Merge pull request #2924 from shengzhou1216/refactor/asr-delete_audio

refactor(asr): 重构speech_to_text方法以接收artifacts参数

main/xiaozhi-server/core/providers/asr/aliyun.py

@@ -213,7 +213,7 @@ class ASRProvider(ASRProviderBase):
             return None
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """将语音数据转换为文本"""
         if self._is_token_expired():
@@ -221,7 +221,6 @@ class ASRProvider(ASRProviderBase):
             self._refresh_token()
 
         try:
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
             # 发送请求并获取文本

main/xiaozhi-server/core/providers/asr/aliyun_stream.py

@@ -341,7 +341,7 @@ class ASRProvider(ASRProviderBase):
 
         logger.bind(tag=TAG).debug("ASR会话清理完成")
 
-    async def speech_to_text(self, opus_data, session_id, audio_format):
+    async def speech_to_text(self, opus_data, session_id, audio_format, artifacts=None):
         """获取识别结果"""
         result = self.text
         self.text = ""

main/xiaozhi-server/core/providers/asr/aliyunbl_stream.py

@@ -325,7 +325,7 @@ class ASRProvider(ASRProviderBase):
 
         logger.bind(tag=TAG).debug("ASR会话清理完成")
 
-    async def speech_to_text(self, opus_data, session_id, audio_format):
+    async def speech_to_text(self, opus_data, session_id, audio_format, artifacts=None):
         """获取识别结果"""
         result = self.text
         self.text = ""

main/xiaozhi-server/core/providers/asr/baidu.py

@@ -30,7 +30,7 @@ class ASRProvider(ASRProviderBase):
         os.makedirs(self.output_dir, exist_ok=True)
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """将语音数据转换为文本"""
         if not opus_data:
@@ -43,7 +43,6 @@ class ASRProvider(ASRProviderBase):
                 logger.bind(tag=TAG).error("百度语音识别配置未设置，无法进行识别")
                 return None, None
 
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
 

main/xiaozhi-server/core/providers/asr/base.py

@@ -25,8 +25,7 @@ logger = setup_logging()
 
 class ASRProviderBase(ABC):
     def __init__(self):
-        self._current_artifacts: Optional[ASRProviderBase.AudioArtifacts] = None
-        """当前正在处理的音频 artifact"""
+        pass
 
     # 打开音频通道
     async def open_audio_channels(self, conn):
@@ -215,9 +214,13 @@ class ASRProviderBase(ABC):
 
     class AudioArtifacts(NamedTuple):
         pcm_frames: List[bytes]
+        """PCM音频帧列表"""
         pcm_bytes: bytes
+        """合并后的PCM音频字节数据"""
         file_path: Optional[str]
+        """WAV文件路径"""
         temp_path: Optional[str]
+        """临时WAV文件路径"""
 
     def get_current_artifacts(self) -> Optional["ASRProviderBase.AudioArtifacts"]:
         return self._current_artifacts
@@ -282,14 +285,19 @@ class ASRProviderBase(ABC):
             ):
                 file_path = self.save_audio_to_file(pcm_data, session_id)
 
-            self._current_artifacts = ASRProviderBase.AudioArtifacts(
-                pcm_frames=pcm_data,
-                pcm_bytes=combined_pcm_data,
-                file_path=file_path,
-                temp_path=temp_path,
-            )
+            if len(combined_pcm_data) == 0:
+                artifacts = None
+            else:
+                artifacts = ASRProviderBase.AudioArtifacts(
+                    pcm_frames=pcm_data,
+                    pcm_bytes=combined_pcm_data,
+                    file_path=file_path,
+                    temp_path=temp_path,
+                )
 
-            text, _ = await self.speech_to_text(opus_data, session_id, audio_format)
+            text, _ = await self.speech_to_text(
+                opus_data, session_id, audio_format, artifacts
+            )
             return text, file_path
         except OSError as e:
             logger.bind(tag=TAG).error(f"文件操作错误: {e}")
@@ -313,9 +321,20 @@ class ASRProviderBase(ABC):
 
     @abstractmethod
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self,
+        opus_data: List[bytes],
+        session_id: str,
+        audio_format="opus",
+        artifacts: Optional[AudioArtifacts] = None,
     ) -> Tuple[Optional[str], Optional[str]]:
-        """将语音数据转换为文本"""
+        """将语音数据转换为文本
+
+        :param opus_data: 输入的Opus音频数据
+        :param session_id: 会话ID
+        :param audio_format: 音频格式，默认"opus"
+        :param artifacts: 音频工件，包含PCM数据、文件路径等
+        :return: 识别结果文本和文件路径（如果有）
+        """
         pass
 
     @staticmethod

main/xiaozhi-server/core/providers/asr/doubao.py

@@ -232,12 +232,11 @@ class ASRProvider(ASRProviderBase):
             yield data[offset:data_len], True
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """将语音数据转换为文本"""
 
         try:
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
 

main/xiaozhi-server/core/providers/asr/doubao_stream.py

@@ -408,7 +408,7 @@ class ASRProvider(ASRProviderBase):
             logger.bind(tag=TAG).error(f"原始响应数据: {res.hex()}")
             raise
 
-    async def speech_to_text(self, opus_data, session_id, audio_format):
+    async def speech_to_text(self, opus_data, session_id, audio_format, artifacts=None):
         result = self.text
         self.text = ""  # 清空text
         return result, None

main/xiaozhi-server/core/providers/asr/fun_local.py

@@ -64,14 +64,13 @@ class ASRProvider(ASRProviderBase):
             )
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """语音转文本主处理逻辑"""
         retry_count = 0
         
         while retry_count < MAX_RETRIES:
             try:
-                artifacts = self.get_current_artifacts()
                 if artifacts is None:
                     return "", None
 

main/xiaozhi-server/core/providers/asr/fun_server.py

@@ -101,7 +101,7 @@ class ASRProvider(ASRProviderBase):
         logger.bind(tag=TAG).debug(f"Sent end message: {end_message}")
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """
         Convert speech data to text using FunASR.
@@ -109,7 +109,6 @@ class ASRProvider(ASRProviderBase):
         :param session_id: Unique session identifier.
         :return: Tuple containing recognized text and optional timestamp.
         """
-        artifacts = self.get_current_artifacts()
         
         if artifacts is None:
             return "", None

main/xiaozhi-server/core/providers/asr/openai.py

@@ -24,10 +24,9 @@ class ASRProvider(ASRProviderBase):
     def requires_file(self) -> bool:
         return True
 
-    async def speech_to_text(self, opus_data: List[bytes], session_id: str, audio_format="opus") -> Tuple[Optional[str], Optional[str]]:
+    async def speech_to_text(self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None) -> Tuple[Optional[str], Optional[str]]:
         file_path = None
         try:
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
             file_path = artifacts.file_path

main/xiaozhi-server/core/providers/asr/qwen3_asr_flash.py

@@ -41,13 +41,12 @@ class ASRProvider(ASRProviderBase):
         return True
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """将语音数据转换为文本"""
         temp_file_path = None
         file_path = None
         try:
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
             temp_file_path = artifacts.temp_path

main/xiaozhi-server/core/providers/asr/sherpa_onnx_local.py

@@ -124,12 +124,11 @@ class ASRProvider(ASRProviderBase):
         return True
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """语音转文本主处理逻辑"""
         file_path = None
         try:
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
             file_path = artifacts.file_path

main/xiaozhi-server/core/providers/asr/tencent.py

@@ -32,7 +32,7 @@ class ASRProvider(ASRProviderBase):
         os.makedirs(self.output_dir, exist_ok=True)
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """将语音数据转换为文本"""
         if not opus_data:
@@ -45,7 +45,6 @@ class ASRProvider(ASRProviderBase):
                 logger.bind(tag=TAG).error("腾讯云语音识别配置未设置，无法进行识别")
                 return None, None
 
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
 

main/xiaozhi-server/core/providers/asr/vosk.py

@@ -44,7 +44,7 @@ class ASRProvider(ASRProviderBase):
             raise
 
     async def speech_to_text(
-        self, opus_data: List[bytes], session_id: str, audio_format="opus"
+        self, opus_data: List[bytes], session_id: str, audio_format="opus", artifacts=None
     ) -> Tuple[Optional[str], Optional[str]]:
         """将语音数据转换为文本"""
         try:
@@ -53,7 +53,6 @@ class ASRProvider(ASRProviderBase):
                 logger.bind(tag=TAG).error("VOSK模型未加载，无法进行识别")
                 return "", None
             
-            artifacts = self.get_current_artifacts()
             if artifacts is None:
                 return "", None
             if not artifacts.pcm_bytes:

main/xiaozhi-server/core/providers/asr/xunfei_stream.py

@@ -334,7 +334,7 @@ class ASRProvider(ASRProviderBase):
 
         logger.bind(tag=TAG).debug("ASR会话清理完成")
 
-    async def speech_to_text(self, opus_data, session_id, audio_format):
+    async def speech_to_text(self, opus_data, session_id, audio_format, artifacts=None):
         """获取识别结果"""
         result = self.text
         self.text = ""