pub fn component_confidence(c: &AtlasComponent) -> f64Expand description
Confidence (0..1) of one component entry: how much of it is graph-derived evidence vs a bare-name heuristic.
A component named only (no purpose, no implementation, no edges) is the
heuristic floor (0.2) — the component compiler can still emit a name
from clustering alone. Every attribute that carries an EXTRACTED/
graph-resolved fact raises the score: a responsibility claim (purpose),
implementation paths + member symbols, consumes/produces and
upstream/downstream edges, ownership claims, failure behavior, and a
hierarchy layer assigned by the clusterer. The strongest components
(all evidence present) reach 1.0.