

Detect potential hallucinations in a response