Struct AnswerRelevancy 

pub struct AnswerRelevancy<M: BaseChatModel, E: Embeddings> { /* private fields */ }

RAGAS answer relevancy: does the answer actually address the question?

The generator produces n questions that the answer could address (one LLM call); the score is the mean cosine similarity between the original question’s embedding and the generated questions’ embeddings. An answer full of on-topic-looking but non-answer text generates off-target questions and scores low, without needing a reference answer.

Uses one chat call (plain-text question list) plus one embedding batch per evaluation.

Implementations

impl<M: BaseChatModel, E: Embeddings> AnswerRelevancy<M, E>

pub fn new(generator: M, embeddings: E) -> Self

Creates an answer-relevancy evaluator.

pub fn with_n_questions(self, n: usize) -> Self

Sets how many questions the generator should produce (clamped to at least 1).

pub fn with_empty_score(self, score: f64) -> Self

Score for an empty prediction (default 0.0: no answer is not relevant).

Trait Implementations

impl<M: BaseChatModel, E: Embeddings> Evaluator for AnswerRelevancy<M, E>

fn eval<'life0, 'life1, 'life2, 'life3, 'async_trait>( &'life0 self, input: &'life1 str, prediction: &'life2 str, _reference: &'life3 str, ) -> Pin<Box<dyn Future<Output = Result<Score, EvalError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait, 'life3: 'async_trait,

Scores a single prediction

fn name(&self) -> &str

Evaluator name (used in report summaries)

impl<M: BaseChatModel, E: Embeddings> RagEvaluator for AnswerRelevancy<M, E>

fn eval_rag<'life0, 'life1, 'life2, 'life3, 'life4, 'async_trait>( &'life0 self, input: &'life1 str, prediction: &'life2 str, _contexts: &'life3 [String], _reference: &'life4 str, ) -> Pin<Box<dyn Future<Output = Result<Score, EvalError>> + Send + 'async_trait>>

where Self: 'async_trait, 'life0: 'async_trait, 'life1: 'async_trait, 'life2: 'async_trait, 'life3: 'async_trait, 'life4: 'async_trait,

Scores a single prediction given the question, retrieved contexts, and reference answer.

fn name(&self) -> &str

Evaluator name (used in report summaries)