// src/main/services/PremiumLLMService.ts
// Phase 3.2: Anthropic Claude 프리미엄 LLM 서비스.
// 내부적으로는 온라인 API를 호출한다.
// Supabase Edge Function(`llm-proxy`)을 경유해 Claude Messages API를 호출.
//
// 특징:
// - 싱글톤 + EventEmitter (설계서 01 패턴)
// - processText(), chatStream() — 시그니처 동일
// - 네트워크 실패 / 401 / 429 / 5xx 감지 시 에러 throw → LLMRouterService가 local로 fallback
// - quota-warning / upgrade-required / fallback-triggered 이벤트 emit

import { EventEmitter } from 'events'
import { getLogger } from './LoggerService'
import { getCloudSyncService } from './CloudSyncService'
import { D3ROError, ErrorCode } from '@d3ro/core/errors'
import type { LLMAction } from '@d3ro/core/types'
import { resolveSystemPrompt } from './llm-prompts'
import { parseAnthropicSSE, AnthropicStreamError } from '../utils/sse-parser'
import {
  fitChatRequest,
  LLM_PROXY_CHAT_LIMITS,
  toChatRequest,
  type ChatRequest,
  type ChatStreamOptions,
  type ChatTurn,
  type RoleMessage,
} from '@d3ro/core/llm-chat'

const logger = getLogger('PremiumLLMService')

// ============================================================
// 내부 타입
// ============================================================

/** llm-proxy Edge Function 요청 body */
interface LlmProxyRequest {
  messages: ChatTurn[]
  system?: string
  max_tokens?: number
  model?: string
  stream?: boolean
}

/** Claude Messages API 비스트리밍 응답 */
interface ClaudeMessageResponse {
  id: string
  model: string
  role: 'assistant'
  content: Array<{ type: 'text'; text: string }>
  stop_reason: string
  usage: { input_tokens: number; output_tokens: number }
}

export interface QuotaUsageSnapshot {
  tier: 'free' | 'pro' | 'pro_plus'
  current: number
  limit: number
  overageCredits: number
}

type UpgradeReason = 'quota_exceeded' | 'model_not_allowed' | 'auth_required'

/** 호출 단위 취소·기한 상태. cancelGeneration() 은 활성 호출 전부를 취소한다. */
type AbortCause = 'timeout' | 'cancelled'

interface PremiumCall {
  controller: AbortController
  abortCause: AbortCause | null
  abort: (cause: AbortCause) => void
  close: () => void
}

/** llm-proxy 출력 토큰 상한 (llm-contract MAX_OUTPUT_TOKENS) */
const PROXY_MAX_OUTPUT_TOKENS = 4096
const DEFAULT_CHAT_MAX_TOKENS = 2048
/** 클라이언트 측 채팅 기한. 프록시가 공급자 호출에 45초 기한을 두므로 여유 있게 잡는다. */
const DEFAULT_CHAT_TIMEOUT_MS = 120_000

/**
 * llm-proxy 오류 메시지를 D3ROError 로 분류한다 (순수 함수).
 * upgradeReason 이 있으면 호출자가 'upgrade-required' 를 emit 한다.
 */
export function classifyProxyError(message: string): { error: D3ROError; upgradeReason: UpgradeReason | null } {
  if (message.includes('401') || message.includes('Unauthorized') || message.includes('auth')) {
    return {
      error: new D3ROError(ErrorCode.LLMServerUnreachable, `인증 실패: ${message}`),
      upgradeReason: 'auth_required',
    }
  }
  if (message.includes('quota_exceeded') || message.includes('429')) {
    return {
      error: new D3ROError(ErrorCode.LLMProcessingFailed, `쿼터 초과: ${message}`),
      upgradeReason: 'quota_exceeded',
    }
  }
  if (message.includes('model_not_allowed') || message.includes('403')) {
    return {
      error: new D3ROError(ErrorCode.LLMInvalidAction, `모델 권한 없음: ${message}`),
      upgradeReason: 'model_not_allowed',
    }
  }
  return {
    error: new D3ROError(ErrorCode.LLMProcessingFailed, `llm-proxy: ${message}`),
    upgradeReason: null,
  }
}

/**
 * invokeFunctionStream 오류가 프록시의 HTTP 응답(`<status>: <body>`)인지 판별한다.
 * 프록시가 응답했다면 같은 요청을 비스트리밍으로 다시 보내도 같은 실패(와 쿼터 소비)만
 * 되풀이되므로, 비스트리밍 폴백은 HTTP 상태가 없는 전송 계층 실패에서만 쓴다.
 */
export function proxyHttpStatus(message: string): number | null {
  const match = /^(\d{3}):/.exec(message)
  return match ? Number(match[1]) : null
}

function resolveMaxTokens(maxTokens: number | undefined): number {
  if (maxTokens === undefined) return DEFAULT_CHAT_MAX_TOKENS
  if (!Number.isSafeInteger(maxTokens) || maxTokens <= 0) {
    throw new D3ROError(ErrorCode.LLMProcessingFailed, 'maxTokens must be a positive safe integer')
  }
  return Math.min(maxTokens, PROXY_MAX_OUTPUT_TOKENS)
}

function resolveTimeoutMs(timeoutMs: number | undefined): number {
  if (timeoutMs === undefined) return DEFAULT_CHAT_TIMEOUT_MS
  if (!Number.isSafeInteger(timeoutMs) || timeoutMs <= 0) {
    throw new D3ROError(ErrorCode.LLMProcessingFailed, 'timeoutMs must be a positive safe integer')
  }
  return timeoutMs
}

/** ChatRequest → llm-proxy body. system 은 최상위 system 필드로 보낸다. */
function toProxyBody(
  request: ChatRequest,
  maxTokens: number,
  model: string | undefined,
  stream: boolean,
): LlmProxyRequest {
  const body: LlmProxyRequest = {
    messages: request.turns,
    max_tokens: maxTokens,
    stream,
  }
  if (request.system !== undefined) body.system = request.system
  if (model !== undefined) body.model = model
  return body
}

function firstText(response: ClaudeMessageResponse): string {
  const firstBlock = response.content?.[0]
  return firstBlock?.type === 'text' ? firstBlock.text : ''
}


interface PremiumLLMEvents {
  'quota-warning': (payload: { current: number; limit: number; overageCredits: number }) => void
  'upgrade-required': (payload: { reason: UpgradeReason }) => void
  'fallback-triggered': (payload: { reason: string }) => void
}

// ============================================================
// PremiumLLMService
// ============================================================

class PremiumLLMService extends EventEmitter {
  private _activeCalls = new Set<PremiumCall>()
  private _disposed = false
  private _lastQuota: QuotaUsageSnapshot | null = null

  /**
   * 사용 가능 여부 — Supabase URL + access token 모두 있어야 true.
   * (LLMRouter가 local/premium 분기 시 호출)
   */
  isAvailable(): boolean {
    const cloud = getCloudSyncService()
    return cloud.isEnabled() && cloud.isAuthenticated()
  }

  /**
   * 마지막 응답에서 서버가 알려준 쿼터 상태.
   * Settings UI에서 "오늘 X/250" 표시에 사용.
   */
  getLastQuota(): QuotaUsageSnapshot | null {
    return this._lastQuota
  }

  /**
   * disposed / 미인증 상태 공통 가드.
   * 문제 시 D3ROError throw, upgrade-required emit.
   */
  private _ensureAuth(): void {
    if (this._disposed) {
      throw new D3ROError(ErrorCode.LLMProcessingFailed, 'PremiumLLMService disposed')
    }
    const cloud = getCloudSyncService()
    if (!cloud.isEnabled() || !cloud.isAuthenticated()) {
      this.emit('upgrade-required', { reason: 'auth_required' })
      throw new D3ROError(
        ErrorCode.LLMServerUnreachable,
        'Premium LLM 사용 전 로그인 필요',
      )
    }
  }

  /**
   * 텍스트 액션 처리.
   * Phase 3.2 MVP는 비스트리밍 (stream=false).
   */
  async processText(
    text: string,
    action: LLMAction,
    targetLanguage?: string,
    customPrompt?: string,
  ): Promise<string> {
    this._ensureAuth()

    const systemPrompt = resolveSystemPrompt(action, targetLanguage, customPrompt)

    const body: LlmProxyRequest = {
      messages: [{ role: 'user', content: text }],
      system: systemPrompt,
      max_tokens: 2048,
      stream: false,
    }

    try {
      const response = await this._invokeProxy(body)
      // Claude 응답 → text 추출
      const firstBlock = response.content?.[0]
      if (!firstBlock || firstBlock.type !== 'text' || !firstBlock.text.trim()) {
        throw new D3ROError(ErrorCode.LLMProcessingFailed, 'Premium LLM returned empty content')
      }
      return firstBlock.text.trim()
    } catch (err) {
      // LLMRouter가 local fallback 처리. 여기서는 에러 전파.
      if (err instanceof D3ROError) throw err
      throw new D3ROError(
        ErrorCode.LLMProcessingFailed,
        `Premium LLM failed: ${err instanceof Error ? err.message : String(err)}`,
      )
    }
  }

  /**
   * 스트리밍 대화 (Voice Conversation / 회의 채팅용).
   * SSE 스트리밍: llm-proxy에 stream=true로 요청, Anthropic SSE를 토큰 단위 yield.
   *
   * @d3ro/core/llm-chat 계약을 따른다:
   * - role:'system' 메시지는 버리지 않고 body.system 으로 보낸다 (한도 초과 시 앞부분 보존).
   * - message_stop 없이 끊긴 스트림, 공급자 오류 이벤트, 빈 응답은 D3ROError 로 throw 한다.
   * - signal / timeoutMs / maxTokens 는 호출 단위로 적용된다. temperature 는 프록시가 받지 않는다.
   * - 비스트리밍 폴백은 전송 계층 실패에서만 쓴다 (프록시 HTTP 오류는 그대로 전파).
   */
  async *chatStream(
    messages: RoleMessage[],
    options?: ChatStreamOptions,
  ): AsyncGenerator<string, string> {
    this._ensureAuth()

    const request = fitChatRequest(toChatRequest(messages), LLM_PROXY_CHAT_LIMITS)
    if (request.turns.length === 0) {
      throw new D3ROError(ErrorCode.LLMProcessingFailed, 'Premium chat requires a user message')
    }
    const body = toProxyBody(request, resolveMaxTokens(options?.maxTokens), options?.model, true)
    const call = this._beginCall(options?.signal, resolveTimeoutMs(options?.timeoutMs))
    let completed = false

    try {
      const cloud = getCloudSyncService()
      const { stream, error } = await cloud.invokeFunctionStream(
        'llm-proxy',
        body as unknown as Record<string, unknown>,
        call.controller.signal,
      )

      if (error || !stream) {
        const msg = error?.message ?? 'Stream unavailable'
        if (call.controller.signal.aborted) {
          throw new D3ROError(ErrorCode.LLMProcessingCancelled, msg)
        }
        logger.error(`SSE stream failed: ${msg}`)

        if (proxyHttpStatus(msg) !== null) {
          // 프록시가 응답한 오류 — 재요청은 같은 실패와 쿼터 소비만 되풀이한다.
          throw this._rejectProxyError(msg)
        }

        // 전송 계층 실패 시 비스트리밍 fallback (같은 system/maxTokens/signal 적용)
        logger.info('Falling back to non-streaming Premium LLM')
        const response = await this._invokeProxy({ ...body, stream: false }, call.controller.signal)
        const text = firstText(response)
        if (!text.trim()) {
          throw new D3ROError(ErrorCode.LLMProcessingFailed, 'Premium LLM returned empty content')
        }
        completed = true
        yield text
        return text
      }

      let accumulated = ''
      for await (const token of parseAnthropicSSE(stream)) {
        accumulated += token
        yield token
      }
      completed = true
      if (!accumulated.trim()) {
        throw new D3ROError(ErrorCode.LLMProcessingFailed, 'Premium LLM returned empty content')
      }
      return accumulated
    } catch (err) {
      throw this._toChatError(err, call)
    } finally {
      // 완료 전 종료(소비자 break, 오류) 시 연결을 끊어 프록시 스트림을 정리한다.
      if (!completed) call.abort('cancelled')
      call.close()
    }
  }

  /**
   * 제목/요약/액션 플랜 등 자유 생성. processText 와 같은 프록시 경로를 탄다.
   */
  async generate(
    text: string,
    options?: { systemPrompt?: string; temperature?: number; maxTokens?: number },
  ): Promise<{ text: string }> {
    this._ensureAuth()
    const response = await this._invokeProxy({
      messages: [{ role: 'user', content: text }],
      system: options?.systemPrompt,
      max_tokens: options?.maxTokens ?? 2048,
      stream: false,
    })
    const firstBlock = response.content?.[0]
    if (!firstBlock || firstBlock.type !== 'text' || !firstBlock.text.trim()) {
      throw new D3ROError(ErrorCode.LLMProcessingFailed, 'Premium LLM generate returned empty content')
    }
    return { text: firstBlock.text.trim() }
  }

  cancelGeneration(): void {
    if (this._activeCalls.size === 0) return
    for (const call of [...this._activeCalls]) {
      call.abort('cancelled')
    }
    logger.info('Premium LLM generation cancelled')
  }

  dispose(): void {
    this._disposed = true
    this.cancelGeneration()
    this.removeAllListeners()
    logger.info('PremiumLLMService disposed')
  }

  // ── 내부: Edge Function 호출 ─────────────────────────────

  private async _invokeProxy(body: LlmProxyRequest, signal?: AbortSignal): Promise<ClaudeMessageResponse> {
    const cloud = getCloudSyncService()

    // Supabase JS 클라이언트의 functions.invoke() 사용 — auth 헤더를 올바르게 처리.
    // raw fetch + Authorization: Bearer 방식은 Supabase gateway가 401로 거부.
    const { data, error } = signal
      ? await cloud.invokeFunction('llm-proxy', body as unknown as Record<string, unknown>, { signal })
      : await cloud.invokeFunction('llm-proxy', body as unknown as Record<string, unknown>)

    if (error) {
      const msg = error.message ?? 'Edge Function error'
      logger.error(`llm-proxy error: ${msg}`)
      throw this._rejectProxyError(msg)
    }

    // functions.invoke는 response body를 자동 파싱해서 data에 넣음
    const result = data as ClaudeMessageResponse
    if (!result?.content) {
      logger.warn('llm-proxy returned unexpected shape — falling back')
      throw new D3ROError(ErrorCode.LLMProcessingFailed, 'llm-proxy returned invalid response')
    }

    return result
  }

  /** 프록시 오류를 분류하고 필요 시 upgrade-required 를 알린다. */
  private _rejectProxyError(message: string): D3ROError {
    const { error, upgradeReason } = classifyProxyError(message)
    if (upgradeReason) this.emit('upgrade-required', { reason: upgradeReason })
    return error
  }

  /** 호출 단위 AbortController 를 만들고 외부 signal·기한·cancelGeneration 에 연결한다. */
  private _beginCall(externalSignal: AbortSignal | undefined, timeoutMs: number): PremiumCall {
    const controller = new AbortController()
    const onExternalAbort = (): void => call.abort('cancelled')
    const timer = setTimeout(() => call.abort('timeout'), timeoutMs)
    const call: PremiumCall = {
      controller,
      abortCause: null,
      abort: (cause) => {
        if (call.abortCause !== null) return
        call.abortCause = cause
        controller.abort()
      },
      close: () => {
        clearTimeout(timer)
        externalSignal?.removeEventListener('abort', onExternalAbort)
        this._activeCalls.delete(call)
      },
    }
    this._activeCalls.add(call)
    if (externalSignal?.aborted) {
      onExternalAbort()
    } else {
      externalSignal?.addEventListener('abort', onExternalAbort, { once: true })
    }
    return call
  }

  /** 채팅 스트림 실패를 호출 단위 원인에 맞는 D3ROError 로 정규화한다. */
  private _toChatError(error: unknown, call: PremiumCall): D3ROError {
    if (call.abortCause === 'timeout') {
      return new D3ROError(ErrorCode.LLMProcessingTimeout, 'Premium LLM generation timed out')
    }
    if (call.abortCause === 'cancelled' || call.controller.signal.aborted) {
      return new D3ROError(ErrorCode.LLMProcessingCancelled, 'Premium LLM generation cancelled')
    }
    if (error instanceof D3ROError) return error
    if (error instanceof AnthropicStreamError) {
      logger.warn(`Premium SSE stream failed (${error.kind}): ${error.message}`)
      return new D3ROError(ErrorCode.LLMProcessingFailed, `Premium LLM stream failed: ${error.message}`)
    }
    return new D3ROError(
      ErrorCode.LLMProcessingFailed,
      `Premium LLM failed: ${error instanceof Error ? error.message : String(error)}`,
    )
  }

  // ── EventEmitter 타입 오버라이드 ───────────────────────

  on<K extends keyof PremiumLLMEvents>(event: K, listener: PremiumLLMEvents[K]): this {
    return super.on(event, listener)
  }

  emit<K extends keyof PremiumLLMEvents>(
    event: K,
    ...args: Parameters<PremiumLLMEvents[K]>
  ): boolean {
    return super.emit(event, ...args)
  }
}

// ── 싱글톤 ─────────────────────────────────────────────────

let _instance: PremiumLLMService | null = null

export function resetPremiumLLMServiceForTests(): void {
  if (_instance) _instance.removeAllListeners()
  _instance = null
}

export function getPremiumLLMService(): PremiumLLMService {
  if (!_instance) {
    _instance = new PremiumLLMService()
  }
  return _instance
}

export type { PremiumLLMService }