Commit Graph

6 Commits

Author SHA1 Message Date
Benjamin Sutter fb029cf0bc feat(ai): AI pre-market rent recommendation from regional comparables, supply & demand
- IAIService.recommendPreMarketRent: recommended price + range, verdict (UNDERPRICED/FAIR/AMBITIOUS), drivers, rationale, confidence
- MockAIService: deterministic recommendation from locationIntelligence — regional comp median, vacancy (supply), demand strength + days-on-market, rent trend (forward for pre-market)
- BackendAIService: LLM prompt with market context + mock fallback
- usePreMarketRentRecommendation hook; PreMarketPriceAdvisor component shows the recommendation per released unit with verdict ("zu günstig" when underpriced) + adjustable expected price + "Empfehlung übernehmen"
- Replaces the simple indexed suggestion with a market-driven AI recommendation that flags underpricing

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
2026-06-21 00:22:14 +02:00
Benjamin Sutter b2530f9e20 feat: Reale Jahresbelastung, must-have fixes, fitOut data coverage
Reale Jahresbelastung (Change 6):
- FitOutCostPanel zeigt Jahresmiete + amortisierte Ausbaukosten für alle fitOut-Werte
- FULL/PREMIUM: Ausbau CHF 0 (bezugsfertig), SHELL/BASIC: CRB/BKP-Richtwerte amortisiert über 5 J.
- Match-Card-Chip zeigt geschätzte Investition (orange wenn >100k)
- FitOutInvestment-Typ + calcFitOutInvestment() in fitOutUtils.ts
- BackendAIService + MockAIService mit generateFitOutAdvice (IAIService-Interface)

Must-have Kriterien (Nicht prüfbar Fix):
- ÖV-Anbindung: Minutengrenze aus Freitext extrahiert, gegen publicTransportMinutes geprüft
- Mindestfläche: m²-Wert aus Freitext extrahiert, gegen areaSqm geprüft
- Ausbaugrad: neues Keyword-Rule für FULL/PREMIUM

fitOut-Datenpflege:
- fitOut-Werte zu 32 fehlenden Properties ergänzt (Logistik=SHELL, Standard=BASIC, Modern=FULL)
- MAB-Werte (200/150/250 CHF/m²) zu 3 BASIC-Objekten hinzugefügt

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-06-06 20:09:55 +02:00
Benjamin Sutter e1f4beb898 refactor: architecture compliance pass — DS tokens, hook boundary, god component split, AI hardening
- DS token migration: Anfragen.tsx + child components (AnfragenInquiryItem, AnfragenMessageBubble)
  fully migrated; DS_TEXT.brandDark added; scoreTheme.ts moved to src/lib/ with re-export proxy
- Hook boundary: Results.tsx no longer calls needService directly — routes through useNeeds()
  with optional refetchOnMount/gcTime overrides
- NewListing.tsx (440L) split into useNewListingForm hook + 8 section components under
  src/components/new-listing/; page shell reduced to 121 lines
- AI hardening: Zod .strict() on all schemas, AIProvenance extended with schemaVersion/
  fallbackReason/traceId/latencyMs, AITraceStore stats with p50/p90/p99 + failure breakdowns,
  MockAIService buildFollowUpQuestions with priority ordering + area-ambiguity detection,
  prompt templates updated (LIGHT_INDUSTRIAL, budget unit, ambiguity detection, decimal precision)
- Tests: all 154 passing; fixed test regression caused by OfferEmailResponseSchema body min(50)

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-24 16:10:39 +02:00
Benjamin Sutter e62391af66 feat: Zod AI validation, AIProvenance governance, fix tests (154 green)
- Add AIProvenance + AIResponse<T> to IAIService — all 11 methods now
  return structured provenance (provider, model, source, fallbackUsed,
  validationPassed) instead of bare ItemResponse<T>
- Add schemas.ts with Zod schemas for all 8 AI response types;
  validateAIResponse() utility returns null on failure, never throws
- Rewrite OpenRouterAIService: every method validates AI JSON against
  its Zod schema; failed validation triggers MockAIService fallback
  with fallbackUsed:true — no invalid data can reach the UI
- Fix MockAIService.generateFollowUpQuestions: replace broken
  mockParseNeed(JSON.stringify(criteria)) with direct ParsedNeedCriteria
  field inspection; returns max 3 prioritised FollowUpQuestion objects
- Add provenance: mockProvenance() to all MockAIService responses
- Improve decisionBriefPrompt: structured JSON schema example,
  confidence vocabulary, availability disclaimer
- Improve matchExplanationPrompt: score-tier vocabulary, isFutureSignal
  flag forbids confirmed-availability language for future signals
- Add 102 new tests: mustHaveScorer (16), softFactorEnrichment (38),
  aiSchemas (52) — 154 total, all passing; 0 TypeScript errors

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-24 13:44:46 +02:00
Benjamin Sutter c972392b78 feat: OpenRouter-ready AI service — all 11 methods implemented
IAIService:
+ generateMatchExplanation, summarizeTradeOffs, generateDataQualitySummary,
  classifyMarketSignal (4 new methods covering all documented AI output types)

OpenRouterAIService:
- All 11 methods now make real API calls via withFallback() pattern
- Every fallback is explicit (console.warn/error) — no silent mock bleed-through
- Proper JSON extraction with type-safe parsers, no any casts
- parseNeed: AI JSON → ParseNeedResult mapping (no TODO stubs)
- generateFollowUpQuestions, generateMatchExplanation, summarizeTradeOffs,
  generateDataQualitySummary, classifyMarketSignal, generateOfferEmail,
  extractCriteria, generateFollowUp: fully implemented

Prompts: +followUpQuestionsPrompt, +tradeOffPrompt, +dataQualityPrompt, +marketSignalPrompt

Factory (index.ts):
- VITE_AI_PROVIDER=mock|openrouter (new, takes priority)
- VITE_USE_REAL_AI=true still supported (legacy compat)
- Missing API key → explicit console.warn + MockAIService fallback

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-24 12:43:51 +02:00
Benjamin Sutter 6515acb7f0 feat: unified error handling + AI service modularisation
Error handling (Prompt 2):
- src/services/errors.ts: AppError class, normalizeError(), throwServiceError() helper
- 6 services wrapped with try/catch (property, match, need, shortlist, futureSignal, inquiry)
- inquiryService aligned from custom ServiceResult<T> to standard ServiceResponse types
- Results, MatchCenter, FutureAvailability pages show <ErrorState onRetry> on query failure

AI modularisation (Prompt 3):
- src/services/aiService.ts reduced from 755 → 19 lines (barrel re-export)
- src/services/ai/IAIService.ts: typed interface + all response types
- src/services/ai/mock/: needParser, compareBuilder, decisionBrief, listingParser, MockAIService
- src/services/ai/openrouter/OpenRouterAIService.ts: model-agnostic skeleton
- src/services/ai/prompts/: 4 prompt template files (needParsing, matchExplanation, compareSummary, decisionBrief)
- src/services/ai/index.ts: factory selects Mock or OpenRouter via VITE_USE_REAL_AI flag
- All existing import paths unchanged — zero call-site modifications

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
2026-05-24 00:32:01 +02:00