Struct VectorSearchRanker 

pub struct VectorSearchRanker {
    pub config: RankerConfig,
}

Re-ranks raw k-NN candidates using a configurable multi-signal scoring model.

Fields

config: RankerConfig

Configuration including signals and the reference timestamp.

Implementations

impl VectorSearchRanker

pub fn new(config: RankerConfig) -> Self

Creates a new VectorSearchRanker with the given configuration.

pub fn score_candidate(&self, candidate: &RawCandidate) -> RankedResult

Scores a single candidate against all configured signals.

Returns a RankedResult with the computed final_score and per-signal breakdown for explainability.

pub fn rank(&self, candidates: &[RawCandidate]) -> Vec<RankedResult>

Scores all candidates and returns them sorted by final_score descending.

pub fn rank_top_k( &self, candidates: &[RawCandidate], k: usize, ) -> Vec<RankedResult>

Returns at most k top-ranked candidates.

pub fn explain(&self, result: &RankedResult) -> String

Returns a human-readable explanation of a RankedResult.

Format: "id=X score=Y.YY [signal=A, ...]"