更新音频任务处理逻辑，使用模型副本替代原始任务对象，以确保数据一致性和完整性。

src/voice_dialogue/services/audio/generator.py

@@ -103,4 +103,4 @@ class TTSAudioGenerator(BaseThread, TaskStatusMixin):
         voice_task.tts_generated_sentence_audio = tts_generated_sentence_audio
         voice_task.tts_end_time = time.time()
 
-        self.audio_output_queue.put(voice_task)
+        self.audio_output_queue.put(voice_task.model_copy())

src/voice_dialogue/services/speech/monitor.py

@@ -291,7 +291,7 @@ class SpeechStateMonitor(BaseThread):
                 # 8. 检查是否需要发送语音任务
                 if self._should_send_voice_task(is_audio_sent_for_processing):
                     voice_task = self._create_voice_task(audio_frames)
-                    self.user_voice_queue.put(voice_task)
+                    self.user_voice_queue.put(voice_task.model_copy(deep=True))
 
                     # 更新状态
                     is_audio_sent_for_processing = True

src/voice_dialogue/services/speech/recognizer.py

@@ -38,7 +38,6 @@ class ASRWorker(BaseThread, PerformanceLogMixin):
             except Empty:
                 continue
 
-
             voice_task.language = self.language
             voice_task.whisper_start_time = time.time()
 
@@ -47,16 +46,16 @@ class ASRWorker(BaseThread, PerformanceLogMixin):
             if not transcribed_text.strip():
                 voice_state_manager.reset_task_id()
                 continue
-                
+
             self.log_task_user_question(voice_task)
 
             voice_task.whisper_end_time = time.time()
 
             task_id = voice_task.id
-            cached_user_question = self.cached_user_questions.get(task_id, [])
-            cached_user_question.append(transcribed_text)
 
+            cached_user_question = self.cached_user_questions.get(task_id, [])
             if voice_task.is_over_audio_frames_threshold:
+                cached_user_question.append(transcribed_text)
                 self.cached_user_questions[task_id] = cached_user_question
 
             answer_id = voice_task.answer_id
@@ -72,4 +71,4 @@ class ASRWorker(BaseThread, PerformanceLogMixin):
             voice_task.transcribed_text = ' '.join(cached_user_question) if cached_user_question else transcribed_text
 
             voice_task.user_voice = []
-            self.transcribed_text_queue.put(voice_task)
+            self.transcribed_text_queue.put(voice_task.model_copy())