refactor: vad独立时间戳，相关残余状态处理

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

@@ -84,6 +84,12 @@ class ASRProviderBase(ABC):
     async def handle_voice_stop(self, conn: "ConnectionHandler", asr_audio_task: List[bytes]):
         """并行处理ASR和声纹识别"""
         try:
+            # 如果处于退出流程中，直接关闭连接，不处理新消息
+            if conn.close_after_chat or conn.is_exiting:
+                logger.bind(tag=TAG).info("退出流程中收到新消息，直接关闭连接")
+                await conn.close()
+                return
+
             total_start_time = time.monotonic()
 
             # 准备音频数据

main/xiaozhi-server/core/providers/vad/silero.py

@@ -107,12 +107,12 @@ class VADProvider(VADProviderBase):
 
                 # 如果之前有声音，但本次没有声音，且与上次有声音的时间差已经超过了静默阈值，则认为已经说完一句话
                 if conn.client_have_voice and not client_have_voice:
-                    stop_duration = time.time() * 1000 - conn.last_activity_time
+                    stop_duration = time.time() * 1000 - conn.vad_last_voice_time
                     if stop_duration >= self.silence_threshold_ms:
                         conn.client_voice_stop = True
                 if client_have_voice:
                     conn.client_have_voice = True
-                    conn.last_activity_time = time.time() * 1000
+                    conn.vad_last_voice_time = time.time() * 1000
 
             return client_have_voice
         except opuslib_next.OpusError as e: