fix(coze/api): fix speech bug

common/changes/@coze/api/fix-speech_2025-03-12-13-51.json

@@ -0,0 +1,11 @@
+{
+  "changes": [
+    {
+      "packageName": "@coze/api",
+      "comment": "fix speech bug",
+      "type": "patch"
+    }
+  ],
+  "packageName": "@coze/api",
+  "email": "[email protected]"
+}

common/changes/@coze/api/fix-speech_2025-03-14-03-48.json

@@ -0,0 +1,11 @@
+{
+  "changes": [
+    {
+      "packageName": "@coze/api",
+      "comment": "fix speech bug",
+      "type": "patch"
+    }
+  ],
+  "packageName": "@coze/api",
+  "email": "[email protected]"
+}

common/changes/@coze/api/fix-speech_2025-03-14-04-06.json

@@ -0,0 +1,11 @@
+{
+  "changes": [
+    {
+      "packageName": "@coze/api",
+      "comment": "Publish fix/speech",
+      "type": "patch"
+    }
+  ],
+  "packageName": "@coze/api",
+  "email": "[email protected]"
+}

common/changes/@coze/api/fix-speech_2025-03-14-04-11.json

@@ -0,0 +1,11 @@
+{
+  "changes": [
+    {
+      "packageName": "@coze/api",
+      "comment": "Publish fix/speech",
+      "type": "patch"
+    }
+  ],
+  "packageName": "@coze/api",
+  "email": "[email protected]"
+}

examples/coze-js-web/src/pages/chat-x/use-ws-api.ts

@@ -306,34 +306,39 @@ const useWsAPI = (
       return;
     }
 
-    const client = new WsSpeechClient({
-      token: config.getPat(),
-      baseWsURL: config.getBaseWsUrl(),
-      allowPersonalAccessTokenInBrowser: true,
-    });
+    if (!speechClientRef.current) {
+      const client = new WsSpeechClient({
+        token: config.getPat(),
+        baseWsURL: config.getBaseWsUrl(),
+        allowPersonalAccessTokenInBrowser: true,
+      });
 
-    client.on('data', data => {
-      console.log('[speech] ws data', data);
-    });
+      client.on('data', data => {
+        console.log('[speech] ws data', data);
+      });
 
-    client.on(WebsocketsEventType.ERROR, data => {
-      console.error('[speech] ws error', data);
-    });
+      client.on(WebsocketsEventType.ERROR, data => {
+        console.error('[speech] ws error', data);
+      });
+      client.on('completed', () => {
+        console.log('[speech] speech completed');
+      });
+      speechClientRef.current = client;
+    }
 
     try {
-      await client.connect();
+      await speechClientRef.current?.connect();
       console.log('[speech] ws connect success');
     } catch (error) {
       console.error('[speech] ws connect error', error);
       return;
     }
 
-    client.appendAndComplete(message);
-    speechClientRef.current = client;
+    speechClientRef.current?.appendAndComplete(message);
   }, []);
 
   const stopSpeech = () => {
-    speechClientRef.current?.interrupt();
+    speechClientRef.current?.disconnect();
   };
 
   const getIsSpeech = () => speechClientRef.current?.isPlaying();

packages/coze-js/README.md

@@ -200,7 +200,7 @@ client.on(WebsocketsEventType.ERROR, data => {
   console.error('[speech] ws error', data);
 });
 
-// Listen for playback completed event
+// Listen for playback completed event, if manually called disconnect, this event will not be triggered
 client.on('completed', () => {
   console.log('[speech] playback completed');
 });
@@ -215,13 +215,11 @@ try {
 }
 
 // Send message and play
-client.appendAndComplete(message);
+client.appendAndComplete('Hello, Coze!');
 
 // Interrupt
 client.interrupt();
 
-// Disconnect, destroy instance
-client.disconnect();
 
 // Pause speech playback
 client.pause();
@@ -235,8 +233,12 @@ client.togglePlay();
 // Check if speech is playing
 client.isPlaying();
 
+// Disconnect, destroy instance
+client.disconnect();
+
 // Send text fragment
-client.append(message);
+client.append('Hello,');
+client.append(' Coze!');
 // End sending text
 client.complete();
 

packages/coze-js/README.zh-CN.md

@@ -199,7 +199,7 @@ client.on(WebsocketsEventType.ERROR, data => {
   console.error('[speech] ws error', data);
 });
 
-// 监听播放完成事件
+// 监听播放完成事件，如果手动调用disconnect，则不会触发此事件
 client.on('completed', () => {
   console.log('[speech] playback completed');
 });
@@ -214,13 +214,11 @@ try {
 }
 
 // 发送消息并播放
-client.appendAndComplete(message);
+client.appendAndComplete('你好，Coze!');
 
 // 打断
 client.interrupt();
 
-// 断开连接，销毁实例
-client.disconnect();
 
 // 暂停语音播放
 client.pause();
@@ -234,8 +232,12 @@ client.togglePlay();
 // 语音是否播放中
 client.isPlaying();
 
+// 断开连接，销毁 websocket
+client.disconnect();
+
 // 发送文本片段
-client.append(message);
+client.append('你好，');
+client.append(' Coze!');
 // 结束发送文本
 client.complete();
 ```

packages/coze-js/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@coze/api",
-  "version": "1.1.0-beta.2",
+  "version": "1.1.0-beta.4",
   "description": "Official Coze Node.js SDK for seamless AI integration into your applications | 扣子官方 Node.js SDK，助您轻松集成 AI 能力到应用中",
   "keywords": [
     "coze",

packages/coze-js/src/ws-tools/speech/index.ts

@@ -88,6 +88,22 @@
             data.event_type === WebsocketsEventType.SPEECH_AUDIO_COMPLETED
           ) {
             console.debug('[speech] totalDuration', this.totalDuration);
+
+            if (this.playbackStartTime) {
+              // 剩余时间 = 总时间 - 已播放时间 - 已暂停时间
+              const now = new Date().getTime();
+              const remaining =
+                this.totalDuration -
+                (now - this.playbackStartTime) / 1000 -
+                this.elapsedBeforePause;
+
+              this.playbackTimeout = setTimeout(() => {
+                this.emit('completed', undefined);
+                this.playbackStartTime = null;
+                this.elapsedBeforePause = 0;
+              }, remaining * 1000);
+            }
+
             this.closeWs();
           }
         };
@@ -135,8 +151,10 @@
   }
 
   async disconnect() {
-    await this.interrupt();
-    this.listeners.clear();
+    if (this.playbackTimeout) {
+      clearTimeout(this.playbackTimeout);
+    }
+    await this.wavStreamPlayer.interrupt();
     this.closeWs();
   }
 
@@ -257,26 +275,6 @@
         this.playbackStartTime = Date.now();
         this.elapsedBeforePause = 0;
       }
-
-      // Clear existing timeout if any
-      if (this.playbackTimeout) {
-        clearTimeout(this.playbackTimeout);
-      }
-
-      // Calculate remaining time
-      const elapsed =
-        this.elapsedBeforePause +
-        (this.playbackPauseTime
-          ? 0
-          : (Date.now() - (this.playbackStartTime || Date.now())) / 1000);
-      const remaining = this.totalDuration - elapsed;
-
-      // Set new timeout
-      this.playbackTimeout = setTimeout(() => {
-        this.emit('completed', undefined);
-        this.playbackStartTime = null;
-        this.elapsedBeforePause = 0;
-      }, remaining * 1000);
     } catch (error) {
       console.warn('[speech] wavStreamPlayer error', error);
     }

packages/coze-js/test/ws-tools/speech.spec.ts

@@ -93,7 +93,7 @@ describe('WsSpeechClient', () => {
         },
         undefined as unknown as MessageEvent,
       );
-
+      await new Promise(resolve => setTimeout(resolve, 50));
       client.ws?.onmessage?.(
         {
           event_type: WebsocketsEventType.SPEECH_AUDIO_COMPLETED,