/**
 * MAMA (Memory-Augmented MCP Architecture) - Relevance Scorer
 *
 * Relevance scoring formula for decision ranking and top-N selection
 * Tasks: 1.1-1.4, 2.1-2.7 (Relevance scoring and top-N selection)
 * AC #1, #4, #5: Decision relevance, failure priority boost, top-N selection
 *
 * @module relevance-scorer
 * @version 1.0
 * @date 2025-11-14
 */
/**
 * Decision object for relevance scoring
 */
export interface DecisionWithEmbedding {
    id?: string;
    topic?: string;
    decision: string;
    reasoning?: string | null;
    outcome?: string | null;
    failure_reason?: string | null;
    user_involvement?: string;
    confidence?: number;
    created_at: number;
    updated_at?: number;
    embedding?: Float32Array;
    relevanceScore?: number;
}
/**
 * Query context for relevance calculation
 */
export interface QueryContext {
    embedding?: Float32Array;
}
/**
 * Formatted context result
 */
export interface FormattedContext {
    full: Array<{
        decision_id?: string;
        topic?: string;
        decision: string;
        reasoning?: string | null;
        outcome?: string | null;
        failure_reason?: string | null;
        user_involvement?: string;
        confidence?: number;
        relevanceScore?: number;
        created_at: number;
    }>;
    summary: {
        count: number;
        duration_days: number;
        failures: Array<{
            decision: string;
            reason: string | null | undefined;
        }>;
    } | null;
}
/**
 * Calculate relevance score for a single decision
 *
 * Task 1.2: Implement calculateRelevance(decision, queryContext) function
 * AC #1, #4: Relevance scoring with failure priority boost
 *
 * Formula:
 *   Relevance = (Recency × 0.2) + (Importance × 0.5) + (Semantic × 0.3)
 *
 * Where:
 *   - Recency: exp(-days_since / 30)  [30-day half-life]
 *   - Importance: OUTCOME_WEIGHTS[outcome]
 *     - FAILED: 1.0 (highest - failures are most valuable)
 *     - PARTIAL: 0.7
 *     - SUCCESS: 0.5
 *     - null: 0.3 (ongoing, lowest)
 *   - Semantic: cosineSimilarity(decision.embedding, query.embedding)
 *
 * @param decision - Decision object
 * @param queryContext - Query context
 * @returns Relevance score (0.0-1.0)
 */
export declare function calculateRelevance(decision: DecisionWithEmbedding, queryContext: QueryContext): number;
/**
 * Select top N most relevant decisions
 *
 * Task 2.1: Add selectTopDecisions(decisions, queryContext, n=3) function
 * AC #1, #5: Top-N selection with threshold filtering
 *
 * @param decisions - Array of decision objects
 * @param queryContext - Query context with embedding
 * @param n - Number of top decisions to return (default: 3)
 * @returns Top N decisions with relevance scores
 */
export declare function selectTopDecisions(decisions: DecisionWithEmbedding[], queryContext: QueryContext, n?: number): DecisionWithEmbedding[];
/**
 * Format decisions with top-N selection and summary
 *
 * Task 8.2-8.3: Format top 3 in full detail, rest as summary
 * AC #5: Top-N selection with summary
 *
 * @param decisions - All decisions (sorted by relevance)
 * @param topN - Number of decisions to show in full detail (default: 3)
 * @returns Formatted context {full: Array, summary: Object}
 */
export declare function formatTopNContext(decisions: DecisionWithEmbedding[], topN?: number): FormattedContext;
/**
 * Test result for relevance scoring
 */
export interface TestResult {
    name: string;
    expected: string;
    calculated: string;
    pass: boolean;
}
/**
 * Test relevance scoring with sample decisions
 *
 * Task 1.4: Test relevance scoring with sample decisions
 * AC #1, #4: Verify scoring formula and failure priority
 *
 * @returns Test results
 */
export declare function testRelevanceScoring(): TestResult[];
//# sourceMappingURL=relevance-scorer.d.ts.map