/**
 * Composite 6-signal scorer for brain observation typed promotion.
 *
 * Replaces the legacy 3-rule OR union in runTierPromotion with a
 * weighted composite score that is explainable and auditable.
 *
 * Signals:
 * 1. citation_count   — retrieval frequency (normalised to 0.0–1.0 via tanh)
 * 2. quality_score    — content richness computed at insert (0.0–1.0)
 * 3. stability_score  — biological-analog consolidation metric (0.0–1.0)
 * 4. recency          — inverse age decay (1.0 = today, approaches 0 over 90 days)
 * 5. user_verified    — hard boost (0 or 1)
 * 6. outcome_correlated — whether tied to a completed task outcome (0 or 1)
 *
 * Formula:
 *   raw = w_citation × normalise(citation_count)
 *       + w_quality  × (quality_score ?? 0.5)
 *       + w_stability × (stability_score ?? 0.5)
 *       + w_recency  × recencyFactor(created_at)
 *       + w_verified × user_verified
 *       + w_outcome  × outcome_correlated
 *
 *   composite_score = clamp(raw / sum(weights), 0.0, 1.0)
 *
 * Weights are designed so that a verified OR high-citation entry always
 * crosses the default threshold (0.6), while pure noise (0 citations,
 * 0.3 quality, 0.5 stability, old age, unverified, no outcome) scores ~0.3.
 *
 * @task T1001
 * @epic T1000
 */
/** Input signals for the composite scorer. All fields are optional / nullable. */
export interface PromotionSignals {
    /** Raw citation count (0+). Normalised via tanh(count/5) to 0–1. */
    citationCount: number;
    /** Content quality score from insert-time scorer (0.0–1.0). Null → 0.5. */
    qualityScore: number | null;
    /** Biological-analog stability from brain_observations.stability_score (0.0–1.0). Null → 0.5. */
    stabilityScore: number | null;
    /** ISO 8601 / SQLite datetime string for age decay. Null → assume 30 days old. */
    createdAt: string | null;
    /** 1 if the entry was verified by the owner, 0 otherwise. */
    userVerified: number;
    /** 1 if the entry is correlated to a completed/verified task outcome, 0 otherwise. */
    outcomeCorrelated: number;
}
/**
 * Compute the composite promotion score for a brain entry.
 *
 * Returns a value in [0.0, 1.0]. Entries above the promotion threshold
 * (default 0.6) are eligible for typed promotion.
 *
 * @param signals - The 6 input signals for this entry
 * @returns Composite score in [0.0, 1.0]
 */
export declare function computePromotionScore(signals: PromotionSignals): number;
/**
 * Default promotion threshold. Entries with composite_score >= this value
 * are eligible for typed promotion via promoteObservationsToTyped.
 */
export declare const PROMOTION_THRESHOLD = 0.6;
/**
 * Determine the best target typed table for a promoted observation.
 *
 * Mapping from BRAIN_OBSERVATION_TYPES to BRAIN typed tables:
 * - 'decision' → promotes to brain_decisions (semantic)
 * - 'feature' | 'refactor' | 'change' → promotes to brain_patterns (procedural)
 * - 'discovery' | 'bugfix' | 'diary' → promotes to brain_learnings (episodic→semantic)
 *
 * All observation types map to one of the three typed tables.
 */
export declare function mapObservationTypeToTier(observationType: string): string;
/** The 6 signal names used in rationale JSON for audit transparency. */
export declare const SIGNAL_NAMES: readonly ["citation_count", "quality_score", "stability_score", "recency", "user_verified", "outcome_correlated"];
/** Detailed per-signal breakdown attached to each brain_promotion_log row. */
export interface PromotionRationale {
    signals: {
        citation_count: number;
        quality_score: number;
        stability_score: number;
        recency: number;
        user_verified: number;
        outcome_correlated: number;
    };
    weighted: {
        citation_count: number;
        quality_score: number;
        stability_score: number;
        recency: number;
        user_verified: number;
        outcome_correlated: number;
    };
    composite_score: number;
    threshold: number;
    decision: 'promote' | 'skip';
}
/**
 * Compute the full rationale breakdown for a promotion decision.
 *
 * The rationale is serialised to JSON and stored in brain_promotion_log.rationale_json
 * to make every promotion explainable (pairs with T997 promote-explain CLI).
 */
export declare function computePromotionRationale(signals: PromotionSignals, threshold?: number): PromotionRationale;
//# sourceMappingURL=promotion-score.d.ts.map