pub struct FeedbackMetrics {
pub entries_applied: usize,
pub suppressed: usize,
pub preserved_useful_siblings: usize,
pub boosted: usize,
pub recurring_fp_suppressed: usize,
pub constraint_penalties_applied: usize,
pub task_bias_boosted: usize,
pub rules_applied: Vec<String>,
}Expand description
Apply feedback verdicts as score adjustments after candidate selection.
Metrics returned by apply_feedback_adjustments, describing how feedback
changed the ranked result set for this query.
Fields§
§entries_applied: usizeFeedback entries whose stored query matched the current query family.
suppressed: usizeTotal symbols removed from results (all causes combined).
preserved_useful_siblings: usizeSymbols in a noisy file preserved because they had an explicit Useful verdict.
boosted: usizeUseful verdicts that boosted a symbol’s score.
recurring_fp_suppressed: usizeSymbols suppressed by the recurring false-positive rule (≥2 noisy verdicts).
constraint_penalties_applied: usizePlan C t-003: symbols suppressed by an active Constraint/Decision
ledger entry carrying a penalty role (stale-api / audit-pending).
Independent of suppressed so callers can tell ranking-time
decisions apart from feedback-time verdicts.
task_bias_boosted: usizePlan C t-006: candidates boosted by +1.0 because they fall inside the active CTX task’s recorded scope. Soft bias only.
rules_applied: Vec<String>Unique rule names that fired, in first-encounter order. Values: “exact_noisy”, “same_file_sibling_suppression”, “useful_boost”, “recurring_fp”, “constraint_penalty” (Plan C t-003), “task_bias” (Plan C t-006), “already_covered” / “diagnostic_only” (Plan C t-005).