feat: F009 matching engine V1 — deterministic multi-criteria scoring
domain/scoring.ts: HARD_FILTER thresholds, DATA_QUALITY/CONFIDENCE modifier tables, ScoringWeightProfile type, DEFAULT_SCORING_PROFILES for Office/Retail/Light Industrial/Logistics/Production/Default (all sum to 1.00), HARD/SOFT_CRITERION_KEYS, MatchEngineOutput type.
features/matching/scoreCalculator.ts: applyHardFilters (asset type mismatch, area < 85% min, budget > 150%, excluded region → hard exclude; occupied → severe penalty), scoreArea/Location/Budget/Timing as ScoreFactor, scoreSoftFactor for 9 keys mapped to property.softFactors, calcDataQualityModifier/calcConfidenceModifier, calculateScore (resolves profile from need.weightingProfile + default, runs filters, computes normalized hard/soft scores, applies modifiers, classifies positive/negative factors, assembles MatchEngineOutput).
features/matching/tradeOffAnalyzer.ts: analyzeTradeOffs (6 patterns: location-vs-budget, area-vs-budget, timing-vs-dataQuality, prestige-vs-flexibility, futureSignal-vs-location, accessibility-vs-commute), analyzeRisks (future signal, data quality, budget, occupied, low confidence, missing critical fields), identifyMissingData (rentPricePerSqm, availabilityDate, softFactors, hardFacts, need.budgetRange).
features/matching/rankingEngine.ts: matchStrengthFromScore (>=78 STRONG, >=52 MODERATE, else WEAK), generateNextBestActions (score-based, future signal SCHEDULE, missing data VERIFY), buildFullMatch → full Match entity with ScoreBreakdown + explainabilitySummary + uncertaintyIndicators, rankMatches (score desc → resultType order → confidence desc), computeRankedMatches batch helper.
services/matchService.ts: computeMatch(need, property) and computeMatchesForNeed(needId) wired to engine.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Benjamin Sutter
3152de004c