Enum RerankFeature 

pub enum RerankFeature {
    EmbeddingScore,
    KeywordOverlap,
    LengthPenalty,
    TitleBoost {
        boost: f64,
    },
    PositionPrior {
        decay: f64,
    },
}

Reranking feature variants.

Variants

EmbeddingScore

Cosine similarity between query.embedding and candidate.embedding.

KeywordOverlap

Jaccard similarity between tokenised query and content term sets.

LengthPenalty

Penalty for very short (<50 chars) or very long (>5000 chars) content. score = 1.0 - |optimal - len| / optimal, optimal = 500.

TitleBoost

Multiply score by boost when the title metadata contains a query term.

Fields

boost: f64

Multiplicative factor applied when the title matches (e.g., 1.5).

PositionPrior

Weighted initial score decay based on candidate rank in initial list. score = initial_score * (1 - decay * rank_fraction)

Fields

decay: f64

Decay factor controlling how quickly position degrades score.

Implementations

impl RerankFeature

pub fn name(&self) -> &'static str

Human-readable feature name used as key in RerankResult::feature_scores.

Trait Implementations

impl Clone for RerankFeature

fn clone(&self) -> RerankFeature

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for RerankFeature

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.