d3ro-voice/apps/desktop/tests/red/voice.usecase.test.ts

import { describe, it, expect, vi, beforeEach } from 'vitest'
import { EventEmitter } from 'events'
import { RecognitionState } from '@d3ro/core/types'
import { ErrorCode } from '@d3ro/core/errors'
import { IPC_CHANNELS } from '@d3ro/core/ipc-channels'
import { FX } from './fixtures'
import { invokeIpc, useRedHarness } from './harness'

const audioBus = new EventEmitter()
const mockAudio = {
  start: vi.fn(async () => undefined),
  stop: vi.fn(async () => undefined),
  on: (ev: string, fn: (...args: unknown[]) => void) => audioBus.on(ev, fn),
  off: (ev: string, fn: (...args: unknown[]) => void) => audioBus.off(ev, fn),
}

const mockSTT = {
  initialize: vi.fn(async () => undefined),
  transcribe: vi.fn(async () => ({
    text: FX.STT_OK,
    segments: [],
    language: 'ko',
    duration: 1,
    processingTime: 10,
  })),
  getModels: vi.fn(() => [
    { id: 'large-v3-turbo', name: 'Turbo', sizeBytes: 0, downloaded: true, languages: [] },
    { id: 'base', name: 'Base', sizeBytes: 0, downloaded: true, languages: [] },
  ]),
  getStatus: vi.fn(() => ({ engineState: 'ready', activeModel: 'large-v3-turbo', engineVersion: null, gpuAccelerated: false })),
}

const mockLLM = {
  isAvailable: vi.fn(() => true),
  processText: vi.fn(async () => FX.LLM_OK),
  generate: vi.fn(async () => ({ text: FX.LLM_OK })),
  cancelGeneration: vi.fn(),
  on: vi.fn(),
  off: vi.fn(),
}

const mockLicense = {
  canUse: vi.fn(() => ({ allowed: true, reason: 'ok' })),
  consumeQuota: vi.fn(),
  promptUpgrade: vi.fn(),
  initialize: vi.fn(),
  getInfo: vi.fn(() => ({ tier: 'free' })),
  on: vi.fn(),
}

const mockCaption = {
  getState: vi.fn(() => 'inactive'),
  start: vi.fn(async () => undefined),
  stop: vi.fn(async () => undefined),
  on: vi.fn(),
  off: vi.fn(),
}

vi.mock('../../src/main/services/AudioCaptureService', () => ({
  getAudioCaptureService: () => mockAudio,
  calculateRMS: () => 0.1,
}))

vi.mock('../../src/main/services/LocalSTTService', () => ({
  getLocalSTTService: () => mockSTT,
  resetLocalSTTServiceForTests: () => undefined,
}))

vi.mock('../../src/main/services/LocalLLMService', () => ({
  getLocalLLMService: () => ({
    isAvailable: vi.fn(() => true),
    processText: (...args: unknown[]) => mockLLM.processText(...args),
    chatStream: mockLLM.chatStream,
    cancelGeneration: vi.fn(),
    getStatus: () => ({
      connectionState: 'connected',
      serverUrl: 'http://localhost:11434',
      activeModel: 'gemma4:e4b',
      serverVersion: null,
    }),
  }),
  resetLocalLLMServiceForTests: () => undefined,
}))

vi.mock('../../src/main/services/PremiumLLMService', () => ({
  getPremiumLLMService: () => mockLLM,
  resetPremiumLLMServiceForTests: () => undefined,
}))

vi.mock('../../src/main/services/TextInsertService', () => ({
  getTextInsertService: () => ({
    insertText: vi.fn(async () => ({ success: true, method: 'clipboard', textLength: 1, durationMs: 1 })),
  }),
}))

vi.mock('../../src/main/services/LicenseService', () => ({
  getLicenseService: () => mockLicense,
  resetLicenseServiceForTests: () => undefined,
  initLicenseService: () => undefined,
}))

vi.mock('../../src/main/services/CaptionService', () => ({
  getCaptionService: () => mockCaption,
  resetCaptionServiceForTests: () => undefined,
}))

vi.mock('../../src/main/services/HotkeyService', () => ({
  getHotkeyService: () => ({
    on: vi.fn(),
    off: vi.fn(),
  }),
}))

useRedHarness()

import { getVoiceModeService } from '../../src/main/services/VoiceModeService'
import { registerVoiceHandlers } from '../../src/main/ipc/voice-handlers'
import { configSet } from '../../src/main/services/ConfigService'
import { getHistoryService } from '../../src/main/services/HistoryService'
import {
  persistCompletedVoiceSession,
  persistCompletedVoiceSessionSafe,
} from '../../src/main/voice-session-persist'
import { unbindTestDatabase } from '../../src/main/db'

function feedAudio(): void {
  audioBus.emit('audio-data', { buffer: Buffer.alloc(16000 * 2) })
}

async function startAndRecord(): Promise<ReturnType<typeof getVoiceModeService>> {
  const svc = getVoiceModeService()
  await svc.startSession('dictation')
  feedAudio()
  await new Promise((r) => setTimeout(r, 750))
  feedAudio()
  return svc
}

describe('유스케이스: 받아쓰기 시작/정지/취소 / STT·LLM 실패 표면화', () => {
  beforeEach(() => {
    mockSTT.initialize.mockReset()
    mockSTT.transcribe.mockReset()
    mockLLM.processText.mockReset()
    mockSTT.initialize.mockResolvedValue(undefined)
    mockSTT.transcribe.mockResolvedValue({
      text: FX.STT_OK,
      segments: [],
      language: 'ko',
      duration: 1,
      processingTime: 10,
    })
    mockLLM.processText.mockResolvedValue(FX.LLM_OK)
    mockLicense.canUse.mockReturnValue({ allowed: true, reason: 'ok' })
    mockCaption.getState.mockReturnValue('inactive')
    mockAudio.start.mockClear()
    audioBus.removeAllListeners()
    configSet('defaultLLMAction', 'refine')
    configSet('autoInsert', false)
  })

  it('초기 음성 상태는 IDLE 이다', () => {
    const state = getVoiceModeService().getState()
    expect(state.recognitionState).toBe(RecognitionState.IDLE)
    expect(state.sessionId).toBeNull()
    expect(getVoiceModeService().isActive).toBe(false)
  })

  it('녹음 시작 시 세션이 활성화된다', async () => {
    const svc = getVoiceModeService()
    await svc.startSession('dictation')
    expect(svc.isActive).toBe(true)
    expect(svc.getState().sessionId).toBeTruthy()
  })

  it('사용자가 취소하면 session-cancelled(user) 가 난다', async () => {
    const svc = getVoiceModeService()
    let reason: string | null = null
    svc.on('session-cancelled', (p: { reason: string }) => {
      reason = p.reason
    })
    await svc.startSession('dictation')
    svc.cancelSession()
    expect(reason).toBe('user')
    expect(svc.getState().recognitionState).toBe(RecognitionState.IDLE)
  })

  it('너무 짧은 녹음은 too-short 로 취소된다', async () => {
    const svc = getVoiceModeService()
    let reason: string | null = null
    svc.on('session-cancelled', (p: { reason: string }) => {
      reason = p.reason
    })
    await svc.startSession('dictation')
    await svc.stopSession()
    expect(reason).toBe('too-short')
  })

  it('정상 녹음 후 STT 픽스처가 전사로 올라온다', async () => {
    const svc = await startAndRecord()
    let finalText: string | null = null
    svc.on('session-completed', (p: { finalText: string }) => {
      finalText = p.finalText
    })
    await svc.stopSession()
    expect(mockSTT.transcribe).toHaveBeenCalled()
    expect(finalText === FX.LLM_OK || finalText === FX.STT_OK).toBe(true)
  })

  it('STT 가 빈 문자열이면 성공 완료가 아니라 error 이벤트다', async () => {
    mockSTT.transcribe.mockResolvedValue({
      text: FX.STT_EMPTY,
      segments: [],
      language: 'ko',
      duration: 1,
      processingTime: 1,
    })
    const svc = await startAndRecord()
    const errors: Array<{ code: number }> = []
    let completed = false
    svc.on('error', (p: { error: { code: number } }) => errors.push(p.error))
    svc.on('session-completed', () => {
      completed = true
    })
    await svc.stopSession()
    expect(completed).toBe(false)
    expect(errors.some((e) => e.code === ErrorCode.STTNoAudioData)).toBe(true)
  })

  it('STT 가 공백만 반환해도 error 다', async () => {
    mockSTT.transcribe.mockResolvedValue({
      text: '   ',
      segments: [],
      language: 'ko',
      duration: 1,
      processingTime: 1,
    })
    const svc = await startAndRecord()
    const errors: number[] = []
    svc.on('error', (p: { error: { code: number } }) => errors.push(p.error.code))
    await svc.stopSession()
    expect(errors).toContain(ErrorCode.STTNoAudioData)
  })

  it('STT throw 는 STTTranscriptionFailed 로 표면화된다', async () => {
    mockSTT.transcribe.mockRejectedValue(new Error('fx.stt.down'))
    const svc = await startAndRecord()
    const errors: number[] = []
    svc.on('error', (p: { error: { code: number } }) => errors.push(p.error.code))
    await svc.stopSession()
    expect(errors).toContain(ErrorCode.STTTranscriptionFailed)
  })

  it('LLM 실패 시 원문으로 조용히 완료하지 않고 error 를 낸다', async () => {
    mockLLM.processText.mockRejectedValue(new Error('fx.llm.timeout'))
    const svc = await startAndRecord()
    const errors: number[] = []
    let completed = false
    svc.on('error', (p: { error: { code: number } }) => errors.push(p.error.code))
    svc.on('session-completed', () => {
      completed = true
    })
    await svc.stopSession()
    expect(completed).toBe(false)
    expect(errors.length).toBeGreaterThan(0)
  })

  it('LLM 액션 none 이면 STT 픽스처가 최종 텍스트다', async () => {
    configSet('defaultLLMAction', 'none')
    const svc = await startAndRecord()
    let finalText: string | null = null
    svc.on('session-completed', (p: { finalText: string }) => {
      finalText = p.finalText
    })
    await svc.stopSession()
    expect(finalText).toBe(FX.STT_OK)
    expect(mockLLM.processText).not.toHaveBeenCalled()
  })

  it('라이선스가 막으면 세션이 시작되지 않고 error 가 난다', async () => {
    mockLicense.canUse.mockReturnValue({ allowed: false, reason: 'quota_exceeded' })
    const svc = getVoiceModeService()
    const errors: unknown[] = []
    svc.on('error', (p) => errors.push(p))
    await svc.startSession('dictation')
    expect(svc.isActive).toBe(false)
    expect(errors.length).toBeGreaterThan(0)
  })

  it('자막 모드가 켜져 있으면 받아쓰기를 시작하지 않고 error 다', async () => {
    mockCaption.getState.mockReturnValue('active')
    const svc = getVoiceModeService()
    const errors: unknown[] = []
    svc.on('error', (p) => errors.push(p))
    await svc.startSession('dictation')
    expect(svc.isActive).toBe(false)
    expect(errors.length).toBeGreaterThan(0)
  })

  it('오디오 시작 실패는 AudioCaptureStartFailed 다', async () => {
    mockAudio.start.mockRejectedValueOnce(new Error('fx.mic.denied'))
    const svc = getVoiceModeService()
    const errors: number[] = []
    svc.on('error', (p: { error: { code: number } }) => errors.push(p.error.code))
    await svc.startSession('dictation')
    expect(errors).toContain(ErrorCode.AudioCaptureStartFailed)
  })

  it('중복 startSession 은 두 번째를 무시하고 기존 세션을 유지한다', async () => {
    const svc = getVoiceModeService()
    await svc.startSession('dictation')
    const first = svc.getState().sessionId
    await svc.startSession('dictation')
    expect(svc.getState().sessionId).toBe(first)
  })

  it('IPC voice:getState 는 현재 상태를 돌려준다', async () => {
    registerVoiceHandlers()
    const res = await invokeIpc(IPC_CHANNELS.VOICE.GET_STATE)
    expect(res.success).toBe(true)
    if (res.success) expect(res.data.recognitionState).toBe(RecognitionState.IDLE)
  })

  it('IPC voice:startRecording 성공 시 sessionId 가 있다', async () => {
    registerVoiceHandlers()
    const res = await invokeIpc(IPC_CHANNELS.VOICE.START_RECORDING, { sessionId: 'ui' })
    expect(res.success).toBe(true)
    if (res.success) expect(res.data.sessionId).toBeTruthy()
  })

  it('IPC voice:cancelRecording 은 활성 세션을 종료한다', async () => {
    registerVoiceHandlers()
    await invokeIpc(IPC_CHANNELS.VOICE.START_RECORDING, { sessionId: 'ui' })
    const res = await invokeIpc(IPC_CHANNELS.VOICE.CANCEL_RECORDING, { sessionId: 'ui' })
    expect(res.success).toBe(true)
    expect(getVoiceModeService().isActive).toBe(false)
  })

  it('hands-free 모드로도 세션을 시작할 수 있다', async () => {
    const svc = getVoiceModeService()
    await svc.startSession('hands-free')
    expect(svc.currentSession?.mode).toBe('hands-free')
    svc.cancelSession()
  })

  it('녹음 완료 → persistCompletedVoiceSession 이 히스토리 행을 만든다', async () => {
    const svc = await startAndRecord()
    svc.on('session-completed', ({ session, finalText }) => {
      persistCompletedVoiceSession(session, finalText)
    })
    await svc.stopSession()
    const page = getHistoryService().list({ page: 0, pageSize: 10 })
    expect(page.total).toBe(1)
    expect(page.entries[0].originalText === FX.STT_OK || page.entries[0].polishedText === FX.LLM_OK).toBe(
      true,
    )
    expect(page.entries[0].status).toBe('completed')
  })

  it('히스토리 저장 실패는 삼키지 않고 error 로 표면화된다', async () => {
    const svc = getVoiceModeService()
    const errors: unknown[] = []
    svc.on('error', (p) => errors.push(p))
    persistCompletedVoiceSessionSafe(
      (error, session) => {
        svc.emit('error', { error, session })
      },
      {
        id: 'fx-session',
        transcription: FX.STT_OK,
        processedText: null,
        mode: 'dictation',
        startedAt: Date.now(),
      },
      FX.STT_OK,
    )
    unbindTestDatabase()
    persistCompletedVoiceSessionSafe(
      (error, session) => {
        svc.emit('error', { error, session })
      },
      {
        id: 'fx-session-fail',
        transcription: FX.STT_OK,
        processedText: null,
        mode: 'dictation',
        startedAt: Date.now(),
      },
      FX.STT_OK,
    )
    expect(errors.length).toBeGreaterThan(0)
  })
})