Struct SemanticFeedbackLoop 

pub struct SemanticFeedbackLoop {
    pub feedback: HashMap<u64, QueryFeedback>,
    pub boosts: HashMap<u64, BoostRecord>,
    pub stats: FeedbackStats,
    /* private fields */
}

Core feedback-loop engine.

Ingests relevance signals emitted during search sessions and uses them to boost (or suppress) future rankings of individual results.

Fields

feedback: HashMap<u64, QueryFeedback>

Per-query feedback records, keyed by query_id.

boosts: HashMap<u64, BoostRecord>

Per-result cumulative boost records, keyed by result_id.

stats: FeedbackStats

Global statistics.

Implementations

impl SemanticFeedbackLoop

pub fn new() -> Self

Create a new, empty feedback loop with the default capacity of 50 000 entries.

pub fn with_max_entries(max_entries: usize) -> Self

Create a new, empty feedback loop with the given maximum entry capacity.

pub fn record_feedback( &mut self, query_id: u64, signal: FeedbackSignal, now_secs: u64, )

Record a single relevance signal for the given query.

• Updates the per-query QueryFeedback record.
• Updates the per-result BoostRecord.
• Updates global FeedbackStats.

pub fn apply_boosts(&self, results: &[(u64, f64)]) -> Vec<(u64, f64)>

Re-score and re-rank a list of (result_id, score) pairs using boosts.

For each entry: new_score = score * (1.0 + effective_boost().max(-0.9))

The result list is returned sorted by new_score descending.

pub fn top_boosted_ids(&self, k: usize) -> Vec<u64>

Return the top-k result IDs sorted by effective_boost descending.

pub fn positive_ids_for_query(&self, query_id: u64) -> Vec<u64>

Return the positive result_ids recorded for a specific query.

Returns an empty Vec if the query has not been seen.

pub fn stats(&self) -> &FeedbackStats

Reference to the global statistics.

pub fn record( &mut self, query_id: &str, doc_id: &str, feedback: FeedbackType, confidence: f64, )

Record a user feedback entry. If the loop is at capacity the oldest entry is evicted before inserting the new one.

pub fn get_summary(&self, query_id: &str) -> Option<QueryFeedbackSummary>

Aggregate feedback for a specific query. Returns None if the query has no recorded entries.

pub fn relevant_docs(&self, query_id: &str) -> Vec<String>

Return document IDs marked Relevant for the given query.

pub fn irrelevant_docs(&self, query_id: &str) -> Vec<String>

Return document IDs marked Irrelevant for the given query.

pub fn precision_at_query(&self, query_id: &str) -> Option<f64>

Precision for a single query: relevant / (relevant + irrelevant). Returns None when the query has no relevant or irrelevant entries.

pub fn overall_precision(&self) -> Option<f64>

Overall precision across all queries: total_relevant / (total_relevant + total_irrelevant). Returns None when there are no relevant or irrelevant entries at all.

pub fn feedback_count(&self) -> usize

Total number of user feedback entries currently stored.

pub fn queries_with_feedback(&self) -> Vec<String>

Unique query IDs that have at least one feedback entry.

pub fn tick(&mut self)

Advance the internal tick counter by one.

pub fn clear_entries(&mut self)

Remove all user feedback entries and reset the tick counter.

pub fn loop_stats(&self) -> FeedbackLoopStats

Compute aggregate FeedbackLoopStats from the current entries.

Trait Implementations

impl Debug for SemanticFeedbackLoop

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for SemanticFeedbackLoop

fn default() -> Self

Returns the “default value” for a type.