update audio output (#516)

modelscope_agent/tools/dashscope_tools/sambert_tts_tool.py

@@ -49,4 +49,4 @@ def call(self, params: str, **kwargs) -> str:
             raise ValueError(
                 f'call sambert tts failed, request id: {response.get_response()}'
             )
-        return str(AudioWrapper(wav_file))
+        return str(AudioWrapper(wav_file, **kwargs))

modelscope_agent/tools/modelscope_tools/text_to_speech_tool.py

@@ -32,10 +32,7 @@ def call(self, params: str, **kwargs) -> str:
             print('speech_generation error: ', result)
             return None
         audio = result['Data']['output_wav']
-        if 'use_tool_api' in kwargs:
-            return audio
-        else:
-            return str(AudioWrapper(audio))
+        return str(AudioWrapper(audio, **kwargs))
 
     def _verify_args(self, params: str):
         # override the args

modelscope_agent/tools/utils/output_wrapper.py

@@ -93,19 +93,22 @@ class AudioWrapper(OutputWrapper):
     Audio wrapper, raw_data is a binary file
     """
 
-    def __init__(self, audio) -> None:
+    def __init__(self, audio, **kwargs) -> None:
 
         super().__init__()
         if isinstance(audio, str):
-            if os.path.isfile(audio):
+            if 'use_tool_api' in kwargs and 'https://' in audio:
                 self._path = audio
             else:
-                self._path = self.get_remote_file(audio, 'wav')
-            try:
-                with open(self._path, 'rb') as f:
-                    self._raw_data = f.read()
-            except FileNotFoundError:
-                raise FileNotFoundError(f'Invalid path: {audio}')
+                if os.path.isfile(audio):
+                    self._path = audio
+                else:
+                    self._path = self.get_remote_file(audio, 'wav')
+                try:
+                    with open(self._path, 'rb') as f:
+                        self._raw_data = f.read()
+                except FileNotFoundError:
+                    raise FileNotFoundError(f'Invalid path: {audio}')
         else:
             self._raw_data = audio
             directory = tempfile.mkdtemp(dir=self.root_path)

tests/tools/test_modelscope_tool.py

@@ -21,7 +21,7 @@ def test_modelscope_speech_generation_with_tool_api():
     kwargs = {'use_tool_api': True}
     txt2speech = TexttoSpeechTool()
     res = txt2speech.call(params, **kwargs)
-    assert res.startswith('https://')
+    assert res.startswith('<audio src="https://')
 
 
 @pytest.mark.skipif(IS_FORKED_PR, reason='only run modelscope-agent main repo')