STT 数据流处理

vetti-admin/src/main/java/com/vetti/socket/ChatWebSocketHandler.java

@@ -198,55 +198,6 @@ public class ChatWebSocketHandler {
         }
     }
 
-    // 接收二进制消息（流数据）
-//    @OnMessage
-//    public void onBinaryMessage(Session session, @PathParam("clientId") String clientId, ByteBuffer byteBuffer) {
-//        log.info("1、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
-//        log.info("客户端ID为:{}", clientId);
-//        // 处理二进制流数据
-//        byte[] bytes = new byte[byteBuffer.remaining()];
-//        //从缓冲区中读取数据并存储到指定的字节数组中
-//        byteBuffer.get(bytes);
-//        log.info("2、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
-//        // 生成唯一文件名
-//        String fileName = clientId + "_" + System.currentTimeMillis() + ".wav";
-//        String pathUrl = RuoYiConfig.getProfile()+VOICE_STORAGE_DIR + fileName;
-//        log.info("文件路径为:{}", pathUrl);
-//        log.info("3、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
-//        try{
-//            log.info("文件流的大小为:{}",bytes.length);
-//            saveAsWebM(bytes,pathUrl);
-//            //接收到数据流后直接就进行SST处理
-//            //语音格式转换
-//            String fileOutName = clientId + "_" + System.currentTimeMillis() + ".pcm";
-//            String pathOutUrl = RuoYiConfig.getProfile()+VOICE_STORAGE_DIR + fileOutName;
-//            handleAudioToPCM(pathUrl,pathOutUrl);
-//            //发送消息
-//            WebSocket webSocket = cacheWebSocket.get(clientId);
-//            log.info("获取的socket对象为:{}",webSocket);
-//            if(webSocket != null){
-////                 1. 启动音频缓冲
-////                webSocket.send("{\"type\": \"input_audio_buffer.start\"}");
-//                log.info("3.1 开始发送数据音频流啦");
-//                File outputFile = new File(pathOutUrl);  // 输出PCM格式文件
-//                ByteBuffer buffer = ByteBuffer.wrap(FileUtils.readFileToByteArray(outputFile));
-//                byte[] outBytes = new byte[buffer.remaining()];
-//                //从缓冲区中读取数据并存储到指定的字节数组中
-//                buffer.get(outBytes);
-//                String base64Audio = Base64.getEncoder().encodeToString(outBytes);
-//                String message = "{ \"type\": \"input_audio_buffer.append\", \"audio\": \"" + base64Audio + "\" }";
-//                webSocket.send(message);
-//                log.info("4、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
-//                // 3. 提交音频并请求转录
-////                webSocket.send("{\"type\": \"input_audio_buffer.commit\"}");
-////                webSocket.send("{\"type\": \"response.create\"}");
-//            }
-//        }catch (Exception e){
-//            e.printStackTrace();
-//        }
-//
-//    }
-
     // 接收二进制消息（流数据）
     @OnMessage
     public void onBinaryMessage(Session session, @PathParam("clientId") String clientId, ByteBuffer byteBuffer) {
@@ -256,17 +207,66 @@ public class ChatWebSocketHandler {
         byte[] bytes = new byte[byteBuffer.remaining()];
         //从缓冲区中读取数据并存储到指定的字节数组中
         byteBuffer.get(bytes);
-
+        log.info("2、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
-        // 1. 获取当前会话的缓存
+        // 生成唯一文件名
-        List<byte[]> fragments = fragmentCache.get(clientId);
+        String fileName = clientId + "_" + System.currentTimeMillis() + ".wav";
-        if (fragments == null) {
+        String pathUrl = RuoYiConfig.getProfile()+VOICE_STORAGE_DIR + fileName;
-            fragments = new ArrayList<>();
+        log.info("文件路径为:{}", pathUrl);
-            fragmentCache.put(clientId, fragments);
+        log.info("3、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
+        try{
+            log.info("文件流的大小为:{}",bytes.length);
+            saveAsWebM(bytes,pathUrl);
+            //接收到数据流后直接就进行SST处理
+            //语音格式转换
+            String fileOutName = clientId + "_" + System.currentTimeMillis() + ".pcm";
+            String pathOutUrl = RuoYiConfig.getProfile()+VOICE_STORAGE_DIR + fileOutName;
+            handleAudioToPCM(pathUrl,pathOutUrl);
+            //发送消息
+            WebSocket webSocket = cacheWebSocket.get(clientId);
+            log.info("获取的socket对象为:{}",webSocket);
+            if(webSocket != null){
+//                 1. 启动音频缓冲
+//                webSocket.send("{\"type\": \"input_audio_buffer.start\"}");
+                log.info("3.1 开始发送数据音频流啦");
+                File outputFile = new File(pathOutUrl);  // 输出PCM格式文件
+                ByteBuffer buffer = ByteBuffer.wrap(FileUtils.readFileToByteArray(outputFile));
+                byte[] outBytes = new byte[buffer.remaining()];
+                //从缓冲区中读取数据并存储到指定的字节数组中
+                buffer.get(outBytes);
+                String base64Audio = Base64.getEncoder().encodeToString(outBytes);
+                String message = "{ \"type\": \"input_audio_buffer.append\", \"audio\": \"" + base64Audio + "\" }";
+                webSocket.send(message);
+                log.info("4、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
+                // 3. 提交音频并请求转录
+//                webSocket.send("{\"type\": \"input_audio_buffer.commit\"}");
+//                webSocket.send("{\"type\": \"response.create\"}");
+            }
+        }catch (Exception e){
+            e.printStackTrace();
         }
-        fragments.add(bytes);
+
-        fragmentCache.put(clientId, fragments);
     }
 
+    // 接收二进制消息（流数据）
+//    @OnMessage
+//    public void onBinaryMessage(Session session, @PathParam("clientId") String clientId, ByteBuffer byteBuffer) {
+//        log.info("1、开始接收数据流时间:{}",System.currentTimeMillis()/1000);
+//        log.info("客户端ID为:{}", clientId);
+//        // 处理二进制流数据
+//        byte[] bytes = new byte[byteBuffer.remaining()];
+//        //从缓冲区中读取数据并存储到指定的字节数组中
+//        byteBuffer.get(bytes);
+//
+//        // 1. 获取当前会话的缓存
+//        List<byte[]> fragments = fragmentCache.get(clientId);
+//        if (fragments == null) {
+//            fragments = new ArrayList<>();
+//            fragmentCache.put(clientId, fragments);
+//        }
+//        fragments.add(bytes);
+//        fragmentCache.put(clientId, fragments);
+//    }
+
     // 连接关闭时调用
     @OnClose
     public void onClose(Session session, CloseReason reason) {