支持豆包流式语音输入 (#377)

Author: ErlichLiu

apps/electron/electron-builder.yml

@@ -67,6 +67,8 @@ mac:
   gatekeeperAssess: false
   entitlements: resources/entitlements.mac.plist
   entitlementsInherit: resources/entitlements.mac.plist
+  extendInfo:
+    NSMicrophoneUsageDescription: "Proma 需要访问麦克风，用于将你的语音实时转写为文本。"
   # 架构由 CLI 参数控制（CI 矩阵分别传 --arm64 / --x64），
   # 每个 runner 只构建其宿主架构对应的产物，与 bun 按 os/cpu 装好的
   # @anthropic-ai/claude-agent-sdk-darwin-{arch} native binary 对齐。

apps/electron/package.json

@@ -120,6 +120,7 @@
     "use-stick-to-bottom": "^1.1.2",
     "vite": "^6.0.3",
     "word-extractor": "1.0.4",
+    "ws": "8.19.0",
     "zod": "^4.0.0"
   },
   "electronmon": {

apps/electron/src/main/index.ts

@@ -57,6 +57,12 @@ import { getDingTalkMultiBotConfig } from './lib/dingtalk-config'
 import { wechatBridge } from './lib/wechat-bridge'
 import { getWeChatConfig } from './lib/wechat-config'
 import { createQuickTaskWindow, toggleQuickTaskWindow, destroyQuickTaskWindow } from './lib/quick-task-window'
+import {
+  createVoiceDictationWindow,
+  toggleVoiceDictationWindow,
+  destroyVoiceDictationWindow,
+  shouldSuppressVoiceDictationActivate,
+} from './lib/voice-dictation-window'
 import { registerGlobalShortcut, unregisterAllGlobalShortcuts } from './lib/global-shortcut-service'
 import { TRAY_IPC_CHANNELS } from '../types'
 
@@ -341,15 +347,23 @@ app.whenReady().then(async () => {
 
   // 预创建快速任务窗口（隐藏状态，首次唤起秒开）
   createQuickTaskWindow()
+  createVoiceDictationWindow()
 
   // 注册全局快捷键
   registerGlobalShortcut('quick-task', toggleQuickTaskWindow)
   registerGlobalShortcut('show-main-window', showAndFocusMainWindow)
+  registerGlobalShortcut('voice-dictation', () => {
+    toggleVoiceDictationWindow({ targetIsProma: mainWindow?.isFocused() === true })
+  })
 
   // 启动所有已注册的 Bridge（飞书/钉钉/微信等）
   await startAllBridges()
 
   app.on('activate', () => {
+    if (shouldSuppressVoiceDictationActivate()) {
+      return
+    }
+
     // 直接检查 mainWindow 引用，避免 getAllWindows() 包含 DevTools 等其他窗口导致误判
     if (!mainWindow || mainWindow.isDestroyed()) {
       createWindow()
@@ -390,6 +404,7 @@ app.on('before-quit', () => {
   unregisterAllGlobalShortcuts()
   // 销毁快速任务窗口
   destroyQuickTaskWindow()
+  destroyVoiceDictationWindow()
   // Clean up system tray before quitting
   destroyTray()
 })

apps/electron/src/main/ipc.ts

@@ -8,8 +8,19 @@ import { ipcMain, nativeTheme, shell, dialog, BrowserWindow, app } from 'electro
 import { join } from 'node:path'
 import { existsSync } from 'node:fs'
 import { IPC_CHANNELS, CHANNEL_IPC_CHANNELS, CHAT_IPC_CHANNELS, AGENT_IPC_CHANNELS, ENVIRONMENT_IPC_CHANNELS, INSTALLER_IPC_CHANNELS, PROXY_IPC_CHANNELS, GITHUB_RELEASE_IPC_CHANNELS, SYSTEM_PROMPT_IPC_CHANNELS, MEMORY_IPC_CHANNELS, CHAT_TOOL_IPC_CHANNELS, FEISHU_IPC_CHANNELS, DINGTALK_IPC_CHANNELS, WECHAT_IPC_CHANNELS } from '@proma/shared'
-import { USER_PROFILE_IPC_CHANNELS, SETTINGS_IPC_CHANNELS, QUICK_TASK_IPC_CHANNELS, APP_ICON_IPC_CHANNELS, DOCK_BADGE_IPC_CHANNELS } from '../types'
-import type { QuickTaskSubmitInput } from '../types'
+import { USER_PROFILE_IPC_CHANNELS, SETTINGS_IPC_CHANNELS, QUICK_TASK_IPC_CHANNELS, VOICE_DICTATION_IPC_CHANNELS, APP_ICON_IPC_CHANNELS, DOCK_BADGE_IPC_CHANNELS } from '../types'
+import type {
+  QuickTaskSubmitInput,
+  VoiceDictationAudioChunkInput,
+  VoiceDictationCommitInput,
+  VoiceDictationCommitResult,
+  VoiceDictationResizeInput,
+  VoiceDictationSettings,
+  VoiceDictationSettingsUpdate,
+  VoiceDictationStartInput,
+  VoiceDictationStopInput,
+  VoiceDictationTestResult,
+} from '../types'
 import type {
   RuntimeStatus,
   GitRepoStatus,
@@ -2607,4 +2618,101 @@ export function registerIpcHandlers(): void {
       return reregisterAllGlobalShortcuts()
     }
   )
+
+  // ===== 语音输入 =====
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.GET_SETTINGS,
+    async (): Promise<VoiceDictationSettings> => {
+      const { getVoiceDictationSettings } = await import('./lib/voice-dictation-settings-service')
+      return getVoiceDictationSettings()
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.UPDATE_SETTINGS,
+    async (_, updates: VoiceDictationSettingsUpdate): Promise<VoiceDictationSettings> => {
+      const { updateVoiceDictationSettings } = await import('./lib/voice-dictation-settings-service')
+      return updateVoiceDictationSettings(updates)
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.TEST_CONNECTION,
+    async (_, updates?: VoiceDictationSettingsUpdate): Promise<VoiceDictationTestResult> => {
+      const { getVoiceDictationSettings } = await import('./lib/voice-dictation-settings-service')
+      const { testDoubaoAsrConnection } = await import('./lib/doubao-asr-service')
+      const settings = { ...getVoiceDictationSettings(), ...(updates ?? {}) }
+      return testDoubaoAsrConnection(settings)
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.TOGGLE,
+    async (event): Promise<void> => {
+      const { toggleVoiceDictationWindow } = await import('./lib/voice-dictation-window')
+      const sourceWindow = BrowserWindow.fromWebContents(event.sender)
+      toggleVoiceDictationWindow({ targetIsProma: !!sourceWindow })
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.START,
+    async (event, input: VoiceDictationStartInput): Promise<void> => {
+      const { getVoiceDictationSettings } = await import('./lib/voice-dictation-settings-service')
+      const { startDoubaoAsrSession } = await import('./lib/doubao-asr-service')
+      const win = BrowserWindow.fromWebContents(event.sender)
+      if (!win) throw new Error('语音输入窗口不存在')
+      await startDoubaoAsrSession(input.sessionId, getVoiceDictationSettings(), win)
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.SEND_AUDIO,
+    async (_, input: VoiceDictationAudioChunkInput): Promise<void> => {
+      const { sendDoubaoAsrAudio } = await import('./lib/doubao-asr-service')
+      sendDoubaoAsrAudio(input.sessionId, input.data)
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.STOP,
+    async (_, input: VoiceDictationStopInput): Promise<void> => {
+      const { stopDoubaoAsrSession } = await import('./lib/doubao-asr-service')
+      await stopDoubaoAsrSession(input.sessionId)
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.CANCEL,
+    async (_, input: VoiceDictationStopInput): Promise<void> => {
+      const { cancelDoubaoAsrSession } = await import('./lib/doubao-asr-service')
+      cancelDoubaoAsrSession(input.sessionId)
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.COMMIT,
+    async (_, input: VoiceDictationCommitInput): Promise<VoiceDictationCommitResult> => {
+      const { getVoiceDictationSettings } = await import('./lib/voice-dictation-settings-service')
+      const { commitVoiceDictationText } = await import('./lib/text-output-service')
+      return commitVoiceDictationText(input.text, getVoiceDictationSettings())
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.HIDE,
+    async (): Promise<void> => {
+      const { hideVoiceDictationWindow } = await import('./lib/voice-dictation-window')
+      hideVoiceDictationWindow()
+    }
+  )
+
+  ipcMain.handle(
+    VOICE_DICTATION_IPC_CHANNELS.RESIZE,
+    async (_, input: VoiceDictationResizeInput): Promise<void> => {
+      const { resizeVoiceDictationWindow } = await import('./lib/voice-dictation-window')
+      resizeVoiceDictationWindow(input.height)
+    }
+  )
 }