Struct BulletSelector 

pub struct BulletSelector {
    pub weights: HashMap<String, f64>,
    pub embedding_model: String,
    pub cache_file: Option<String>,
    pub embedding_cache: EmbeddingCache,
}

Selects most relevant bullets for a given query.

Implements hybrid retrieval with three scoring factors:

• Effectiveness: Based on helpful/harmful feedback
• Recency: Prefers recently updated bullets
• Semantic: Query-to-bullet similarity using embeddings

Fields

weights: HashMap<String, f64>

embedding_model: String

cache_file: Option<String>

embedding_cache: EmbeddingCache

Implementations

impl BulletSelector

pub fn new( weights: Option<HashMap<String, f64>>, embedding_model: &str, cache_file: Option<&str>, ) -> Self

Create a new bullet selector.

pub fn select( &self, bullets: &[Bullet], max_count: usize, query: Option<&str>, ) -> Vec<Bullet>

Select top-K most relevant bullets.

pub fn score_bullet(&self, bullet: &Bullet, query: Option<&str>) -> ScoredBullet

Score a single bullet.

pub fn selection_stats( &self, all_bullets: &[Bullet], selected: &[Bullet], ) -> HashMap<String, f64>

Get statistics about a selection.

Trait Implementations

impl Default for BulletSelector

fn default() -> Self

Returns the “default value” for a type.