add banbanmini backend

talkingq-url/services/tts_synthesizer.py

@@ -0,0 +1,101 @@
+import asyncio
+import time
+from typing import Dict
+from utils.logger import session_logger
+from services.interrupt_handler import interrupt_handler
+from utils.text_splitter import should_skip_tts
+from services.tts_config import TTSConfig
+from services.tts_audio_cleaner import TTSAudioCleaner
+from services.interruption_helper import InterruptionHelper
+
+class TTSSynthesizer:
+    def __init__(self, device_id: str, session_id: str):
+        self.device_id = device_id
+        self.session_id = session_id
+        self.session_key = (device_id, session_id)
+        self.request_timeout = TTSConfig.get_request_timeout()
+        self.max_tts_errors = TTSConfig.get_max_tts_errors()
+
+    def check_interruption(self):
+        return interrupt_handler.is_interrupted(self.session_key)
+
+    async def synthesize_audio(
+        self,
+        tts_service,
+        text_queue: asyncio.Queue,
+        url_queue: asyncio.Queue,
+        selected_role: Dict,
+        language: str = None,  
+    ):
+        sequence_number = 0
+        tts_errors_count = 0
+        error_notified = False
+        first_tts_submit_time = None
+        await TTSAudioCleaner.prepare_output_directory()
+        try:
+            while True:
+                if self.check_interruption():
+                    session_logger.info(
+                        self.device_id, self.session_id, "TTS合成过程收到中断请求，退出"
+                    )
+                    break
+                
+                text_item = await text_queue.get()
+                if text_item is None:
+                    session_logger.info(self.device_id, self.session_id, "TTS文本队列为空，合成结束")
+                    break
+                
+                if should_skip_tts(text_item):
+                    session_logger.info(self.device_id, self.session_id, f"跳过不需要TTS的文本: {text_item}")
+                    continue
+                
+                if first_tts_submit_time is None:
+                    first_tts_submit_time = time.perf_counter()
+                    url_queue.first_tts_submit_time = first_tts_submit_time
+                
+                try:
+                    sequence_number += 1
+                    tts_start_time = time.perf_counter()
+                    session_logger.info(self.device_id, self.session_id, f"开始合成音频 #{sequence_number}: {text_item}")
+                    
+                    output_file_prefix = TTSAudioCleaner.generate_filename(
+                        self.device_id, self.session_id, sequence_number
+                    )
+                    
+                    urls = await tts_service.tts(
+                        text_item,
+                        output_file_prefix=output_file_prefix,
+                        tts_format="mp3",
+                        selected_role=selected_role,
+                        language=language
+                    )
+                    
+                    tts_end_time = time.perf_counter()
+                    session_logger.info(
+                        self.device_id,
+                        self.session_id,
+                        f"音频合成完成 #{sequence_number}, 耗时: {tts_end_time - tts_start_time:.2f}秒"
+                    )
+                    
+                    if urls and not self.check_interruption():
+                        for url in urls:
+                            await url_queue.put(url)
+                except Exception as e:
+                    tts_errors_count += 1
+                    session_logger.error(
+                        self.device_id,
+                        self.session_id,
+                        f"TTS合成失败 (第{tts_errors_count}次错误): {str(e)}",
+                        exc_info=True
+                    )
+                    
+                    if tts_errors_count >= self.max_tts_errors and not error_notified:
+                        error_notified = True
+                        await InterruptionHelper.notify_client_tts_error(self.device_id, self.session_id)
+                        await InterruptionHelper.send_error_prompt_sound(self.device_id, self.session_id)
+        except Exception as e:
+            session_logger.error(
+                self.device_id, self.session_id, f"TTS合成任务出错: {e}", exc_info=True
+            )
+        finally:
+            await url_queue.put(None)