funasr support payload.input and bug fix in transcription

src/main/java/com/alibaba/dashscope/api/AsynchronousApi.java

@@ -6,6 +6,7 @@
 
 import com.alibaba.dashscope.base.HalfDuplexParamBase;
 import com.alibaba.dashscope.common.DashScopeResult;
+import com.alibaba.dashscope.common.Status;
 import com.alibaba.dashscope.common.TaskStatus;
 import com.alibaba.dashscope.exception.ApiException;
 import com.alibaba.dashscope.exception.NoApiKeyException;
@@ -76,12 +77,17 @@ public DashScopeResult asyncCall(ParamT param, ServiceOption serviceOption)
    * @param apiKey The api-key.
    * @param baseUrl The base http url.
    * @param customHeaders The custom headers.
+   * @param timeoutSeconds The maximum time to wait in seconds.
    * @return The task result.
    * @throws NoApiKeyException Can not find api key
-   * @throws ApiException The request failed, possibly due to a network or data error.
+   * @throws ApiException The request failed, possibly due to a network or data error, or timeout.
    */
   public DashScopeResult wait(
-      String taskId, String apiKey, String baseUrl, Map<String, String> customHeaders)
+      String taskId,
+      String apiKey,
+      String baseUrl,
+      Map<String, String> customHeaders,
+      long timeoutSeconds)
       throws ApiException, NoApiKeyException {
     AsyncTaskOption serviceOption =
         AsyncTaskOption.builder()
@@ -98,7 +104,23 @@ public DashScopeResult wait(
     int maxWaitMilliseconds = 5 * 1000;
     int incrementSteps = 3;
     int step = 0;
+    long startTime = System.currentTimeMillis();
+    long timeoutMillis = timeoutSeconds > 0 ? timeoutSeconds * 1000L : -1L;
     while (true) {
+      if (timeoutMillis > 0) {
+        long elapsed = System.currentTimeMillis() - startTime;
+        if (elapsed >= timeoutMillis) {
+          throw new ApiException(
+              Status.builder()
+                  .statusCode(HttpURLConnection.HTTP_CLIENT_TIMEOUT)
+                  .code("TaskWaitTimeout")
+                  .message(
+                      StringUtils.format(
+                          "Waiting for task [%s] timed out after %d ms (timeoutSeconds=%d).",
+                          taskId, elapsed, timeoutSeconds))
+                  .build());
+        }
+      }
       try {
         DashScopeResult taskResult = client.send(req);
         JsonObject output = (JsonObject) taskResult.getOutput();
@@ -121,8 +143,18 @@ public DashScopeResult wait(
                     ? maxWaitMilliseconds
                     : waitMilliseconds * 2;
           }
+          long sleepMs = waitMilliseconds;
+          if (timeoutMillis > 0) {
+            long remaining = timeoutMillis - (System.currentTimeMillis() - startTime);
+            if (remaining <= 0) {
+              continue;
+            }
+            if (remaining < sleepMs) {
+              sleepMs = remaining;
+            }
+          }
           try {
-            Thread.sleep(waitMilliseconds);
+            Thread.sleep(sleepMs);
           } catch (InterruptedException ignored) {
           }
         }
@@ -135,6 +167,23 @@ public DashScopeResult wait(
     }
   }
 
+  /**
+   * Wait for async task completed and return task result.
+   *
+   * @param taskId The async task id.
+   * @param apiKey The api-key.
+   * @param baseUrl The base http url.
+   * @param customHeaders The custom headers.
+   * @return The task result.
+   * @throws NoApiKeyException Can not find api key
+   * @throws ApiException The request failed, possibly due to a network or data error.
+   */
+  public DashScopeResult wait(
+      String taskId, String apiKey, String baseUrl, Map<String, String> customHeaders)
+      throws ApiException, NoApiKeyException {
+    return wait(taskId, apiKey, baseUrl, customHeaders, -1);
+  }
+
   /**
    * Wait for async task completed and return task result.
    *

src/main/java/com/alibaba/dashscope/audio/asr/recognition/Recognition.java

@@ -9,6 +9,7 @@
 import com.alibaba.dashscope.exception.InputRequiredException;
 import com.alibaba.dashscope.exception.NoApiKeyException;
 import com.alibaba.dashscope.protocol.ApiServiceOption;
+import com.alibaba.dashscope.protocol.ConnectionOptions;
 import com.alibaba.dashscope.protocol.Protocol;
 import com.alibaba.dashscope.protocol.StreamingMode;
 import com.google.gson.Gson;
@@ -85,6 +86,9 @@ public static RecognitionParamWithStream FromRecognitionParam(
       if (param.getPhraseId() != null && !param.getPhraseId().isEmpty()) {
         recognitionParamWithStream.setPhraseId(param.getPhraseId());
       }
+      if (param.getInputs() != null && !param.getInputs().isEmpty()) {
+        recognitionParamWithStream.setInput(param.getInputs());
+      }
 
       return recognitionParamWithStream;
     }
@@ -103,6 +107,23 @@ public Recognition() {
     duplexApi = new SynchronizeFullDuplexApi<>(serviceOption);
   }
 
+  public Recognition(ConnectionOptions connectionOptions) {
+    serviceOption =
+        ApiServiceOption.builder()
+            .protocol(Protocol.WEBSOCKET)
+            .streamingMode(StreamingMode.DUPLEX)
+            .outputMode(OutputMode.ACCUMULATE)
+            .taskGroup(TaskGroup.AUDIO.getValue())
+            .task(Task.ASR.getValue())
+            .function(Function.RECOGNITION.getValue())
+            .build();
+    duplexApi = new SynchronizeFullDuplexApi<>(connectionOptions, serviceOption);
+  }
+
+  public void setWebsocketUrl(String websocketUrl) {
+    serviceOption.setBaseWebSocketUrl(websocketUrl);
+  }
+
   public Flowable<RecognitionResult> streamCall(
       RecognitionParam param, Flowable<ByteBuffer> audioFrame)
       throws ApiException, NoApiKeyException {

src/main/java/com/alibaba/dashscope/audio/asr/recognition/RecognitionParam.java

@@ -32,6 +32,8 @@ public class RecognitionParam extends FullDuplexServiceParam {
 
   private String vocabularyId;
 
+  private Map<String, Object> input;
+
   @Override
   public Map<String, Object> getParameters() {
     Map<String, Object> params = new HashMap<>();
@@ -45,6 +47,11 @@ public Map<String, Object> getParameters() {
     return params;
   }
 
+  @Override
+  public Map<String, Object> getInputs() {
+    return input;
+  }
+
   @Override
   public Object getResources() {
     if (phraseId == null || phraseId.isEmpty()) {

src/main/java/com/alibaba/dashscope/audio/asr/transcription/Transcription.java

@@ -45,13 +45,18 @@ public TranscriptionResult asyncCall(TranscriptionParam param) {
   }
 
   public TranscriptionResult wait(TranscriptionQueryParam queryParam) {
+    return wait(queryParam, -1);
+  }
+
+  public TranscriptionResult wait(TranscriptionQueryParam queryParam, long timeoutSeconds) {
     try {
       return TranscriptionResult.fromDashScopeResult(
           asyncApi.wait(
               queryParam.getTaskId(),
               queryParam.getApiKey(),
               baseUrl,
-              queryParam.getCustomHeaders()));
+              queryParam.getCustomHeaders(),
+              timeoutSeconds));
     } catch (NoApiKeyException e) {
       throw new ApiException(e);
     }

src/main/java/com/alibaba/dashscope/audio/qwen_asr/QwenTranscription.java

@@ -45,13 +45,18 @@ public QwenTranscriptionResult asyncCall(QwenTranscriptionParam param) {
   }
 
   public QwenTranscriptionResult wait(QwenTranscriptionQueryParam queryParam) {
+    return wait(queryParam, -1);
+  }
+
+  public QwenTranscriptionResult wait(QwenTranscriptionQueryParam queryParam, long timeoutSeconds) {
     try {
       return QwenTranscriptionResult.fromDashScopeResult(
           asyncApi.wait(
               queryParam.getTaskId(),
               queryParam.getApiKey(),
               baseUrl,
-              queryParam.getCustomHeaders()));
+              queryParam.getCustomHeaders(),
+              timeoutSeconds));
     } catch (NoApiKeyException e) {
       throw new ApiException(e);
     }