Struct Ranker 

pub struct Ranker<'a> { /* private fields */ }

Implementations

impl<'a> Ranker<'a>

pub fn new(engine: &'a Engine<'a>) -> Self

pub fn pagerank_global(&self) -> Result<Vec<(String, f64)>>

Raw global PageRank vector: (node_id, score) over the full heterogeneous universe, id-sorted. No projection, no blending.

pub fn pagerank_global_with( &self, contributors: &[EdgeWeightFn], ) -> Result<Vec<(String, f64)>>

Global vector with edge-weight contributors applied.

pub fn pagerank_task(&self, goal: &str) -> Result<Vec<(String, f64)>>

Raw task-personalized PPR vector for goal.

pub fn pagerank_task_with( &self, goal: &str, contributors: &[EdgeWeightFn], ) -> Result<Vec<(String, f64)>>

Task vector with edge-weight contributors applied.

pub fn candidates(&self, goal: &str, limit: usize) -> Result<Vec<ScoredEntity>>

Lexical candidate generation for goal (stage 1).

pub fn candidates_with( &self, goal: &str, limit: usize, hooks: &RankHooks, ) -> Result<Vec<ScoredEntity>>

Candidates with explicit plugin providers (one call, deterministic order). Provider rows merge by canonical id: max score wins, the provider reason is tagged plugin:<id> for explainability.

pub fn symbols(&self, req: &RankRequest) -> Result<RankResult>

Full task/global blend per symbol with feature decomposition. Same math as build_surface (no MMR/quotas/budget — pure ranking). Plugin seed/feature/rerank hooks apply here and are recorded.

pub fn symbols_with_hooks( &self, req: &RankRequest, hooks: &RankHooks, ) -> Result<RankResult>

symbols() with explicit plugin hooks (one call, deterministic order).

Derivation shares the pipeline’s inputs BY CONSTRUCTION: per-entry confidence, required set, seed membership, and lexical scores come from the compiled surface map and its helpers — the same values build_surface blends. Per-symbol totals take the max over that symbol’s entries (overloads); blends are otherwise identical math.