test-api-fix

app/services/cosyvoice_service.py

@@ -33,10 +33,20 @@ class CosyVoiceService:
                 if not self.connect():
                     return []
             
-            voices = self.client.predict(api_name="/refresh_sft_spk")
+            result = self.client.predict(api_name="/refresh_sft_spk")
-            # 过滤掉不需要的音色
+            
-            filtered_voices = [voice for voice in voices if voice != '.ipynb_checkpoints']
+            # 处理返回的字典格式
-            return filtered_voices
+            if isinstance(result, dict) and 'choices' in result:
+                # 从choices中提取音色名称
+                voices = [choice[0] for choice in result['choices'] if choice[0] != '.ipynb_checkpoints']
+                return voices
+            elif isinstance(result, list):
+                # 直接是列表格式
+                return [voice for voice in result if voice != '.ipynb_checkpoints']
+            else:
+                logger.error(f"未知的音色列表格式: {result}")
+                return []
+                
         except Exception as e:
             logger.error(f"获取音色列表失败: {str(e)}")
             return []
@@ -48,8 +58,17 @@ class CosyVoiceService:
                 if not self.connect():
                     return []
             
-            audio_files = self.client.predict(api_name="/refresh_prompt_wav")
+            result = self.client.predict(api_name="/refresh_prompt_wav")
-            return audio_files
+            
+            # 处理返回的字典格式
+            if isinstance(result, dict) and 'choices' in result:
+                audios = [choice[0] for choice in result['choices']]
+                return audios
+            elif isinstance(result, list):
+                return result
+            else:
+                return []
+                
         except Exception as e:
             logger.error(f"获取参考音频列表失败: {str(e)}")
             return []
@@ -65,7 +84,7 @@ class CosyVoiceService:
                 prompt_wav=handle_file(audio_file_path),
                 api_name="/prompt_wav_recognition"
             )
-            return text
+            return text if isinstance(text, str) else ""
         except Exception as e:
             logger.error(f"语音识别失败: {str(e)}")
             return ""
@@ -84,16 +103,41 @@ class CosyVoiceService:
                 if not self.connect():
                     return None, None
             
+            # 创建临时空音频文件用于占位
+            import tempfile
+            import wave
+            import numpy as np
+            
+            # 创建一个短的静音音频作为占位符
+            temp_audio = tempfile.NamedTemporaryFile(suffix='.wav', delete=False)
+            with wave.open(temp_audio.name, 'w') as wav_file:
+                wav_file.setnchannels(1)  # 单声道
+                wav_file.setsampwidth(2)  # 16位
+                wav_file.setframerate(16000)  # 16kHz采样率
+                # 写入很短的静音（0.1秒）
+                silence = np.zeros(1600, dtype=np.int16)  
+                wav_file.writeframes(silence.tobytes())
+            
             result = self.client.predict(
                 tts_text=text,
                 mode_checkbox_group="预训练音色",
                 sft_dropdown=voice,
-                seed=seed,
+                prompt_text="",
-                speed=speed,
+                prompt_wav_upload=handle_file(temp_audio.name),
-                stream="true" if stream else "false",
+                prompt_wav_record=handle_file(temp_audio.name),
+                instruct_text="",
+                seed=float(seed),
+                stream="True" if stream else "False",
+                speed=float(speed),
                 api_name="/generate_audio"
             )
             
+            # 清理临时文件
+            try:
+                os.unlink(temp_audio.name)
+            except:
+                pass
+            
             # result是一个元组 [流式音频路径, 完整音频路径]
             if isinstance(result, (list, tuple)) and len(result) >= 2:
                 return result[0], result[1]
@@ -127,9 +171,14 @@ class CosyVoiceService:
             result = self.client.predict(
                 tts_text=text,
                 mode_checkbox_group="3s极速复刻",
+                sft_dropdown="中文女",
                 prompt_text=reference_text,
                 prompt_wav_upload=handle_file(reference_audio_path),
-                seed=seed,
+                prompt_wav_record=handle_file(reference_audio_path),
+                instruct_text="",
+                seed=float(seed),
+                stream="False",
+                speed=1.0,
                 api_name="/generate_audio"
             )
             
@@ -154,14 +203,41 @@ class CosyVoiceService:
                 if not self.connect():
                     return None, None
             
+            # 创建临时空音频文件用于占位
+            import tempfile
+            import wave
+            import numpy as np
+            
+            # 创建一个短的静音音频作为占位符
+            temp_audio = tempfile.NamedTemporaryFile(suffix='.wav', delete=False)
+            with wave.open(temp_audio.name, 'w') as wav_file:
+                wav_file.setnchannels(1)  # 单声道
+                wav_file.setsampwidth(2)  # 16位
+                wav_file.setframerate(16000)  # 16kHz采样率
+                # 写入很短的静音（0.1秒）
+                silence = np.zeros(1600, dtype=np.int16)  
+                wav_file.writeframes(silence.tobytes())
+            
             result = self.client.predict(
                 tts_text=text,
                 mode_checkbox_group="自然语言控制",
+                sft_dropdown="中文女",
+                prompt_text="",
+                prompt_wav_upload=handle_file(temp_audio.name),
+                prompt_wav_record=handle_file(temp_audio.name),
                 instruct_text=instruction,
-                seed=seed,
+                seed=float(seed),
+                stream="False",
+                speed=1.0,
                 api_name="/generate_audio"
             )
             
+            # 清理临时文件
+            try:
+                os.unlink(temp_audio.name)
+            except:
+                pass
+            
             if isinstance(result, (list, tuple)) and len(result) >= 2:
                 return result[0], result[1]
             else:
@@ -178,8 +254,20 @@ class CosyVoiceService:
                 if not self.connect():
                     return 42
             
-            seed = self.client.predict(api_name="/generate_random_seed")
+            result = self.client.predict(api_name="/generate_random_seed")
-            return int(seed) if seed else 42
+            
+            # 处理返回的字典格式
+            if isinstance(result, dict) and 'value' in result:
+                seed = int(result['value'])
+            elif isinstance(result, (int, float)):
+                seed = int(result)
+            elif isinstance(result, str) and result.isdigit():
+                seed = int(result)
+            else:
+                logger.warning(f"未知的随机种子格式: {result}")
+                seed = 42
+            
+            return seed
         except Exception as e:
             logger.error(f"生成随机种子失败: {str(e)}")
             return 42