import type { CreateNeedInput } from '../../../domain/need'
import type { ParsedNeedCriteria, FollowUpQuestion } from '../../../domain/needBuilder'
import type { UnifiedMatchResult } from '../../../domain/unifiedResult'
import type {
  IAIService,
  AIResponse,
  DecisionBrief,
  ComparisonSummary,
  CriteriaExtractionResult,
  OfferEmailPayload,
  MatchExplanationInput,
  MatchExplanation,
  TradeOffInput,
  TradeOffSummary,
  DataQualityInput,
  DataQualitySummary,
  MarketSignalClassification,
} from '../IAIService'
import { mockProvenance } from '../IAIService'
import { aiTraceStore } from '../tracing'
import { mockParseNeed } from './needParser'
import { buildComparisonSummary } from './compareBuilder'
import { buildMockDecisionBrief } from './decisionBrief'

const SIMULATED_DELAY = { fast: 300, medium: 600, slow: 1800 }
const delay = (ms: number) => new Promise(r => setTimeout(r, ms))

// ── Tracing wrapper ───────────────────────────────────────────────────────────

async function traceMock<T>(method: string, fn: () => Promise<AIResponse<T>>): Promise<AIResponse<T>> {
  const startMs = Date.now()
  const result = await fn()
  aiTraceStore.add({
    id: crypto.randomUUID(),
    method,
    provider: 'mock',
    model: 'mock',
    promptVersion: 'mock',
    latencyMs: Date.now() - startMs,
    fallbackUsed: false,
    validationPassed: true,
    responseValidationStatus: 'valid',
    source: 'mock',
    createdAt: new Date().toISOString(),
  })
  return result
}

// ── Follow-up question templates keyed by ParsedNeedCriteria field ────────────

interface QuestionTemplate {
  questionText: string
  reason: string
  suggestedAnswerOptions?: string[]
  importance: FollowUpQuestion['importance']
}

const FOLLOW_UP_TEMPLATES: Partial<Record<keyof ParsedNeedCriteria, QuestionTemplate>> = {
  assetType: {
    questionText: 'Welchen Nutzungstyp suchen Sie?',
    reason: 'Nutzungstyp ist zwingend für die Matchsuche',
    suggestedAnswerOptions: ['Büro', 'Retail', 'Logistik', 'Produktion', 'Leichtindustrie', 'Gastro'],
    importance: 'required',
  },
  areaRange: {
    questionText: 'Welche Fläche benötigen Sie (min–max in m²)?',
    reason: 'Flächenbedarf ist zwingend für die Filterung',
    importance: 'required',
  },
  preferredLocations: {
    questionText: 'In welchen Städten oder Regionen suchen Sie?',
    reason: 'Standortpräferenz fehlt',
    suggestedAnswerOptions: ['Zürich', 'Basel', 'Bern', 'Zug', 'Genf', 'Lausanne'],
    importance: 'required',
  },
  budgetRange: {
    questionText: 'Was ist Ihr maximales Budget pro m² und Jahr (CHF)?',
    reason: 'Budget ist wichtig für die Filterung unpassender Objekte',
    importance: 'recommended',
  },
  timing: {
    questionText: 'Wann möchten Sie spätestens einziehen?',
    reason: 'Verfügbarkeitstermin fehlt',
    importance: 'recommended',
  },
  mustHaveCriteria: {
    questionText: 'Haben Sie zwingende Anforderungen (ÖV-Anbindung, Parkplätze, Laderampe)?',
    reason: 'Pflichtkriterien sind für die Filterung relevant',
    importance: 'optional',
  },
}

const AREA_AMBIGUITY_RATIO_THRESHOLD = 8

const AREA_AMBIGUITY_QUESTION: QuestionTemplate = {
  questionText: 'Ihre Flächenangabe ist sehr weit gefasst — können Sie den Bereich präzisieren (z.B. min 300 m², max 600 m²)?',
  reason: 'Zu grosse Spanne reduziert die Matchgenauigkeit erheblich',
  importance: 'required',
}

function isAreaAmbiguous(areaRange: NonNullable<ParsedNeedCriteria['areaRange']>): boolean {
  const { min, max } = areaRange
  if (min <= 0 || max <= 0) return true
  return max / min > AREA_AMBIGUITY_RATIO_THRESHOLD
}

// Priority order: required fields first, recommended next, optional last.
// Max 3 questions returned. Area range ambiguity is detected and raised as a
// required clarification even when areaRange is nominally present.
const FIELD_PRIORITY: Array<keyof ParsedNeedCriteria> = [
  'assetType',
  'areaRange',
  'preferredLocations',
  'budgetRange',
  'timing',
  'mustHaveCriteria',
]

function buildFollowUpQuestions(criteria: ParsedNeedCriteria): FollowUpQuestion[] {
  const questions: FollowUpQuestion[] = []
  let idx = 0

  for (const field of FIELD_PRIORITY) {
    if (questions.length >= 3) break

    if (field === 'areaRange') {
      if (!criteria.areaRange) {
        const tpl = FOLLOW_UP_TEMPLATES['areaRange']!
        questions.push({ id: `fq-mock-${idx++}`, questionText: tpl.questionText, targetField: 'areaRange', reason: tpl.reason, importance: tpl.importance })
      } else if (isAreaAmbiguous(criteria.areaRange)) {
        questions.push({ id: `fq-mock-${idx++}`, questionText: AREA_AMBIGUITY_QUESTION.questionText, targetField: 'areaRange', reason: AREA_AMBIGUITY_QUESTION.reason, importance: AREA_AMBIGUITY_QUESTION.importance })
      }
      continue
    }

    const isMissing =
      field === 'preferredLocations' ? !criteria.preferredLocations?.length
      : field === 'mustHaveCriteria' ? !criteria.mustHaveCriteria?.length
      : !criteria[field]

    if (isMissing) {
      const tpl = FOLLOW_UP_TEMPLATES[field]
      if (!tpl) continue
      questions.push({
        id:                     `fq-mock-${idx++}`,
        questionText:           tpl.questionText,
        targetField:            field,
        reason:                 tpl.reason,
        importance:             tpl.importance,
        suggestedAnswerOptions: tpl.suggestedAnswerOptions,
      })
    }
  }

  return questions
}

// ── Service ───────────────────────────────────────────────────────────────────

export const MockAIService: IAIService = {
  parseNeed: (input: string) =>
    traceMock('parseNeed', async () => {
      await delay(SIMULATED_DELAY.fast)
      return { data: mockParseNeed(input), provenance: mockProvenance() }
    }),

  generateFollowUpQuestions: (criteria: ParsedNeedCriteria) =>
    traceMock('generateFollowUpQuestions', async () => {
      await delay(SIMULATED_DELAY.medium)
      return { data: buildFollowUpQuestions(criteria), provenance: mockProvenance() }
    }),

  generateMatchExplanation: (input: MatchExplanationInput) =>
    traceMock('generateMatchExplanation', async () => {
      await delay(SIMULATED_DELAY.medium)
      const isStrong = input.matchScore >= 78
      const isMedium = input.matchScore >= 52
      const headline = isStrong
        ? `Starkes Match — ${input.propertyTitle} erfüllt Ihre Kernkriterien hervorragend`
        : isMedium
          ? `Gutes Match mit einzelnen Kompromissen für ${input.propertyTitle}`
          : `Schwaches Match — mehrere Kriterien nicht erfüllt bei ${input.propertyTitle}`
      const positiveText = input.positiveFactors.slice(0, 2).map(f => f.explanation).join('; ')
      const negativeText = input.negativeFactors.slice(0, 1).map(f => f.explanation).join('; ')
      const summary = `${input.propertyTitle} in ${input.propertyCity} erreicht ${input.matchScore}/100 Punkte.${positiveText ? ` Hauptstärken: ${positiveText}.` : ''}${negativeText ? ` Einschränkung: ${negativeText}.` : ''}`
      return {
        data: {
          headline,
          summary,
          keyReasons: [
            ...input.positiveFactors.slice(0, 2).map(f => `+ ${f.explanation}`),
            ...input.negativeFactors.slice(0, 1).map(f => `− ${f.explanation}`),
          ],
        },
        provenance: mockProvenance(),
      }
    }),

  summarizeTradeOffs: (tradeoffs: TradeOffInput[]) =>
    traceMock('summarizeTradeOffs', async () => {
      await delay(SIMULATED_DELAY.fast)
      const critical = tradeoffs.filter(t => t.severity === 'HIGH')
      const overallRisk: TradeOffSummary['overallRisk'] =
        critical.length >= 2 ? 'HIGH' : critical.length === 1 ? 'MEDIUM' : 'LOW'
      const riskLabel = overallRisk === 'HIGH' ? 'Hoch' : overallRisk === 'MEDIUM' ? 'Mittel' : 'Gering'
      return {
        data: {
          headline: tradeoffs.length === 0
            ? 'Keine wesentlichen Trade-offs identifiziert'
            : `${tradeoffs.length} Trade-off${tradeoffs.length > 1 ? 's' : ''} — Gesamtrisiko: ${riskLabel}`,
          items: tradeoffs.map(t => ({ concern: t.concern, severity: t.severity, mitigation: t.mitigation })),
          overallRisk,
        },
        provenance: mockProvenance(),
      }
    }),

  summarizeComparison: (items: UnifiedMatchResult[]) =>
    traceMock('summarizeComparison', async () => {
      await delay(SIMULATED_DELAY.medium)
      return { data: buildComparisonSummary(items), provenance: mockProvenance() }
    }),

  generateDecisionBrief: (shortlistId: string) =>
    traceMock('generateDecisionBrief', async () => {
      await delay(SIMULATED_DELAY.slow)
      return { data: buildMockDecisionBrief(shortlistId), provenance: mockProvenance() }
    }),

  generateDataQualitySummary: (_propertyId: string, quality: DataQualityInput) =>
    traceMock('generateDataQualitySummary', async () => {
      await delay(SIMULATED_DELAY.fast)
      const level =
        quality.score >= 0.85 ? 'excellent'
        : quality.score >= 0.70 ? 'good'
        : quality.score >= 0.55 ? 'fair'
        : quality.score >= 0.40 ? 'poor'
        : 'critical'
      const assessments: Record<string, string> = {
        excellent: 'Exzellente Datenqualität — alle Kernfelder vollständig und aktuell.',
        good: 'Gute Datenqualität — kleinere Lücken beeinflussen die Matchgenauigkeit nicht wesentlich.',
        fair: 'Ausreichende Datenqualität — fehlende Felder können die Matchgenauigkeit beeinträchtigen.',
        poor: 'Geringe Datenqualität — wichtige Felder fehlen, Match-Score mit Vorsicht interpretieren.',
        critical: 'Kritische Datenqualität — fundamentale Felder fehlen, Match-Ergebnis stark eingeschränkt.',
      }
      const hasCritical = quality.missingCriticalFields.length > 0
      return {
        data: {
          overallAssessment: assessments[level],
          missingCriticalFields: quality.missingCriticalFields,
          recommendation: hasCritical
            ? `Fehlende Pflichtfelder ergänzen: ${quality.missingCriticalFields.join(', ')}`
            : quality.score < 0.70
              ? 'Daten aktualisieren und optionale Felder ergänzen für bessere Matchgenauigkeit.'
              : 'Keine sofortigen Massnahmen erforderlich.',
          confidence: quality.score,
        },
        provenance: mockProvenance(),
      }
    }),

  classifyMarketSignal: (signalText: string) =>
    traceMock('classifyMarketSignal', async () => {
      await delay(SIMULATED_DELAY.medium)
      const t = signalText.toLowerCase()
      let signalType: MarketSignalClassification['signalType'] = 'UNKNOWN'
      if (t.includes('neubau') || t.includes('baubewilligung') || t.includes('umbau')) signalType = 'CONSTRUCTION'
      else if (t.includes('expansion') || t.includes('wachstum') || t.includes('sucht fläche')) signalType = 'EXPANSION'
      else if (t.includes('verlegt') || t.includes('umzug') || t.includes('relocation')) signalType = 'RELOCATION'
      else if (t.includes('stellenabbau') || t.includes('restruktur') || t.includes('fusion')) signalType = 'RESTRUCTURING'
      else if (t.includes('frei') || t.includes('kündigung') || t.includes('schliessung') || t.includes('leerstand')) signalType = 'VACANCY'
      const areaMatch = signalText.match(/(\d{2,5})\s*m²/)
      const monthsMatch = signalText.match(/(\d{1,2})\s*Monate?n?/)
      return {
        data: {
          signalType,
          probability: 0.65,
          timeHorizonMonths: monthsMatch ? parseInt(monthsMatch[1]) : null,
          areaSqmEstimate: areaMatch ? parseInt(areaMatch[1]) : null,
          credibility: 'MEDIUM',
          reasoning: `Keyword-basierte Klassifikation (Mock). Signaltyp: ${signalType}.`,
        },
        provenance: mockProvenance(),
      }
    }),

  generateOfferEmail: (payload: OfferEmailPayload) =>
    traceMock('generateOfferEmail', async () => {
      await delay(SIMULATED_DELAY.medium * 2)
      return {
        data: {
          subject: `Passende Gewerbeflächen zu Ihrer Anfrage: ${payload.needTitle}`,
          body:
            `Sehr geehrte Damen und Herren,\n\nvielen Dank für Ihr Interesse. Gerne unterbreiten wir Ihnen folgende passende Gewerbeobjekte aus unserem Portfolio:\n\n` +
            payload.properties.map((p, i) => `• ${p} (Match-Score: ${payload.matchScores[i]}%)`).join('\n') +
            `\n\nGerne arrangieren wir Besichtigungstermine für die genannten Objekte und stehen für alle weiteren Fragen zur Verfügung.\n\nFreundliche Grüsse\nWincasa AG`,
        },
        provenance: mockProvenance(),
      }
    }),

  // Legacy methods
  extractCriteria: (_input: string) =>
    traceMock('extractCriteria', async () => ({
      data: {
        extractedCriteria: {
          companyName: 'Unbekannt (bitte bestätigen)',
          requiredArea: { min: 400, max: 900 },
          budgetRange: { maxPerSqm: 40, currency: 'CHF' },
        },
        confidence: 0.72,
        missingFields: ['assetType', 'timing', 'preferredLocations'],
        assumptions: ['Fläche aus Zahlenangabe geschätzt', 'Budget aus Kostennennung abgeleitet'],
        followUpQuestions: [
          'Welchen Nutzungstyp suchen Sie (Büro, Retail, Logistik)?',
          'In welchen Städten oder Regionen suchen Sie?',
          'Wann möchten Sie spätestens einziehen?',
        ],
      },
      provenance: mockProvenance(),
    })),

  generateFollowUp: (partialNeed: Partial<CreateNeedInput>) =>
    traceMock('generateFollowUp', async () => {
      const questions: string[] = []
      if (!partialNeed.assetType) questions.push('Welchen Nutzungstyp suchen Sie?')
      if (!partialNeed.preferredLocations?.length) questions.push('In welchen Regionen suchen Sie?')
      if (!partialNeed.timing) questions.push('Was ist Ihr gewünschter Einzugstermin?')
      if (!partialNeed.budgetRange) questions.push('Was ist Ihr maximales monatliches Budget?')
      return { data: questions, provenance: mockProvenance() }
    }),
}