Enum RecallScore 

#[non_exhaustive]
pub enum RecallScore {
    #[non_exhaustive]
    Hybrid {
        rrf_score: f64,
        text_contrib: f64,
        vector_contrib: f64,
        confidence: f32,
        effective_confidence: Option<f32>,
    },
    #[non_exhaustive]
    TextOnly {
        rank: usize,
        bm25_score: f32,
        confidence: f32,
        effective_confidence: Option<f32>,
    },
}

Per-channel signal breakdown for a recalled fact.

These are the input signals that the retrieval engine used to rank results — they explain what each channel contributed, not the final composite ranking score. The result ordering in recall_scored() is the authoritative ranking; inspect these fields to understand why a given channel dominated or was weak for a particular fact.

Every variant includes confidence — the fact-level confidence score from the underlying Fact (default 1.0). This lets callers weight or filter results by trustworthiness alongside retrieval signals.

The variant indicates which retrieval path produced the result:

• Hybrid — RRF fusion input signals (text + vector channels). The engine’s two-stage reranker uses these as inputs alongside temporal feasibility to determine final ordering.
• TextOnly — fulltext search with BM25 relevance score.

Variants (Non-exhaustive)

Non-exhaustive enums could have additional variants added in future. Therefore, when matching against variants of non-exhaustive enums, an extra wildcard arm must be added to account for any future variants.

#[non_exhaustive]

Hybrid

Input signals from hybrid retrieval (text + vector channels).

The rrf_score is the pre-rerank RRF fusion score (sum of text + vector contributions). Final result ordering may differ because the two-stage reranker applies adaptive weighting and temporal feasibility filtering on top of these signals.

Fields

Non-exhaustive enum variants could have additional fields added in future. Therefore, non-exhaustive enum variants cannot be constructed in external crates and cannot be matched against.

rrf_score: f64

Pre-rerank RRF fusion score (text + vector sum).

text_contrib: f64

Text-channel contribution from weighted RRF.

vector_contrib: f64

Vector-channel contribution from weighted RRF.

confidence: f32

Fact-level confidence [0.0, 1.0] from the stored fact.

effective_confidence: Option<f32>

Effective confidence after uncertainty model (age decay × source weight). None when uncertainty modeling is disabled.

#[non_exhaustive]

TextOnly

Result from fulltext-only retrieval.

Fields

Non-exhaustive enum variants could have additional fields added in future. Therefore, non-exhaustive enum variants cannot be constructed in external crates and cannot be matched against.

rank: usize

Ordinal rank in the result set (0-indexed).

bm25_score: f32

Tantivy BM25 relevance score. Higher = stronger lexical match. Comparable within a single query but not across queries.

confidence: f32

Fact-level confidence [0.0, 1.0] from the stored fact.

effective_confidence: Option<f32>

Effective confidence after uncertainty model (age decay × source weight). None when uncertainty modeling is disabled.

Implementations

impl RecallScore

pub fn display_tag(&self) -> String

Human-readable score tag suitable for debug output or LLM context.

Returns a decimal RRF score "0.032" for hybrid results, or a BM25 score with rank "#1 bm25:4.21" for text-only results.

pub fn confidence(&self) -> f32

The fact-level confidence score, regardless of retrieval path.

pub fn effective_confidence(&self) -> Option<f32>

The effective confidence after uncertainty model processing.

Returns None when uncertainty modeling is disabled. When Some, this reflects: base_confidence × age_decay × source_weight.

Trait Implementations

impl Clone for RecallScore

fn clone(&self) -> RecallScore

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for RecallScore

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

Formats the value using the given formatter.

impl PartialEq for RecallScore

fn eq(&self, other: &RecallScore) -> bool

Tests for self and other values to be equal, and is used by ==.

fn ne(&self, other: &Rhs) -> bool

Tests for !=. The default implementation is almost always sufficient, and should not be overridden without very good reason.

impl Copy for RecallScore

impl StructuralPartialEq for RecallScore