Files
Cyrene-For-Android/devdocs/01-voice-assistant-system.md
T

257 lines
8.1 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
# 01 — 系统语音助手集成规范
> **目标**:让昔涟成为 Android 系统级默认语音助手,替换 Google Assistant / Bixby
> **核心 API**`VoiceInteractionService` + `VoiceInteractionSession`
---
## 1. 功能目标
- 用户可在 **系统设置 → 默认应用 → 数字助理** 中选择昔涟
- 长按 Home 键呼出昔涟(非全屏,悬浮覆盖层)
- 屏幕底部两角向内滑动触发昔涟
- 长按电源键可配置为呼出昔涟
- 息屏状态下热词唤醒昔涟
- 有线/蓝牙耳机按键呼出昔涟
## 2. AndroidManifest.xml 声明
```xml
<!-- VoiceInteractionService -->
<service
android:name=".service.CyreneVoiceInteractionService"
android:exported="true"
android:permission="android.permission.BIND_VOICE_INTERACTION">
<meta-data
android:name="android.voice_interaction"
android:resource="@xml/voice_interaction_config" />
<intent-filter>
<action android:name="android.service.voice.VoiceInteractionService" />
</intent-filter>
</service>
<!-- AssistService (Android 14+) -->
<service
android:name=".service.CyreneAssistService"
android:exported="true"
android:permission="android.permission.BIND_ASSIST">
<intent-filter>
<action android:name="android.service.voice.AssistService" />
</intent-filter>
</service>
```
## 3. 配置文件
### res/xml/voice_interaction_config.xml
```xml
<?xml version="1.0" encoding="utf-8"?>
<voice-interaction-service
xmlns:android="http://schemas.android.com/apk/res/android"
android:sessionService=".service.CyreneVoiceInteractionSession"
android:recognitionService=".service.CyreneRecognitionService"
android:supportsAssist="true"
android:supportsLaunchVoiceAssistFromKeyguard="true"
android:supportsLocalRecognition="true"
android:serviceIcon="@drawable/ic_cyrene"
android:serviceLabel="@string/voice_assistant_name" />
```
## 4. VoiceInteractionService 实现
```kotlin
class CyreneVoiceInteractionService : VoiceInteractionService() {
override fun onReady() {
super.onReady()
// 服务就绪,可在此初始化 TTS 引擎等
}
override fun onCreateSession(args: Bundle?): VoiceInteractionSession {
return CyreneVoiceInteractionSession(this)
}
override fun onLaunchVoiceAssistFromKeyguard() {
// 锁屏启动 → 进入简化模式,仅显示对话,IoT 控制等需先解锁
}
// Android 14+: AssistAction 回调
override fun onHandleAssist(
request: AssistRequest?,
cancellationSignal: CancellationSignal?,
callback: OutcomeCallback<AssistResult?>?
) {
if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.UPSIDE_DOWN_CAKE) {
request?.let {
val assistContent = it.assistContent
// 提取当前屏幕上下文(可选,用于后续上下文感知)
callback?.onResult(AssistResult(assistContent))
}
}
}
}
```
## 5. VoiceInteractionSession 实现(悬浮窗界面)
```kotlin
class CyreneVoiceInteractionSession(context: Context) :
VoiceInteractionSession(context) {
override fun onCreateContentView(): View {
// 返回 ComposeView 作为悬浮窗的内容
return ComposeView(context).apply {
setContent {
CyreneTheme {
OverlayScreen(
viewModel = overlayViewModel,
onDismiss = { finish() }
)
}
}
}
}
override fun onShow(args: Bundle?, showFlags: Int) {
super.onShow(args, showFlags)
// 设置窗口属性:透明背景 + 底部卡片式布局
window?.apply {
// 半透明遮罩
setBackgroundDrawable(ColorDrawable(0x80000000.toInt()))
// FLAG_DIM_BEHIND 可实现模糊效果
if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.S) {
addSystemGestureExclusionRects(...)
}
}
}
override fun onComputeInsets(outInsets: Insets?) {
super.onComputeInsets(outInsets)
// 控制悬浮窗内容区域
}
override fun onHide() {
super.onHide()
// 悬浮窗隐藏时清理状态
}
}
```
### 关键窗口属性
| 属性 | 值 | 说明 |
|------|-----|------|
| 背景 | `ColorDrawable(0x80000000)` | 半透明黑色遮罩,透出底层 APP |
| 内容区域 | 自适应高度 | 底部弹出,类似 Google Assistant |
| 触摸外区域行为 | 关闭悬浮窗 | 用户点击遮罩区域关闭 |
| 键盘弹出 | 推高内容区域 | 文本输入时自动调整 |
## 6. 权限清单
```xml
<!-- 核心语音助手权限 -->
<uses-permission android:name="android.permission.BIND_VOICE_INTERACTION" />
<uses-permission android:name="android.permission.BIND_ASSIST" />
<!-- 音频相关 -->
<uses-permission android:name="android.permission.RECORD_AUDIO" />
<uses-permission android:name="android.permission.MODIFY_AUDIO_SETTINGS" />
<!-- 热词唤醒 -->
<uses-permission android:name="android.permission.CAPTURE_AUDIO_HOTWORD" />
<!-- 后台服务 -->
<uses-permission android:name="android.permission.FOREGROUND_SERVICE" />
<uses-permission android:name="android.permission.FOREGROUND_SERVICE_MICROPHONE" />
<!-- 网络 -->
<uses-permission android:name="android.permission.INTERNET" />
<uses-permission android:name="android.permission.ACCESS_NETWORK_STATE" />
<!-- 推送 -->
<uses-permission android:name="android.permission.POST_NOTIFICATIONS" />
<!-- 锁屏交互 -->
<uses-permission android:name="android.permission.USE_FULL_SCREEN_INTENT" />
```
## 7. 引导用户设为默认助手
首次启动时检测并引导:
```kotlin
fun checkAndPromptDefaultAssistant(context: Context) {
val isDefault = if (Build.VERSION.SDK_INT >= Build.VERSION_CODES.Q) {
val componentName = ComponentName(context, CyreneVoiceInteractionService::class.java)
context.packageManager
.queryIntentServices(
Intent(VoiceInteractionService.SERVICE_INTERFACE),
PackageManager.MATCH_DEFAULT_ONLY
)
.any { it.serviceInfo.packageName == context.packageName }
} else {
false
}
if (!isDefault) {
// 显示引导 UI → 跳转到 Settings.ACTION_VOICE_INPUT_SETTINGS
val intent = Intent(Settings.ACTION_VOICE_INPUT_SETTINGS)
context.startActivity(intent)
}
}
```
## 8. 热词唤醒检测
### 方案选型
| 方案 | 优点 | 缺点 | 适用场景 |
|------|------|------|---------|
| 系统 Always-On Hotword API | 低功耗、系统级支持 | 限 Android 8+,某些 ROM 不支持 | **首选** |
| Porcupine (Picovoice) | 跨平台、离线 | 商业许可,需额外集成 | 兜底 |
| 自建模型 (openWakeWord) | 完全可控、低成本 | 需要本地推理能力 | 长期方案 |
### 唤醒词配置
| 优先级 | 唤醒词 | 说明 |
|--------|--------|------|
| P0 | "昔涟" (Xī Lián) | 角色名,默认唤醒词 |
| P1 | "Hey 昔涟" | 与 "Hey Google" 习惯对齐 |
| P2 | 自定义 | 用户可在设置中自定义 |
### 息屏唤醒流程
```
用户说出唤醒词
→ HotwordDetector 识别成功(<800ms
→ 系统触发 VoiceInteractionService
→ CyreneVoiceInteractionSession.onCreateContentView()
→ Overlay 显示,播放连接提示音
→ 用户说话 → STT → AI-Core → TTS → 语音回复
→ 对话结束 → finish() → 息屏
```
## 9. Dismiss 时机
悬浮窗在以下情况关闭:
| 条件 | 行为 |
|------|------|
| 用户说"再见" / "退下" | 自然对话结束,收起悬浮窗 |
| 用户点击遮罩区域 | 立即关闭 |
| 对话静默 10 秒 | 自动收起 |
| 用户主动滑动关闭 | 手势关闭,同 Google Assistant |
| 收到系统电话等中断 | 暂停语音,进入后台等待 |
## 10. 降级策略
当系统不支持 `VoiceInteractionService` 或未设为默认助手时:
- **保底方案**:PWA(利用主项目已有的 PWA 支持)
- **WebView 封装**:内嵌 H5 对话界面作为过渡
- **通知栏常驻**:提供快速对话入口,但功能受限