feat: 实现 TTS 语音播放，完成 MVP P0 全部用户故事

- 新增 TTSPlayer：收集流式 tts_audio 片段，is_last 时拼接解码播放
- useVisionSession 接入 TTS 播放器，interrupt/stopSession 时停止播放
- App 显示 '🔊 正在播放...' 指示器
- MVP P0 四个用户故事前端全部实现（US-01 ~ US-04）

Co-Authored-By: Claude <noreply@anthropic.com>

frontend/src/App.css

@@ -112,6 +112,12 @@ body {
   animation: none;
 }
 
+.vad-indicator--audio {
+  left: auto;
+  right: 12px;
+  color: var(--color-primary);
+}
+
 @keyframes pulse {
   0%, 100% { opacity: 1; }
   50% { opacity: 0.6; }

frontend/src/App.tsx

@@ -13,6 +13,7 @@ function App() {
     messages,
     currentReply,
     isProcessing,
+    isAudioPlaying,
     isSpeaking,
     isVADReady,
     vadError,
@@ -43,6 +44,7 @@ function App() {
         <div className="video-section">
           <VideoPreview ref={videoRef} isStreaming={!!stream} />
           {isSpeaking && <div className="vad-indicator">🎤 正在聆听...</div>}
+          {isAudioPlaying && <div className="vad-indicator vad-indicator--audio">🔊 正在播放...</div>}
           {isConnected && !isVADReady && !vadError && (
             <div className="vad-indicator vad-indicator--loading">
               正在初始化语音检测...

frontend/src/hooks/useVisionSession.ts

@@ -9,6 +9,7 @@ import { v4 as uuidv4 } from "uuid";
 import { wsClient } from "../lib/websocket";
 import { encodeAudioToBase64, dataUrlToBase64 } from "../lib/audio";
 import { getErrorMessage } from "../lib/errors";
+import { TTSPlayer } from "../lib/ttsPlayer";
 import { showToast } from "../lib/toast";
 import { useCamera } from "../components/CameraManager";
 import { useMicrophone } from "../components/MicManager";
@@ -20,6 +21,18 @@ export function useVisionSession() {
   const [messages, setMessages] = useState<ChatMessage[]>([]);
   const [currentReply, setCurrentReply] = useState<string>("");
   const [isProcessing, setIsProcessing] = useState(false);
+  const [isAudioPlaying, setIsAudioPlaying] = useState(false);
+
+  // TTS 播放器
+  const ttsPlayerRef = useRef<TTSPlayer | null>(null);
+  const getTTSPlayer = useCallback(() => {
+    if (!ttsPlayerRef.current) {
+      const player = new TTSPlayer();
+      player.onEnd(() => setIsAudioPlaying(false));
+      ttsPlayerRef.current = player;
+    }
+    return ttsPlayerRef.current;
+  }, []);
 
   const { videoRef, captureFrame, startCamera, stopCamera, stream } = useCamera();
   const { startMic, stopMic } = useMicrophone();
@@ -115,7 +128,10 @@ export function useVisionSession() {
         }
 
         case "tts_audio":
-          // TODO: 阶段 5 音频流播放
+          getTTSPlayer().enqueue(msg.audio, msg.mime_type, msg.is_last);
+          if (!msg.is_last) {
+            setIsAudioPlaying(true);
+          }
           break;
 
         case "error":
@@ -127,7 +143,7 @@ export function useVisionSession() {
     });
 
     return unsub;
-  }, []);
+  }, [getTTSPlayer]);
 
   // 连接断开时显示提示
   useEffect(() => {
@@ -160,7 +176,9 @@ export function useVisionSession() {
     stopMic();
     stopCamera();
     disconnect();
-    // 清理所有对话状态
+    // 停止 TTS 并清理状态
+    ttsPlayerRef.current?.stop();
+    setIsAudioPlaying(false);
     setMessages([]);
     setCurrentReply("");
     setIsProcessing(false);
@@ -169,6 +187,9 @@ export function useVisionSession() {
   /** 打断当前回复 */
   const interrupt = useCallback(() => {
     send({ type: "interrupt" });
+    // 停止 TTS 播放
+    ttsPlayerRef.current?.stop();
+    setIsAudioPlaying(false);
     // 将未完成的流式内容保存为最终消息
     if (currentReply) {
       setMessages((prev) => [
@@ -184,6 +205,7 @@ export function useVisionSession() {
     messages,
     currentReply,
     isProcessing,
+    isAudioPlaying,
     isSpeaking,
     isVADReady,
     vadError,

frontend/src/lib/ttsPlayer.ts

@@ -0,0 +1,104 @@
+// ============================================================
+// TTS Player — 语音播放器
+// 职责：收集后端流式 tts_audio 片段，拼接后播放
+// 格式：MVP 仅支持 audio/mp3，pcm 为 TODO
+// ============================================================
+
+type OnEndCallback = () => void;
+
+export class TTSPlayer {
+  private chunks: string[] = [];
+  private audio: HTMLAudioElement | null = null;
+  private _isPlaying = false;
+  private onEndCallback: OnEndCallback | null = null;
+
+  /** 注册播放完成回调 */
+  onEnd(cb: OnEndCallback): void {
+    this.onEndCallback = cb;
+  }
+
+  /** 当前是否正在播放 */
+  get isPlaying(): boolean {
+    return this._isPlaying;
+  }
+
+  /**
+   * 入队一个 TTS 音频片段
+   * @param base64 Base64 编码的音频数据
+   * @param mimeType 音频格式（"audio/mp3" 或 "audio/pcm"）
+   * @param isLast 是否为最后一个片段
+   */
+  enqueue(base64: string, mimeType: string, isLast: boolean): void {
+    this.chunks.push(base64);
+
+    if (isLast) {
+      this.play(mimeType);
+    }
+  }
+
+  /** 停止播放并清空缓冲区 */
+  stop(): void {
+    if (this.audio) {
+      this.audio.pause();
+      this.audio.removeAttribute("src");
+      this.audio = null;
+    }
+    this.chunks = [];
+    this._isPlaying = false;
+  }
+
+  /** 暂停播放 */
+  pause(): void {
+    this.audio?.pause();
+  }
+
+  /** 恢复播放 */
+  resume(): void {
+    this.audio?.play();
+  }
+
+  /** 拼接所有片段并播放 */
+  private play(mimeType: string): void {
+    if (this.chunks.length === 0) return;
+
+    // 拼接所有 Base64 片段
+    const combined = this.chunks.join("");
+    this.chunks = [];
+
+    // Base64 → Uint8Array → Blob
+    const binary = atob(combined);
+    const bytes = new Uint8Array(binary.length);
+    for (let i = 0; i < binary.length; i++) {
+      bytes[i] = binary.charCodeAt(i);
+    }
+    const blob = new Blob([bytes], { type: mimeType });
+    const url = URL.createObjectURL(blob);
+
+    // 播放
+    const audio = new Audio(url);
+    this.audio = audio;
+    this._isPlaying = true;
+
+    audio.onended = () => {
+      URL.revokeObjectURL(url);
+      this._isPlaying = false;
+      this.audio = null;
+      this.onEndCallback?.();
+    };
+
+    audio.onerror = () => {
+      console.error("[TTS] 播放失败");
+      URL.revokeObjectURL(url);
+      this._isPlaying = false;
+      this.audio = null;
+      this.onEndCallback?.();
+    };
+
+    audio.play().catch((err) => {
+      console.error("[TTS] play() 被拒绝:", err);
+      this._isPlaying = false;
+      this.audio = null;
+      this.onEndCallback?.();
+    });
+  }
+}