/**
 * Entity tagger — regex-based proper-noun / structured-token extractor.
 *
 * Used by the `hybridSearch` controller as a third RRF arm alongside the
 * dense (vector) and sparse (BM25/FTS5) signals (ADR-147, ruvnet/ruflo#2317).
 *
 * Entity matching is a distinct signal because BM25 weights documents by
 * overall token frequency; a per-entity exact match avoids downweighting
 * for tokens that happen to be common but mention an entity by name.
 * Example: querying "Alice OAuth tokens" — BM25 may rank a doc about
 * generic OAuth above one mentioning Alice specifically; entity arm
 * surfaces the Alice doc independently.
 *
 * P1 is regex-only — no NLP model dependency. Tags:
 *   - Emails: foo@bar.com
 *   - URLs: http(s)://...
 *   - File paths: ./foo/bar.ts, C:\foo\bar.ts, src/foo
 *   - Quoted phrases: "..." or '...'
 *   - Proper-noun 2-grams: "Alice Smith", "Acme Corp"
 *
 * Deliberately conservative — false negatives are fine (the dense + sparse
 * arms still cover the query); false positives would dilute the RRF score
 * by adding noise rows. A future P2 can swap this for a CRF/spaCy tagger
 * behind the same `extractEntities(text)` contract.
 */
/**
 * Extract entity-like tokens from free text. Returns a unique list,
 * trimmed and deduplicated, in extraction order.
 */
export declare function extractEntities(text: string): string[];
//# sourceMappingURL=entity-tagger.d.ts.map