Struct MemoryRecallPipeline 

pub struct MemoryRecallPipeline { /* private fields */ }

Composable scoring pipeline for memory recall candidates.

Wraps a WeightedObjective<NoteCandidate> with the three standard memory scoring components (RRF relevance, amplified salience, temporal recency) weighted by the recall config parameters. Pack code uses this type to avoid a direct dependency on khive-fold.

Implementations

impl MemoryRecallPipeline

pub fn new( relevance_weight: f64, salience_weight: f64, temporal_weight: f64, half_life_days: f64, salience_alpha: f64, ) -> Self

Build a pipeline from explicit component weights and temporal half-life.

relevance_weight, salience_weight, temporal_weight correspond to RecallConfig’s three weight fields. half_life_days drives TemporalRecencyObjective. salience_alpha is the amplification exponent for AmplifiedDecayAwareSalienceObjective (default 1.5).

pub fn default_memory() -> Self

Build a pipeline using the standard memory recall defaults.

Weights: relevance=0.70, salience=0.20, temporal=0.10; half_life=30 days; alpha=1.5.

pub fn score(&self, candidate: &NoteCandidate) -> f64

Score a NoteCandidate through the pipeline.

The result is in [0.0, 1.0]. The NoteCandidate.rrf_score field should carry the pre-normalized relevance (output of normalize_relevance / RrfFusionObjective).