Hass/xiaozhi-esp32-server
1
0
Fork 0
mirror of https://github.com/xinnan-tech/xiaozhi-esp32-server.git synced 2026-07-22 07:03:53 +08:00
Code Issues Packages Projects Releases Wiki Activity

调整流控参数优化

Browse Source
This commit is contained in:
Chingfeng Li
2025-08-06 10:09:28 +08:00
parent 095fe72db0
commit ce6a711c8e
3 changed files with 30 additions and 34 deletions
Download Patch File Download Diff File Expand all files Collapse all files
main/xiaozhi-server/core/providers/tts/aliyun_stream.py
+9 -15
View File
@@ -447,21 +447,15 @@ class TTSProvider(TTSProviderBase):
elif event_name == "SentenceBegin":
opus_datas_cache = []
elif event_name == "SentenceEnd":
if (
not is_first_sentence
or first_sentence_segment_count > 10
):
# 发送缓存的数据
if self.conn.tts_MessageText:
logger.bind(tag=TAG).info(
f"句子语音生成成功： {self.conn.tts_MessageText}"
)
self.tts_audio_queue.put(
(SentenceType.MIDDLE, [], self.conn.tts_MessageText)
)
self.conn.tts_MessageText = None
# 第一句话结束后，将标志设置为False
is_first_sentence = False
# 发送缓存的数据
if self.conn.tts_MessageText:
logger.bind(tag=TAG).info(
f"句子语音生成成功： {self.conn.tts_MessageText}"
)
self.tts_audio_queue.put(
(SentenceType.FIRST, [], self.conn.tts_MessageText)
)
self.conn.tts_MessageText = None
elif event_name == "SynthesisCompleted":
logger.bind(tag=TAG).debug(f"会话结束～～")
self._process_before_stop_play_files()
main/xiaozhi-server/core/providers/tts/base.py
+19 -17
View File
@@ -253,7 +253,7 @@ class TTSProviderBase(ABC):
text = None
try:
try:
sentence_type, audio_datas, text = self.tts_audio_queue.get(timeout=1)
sentence_type, audio_datas, text = self.tts_audio_queue.get(timeout=0.1)
except queue.Empty:
if self.conn.stop_event.is_set():
break
@@ -288,15 +288,14 @@ class TTSProviderBase(ABC):
# 短暂等待后重试
time.sleep(retry_interval)
# 更新设备消费估计（这里假设设备以恒定速率消费）
# 实际应用中可能需要从设备端获取真实的消费情况
estimated_consumption = int(retry_interval * FlowControlConfig.DEFAULT_REFILL_RATE)
self.flow_controller.update_device_consumption(estimated_consumption)
status = self.flow_controller.get_status()
logger.bind(tag=TAG).debug(
f"流控状态: 缓冲区使用率={status['buffer_usage_percent']:.1f}%, "
f"可用令牌={status['available_tokens']}..."
)
# status = self.flow_controller.get_status()
# logger.bind(tag=TAG).debug(
# f"流控状态: 缓冲区使用率={status['buffer_usage_percent']:.1f}%, "
# f"可用令牌={status['available_tokens']}..."
# f"发送帧数={status['sent_frames']}..."
# f"消费帧数={status['consumed_frames']}..."
# f"代播放帧数={status['sent_frames'] - status['consumed_frames']}..."
# )
else:
# 可以发送，记录发送的帧数
self.flow_controller.record_sent_frames(frame_count)
@@ -314,16 +313,19 @@ class TTSProviderBase(ABC):
enqueue_tts_report(self.conn, text, audio_datas)
# 输出流控状态（调试用）
if frame_count > 0: # 只在较大的音频块时输出状态
status = self.flow_controller.get_status()
logger.bind(tag=TAG).debug(
f"流控状态: 缓冲区使用率={status['buffer_usage_percent']:.1f}%, "
f"可用令牌={status['available_tokens']}..."
)
# if frame_count > 0: # 只在较大的音频块时输出状态
# status = self.flow_controller.get_status()
# logger.bind(tag=TAG).debug(
# f"流控状态: 缓冲区使用率={status['buffer_usage_percent']:.1f}%, "
# f"可用令牌={status['available_tokens']}..."
# f"发送帧数={status['sent_frames']}..."
# f"消费帧数={status['consumed_frames']}..."
# f"代播放帧数={status['sent_frames'] - status['consumed_frames']}..."
# )
else:
# 没有音频数据，直接发送
future = asyncio.run_coroutine_threadsafe(
sendAudioMessage(self.conn, sentence_type, audio_datas, text),
self._send_audio_with_flow_control(sentence_type, audio_datas, text),
self.conn.loop,
)
future.result()
main/xiaozhi-server/core/utils/audio_flow_control.py
+2 -2
View File
@@ -174,9 +174,9 @@ class FlowControlConfig:
# 默认流控参数
DEFAULT_MAX_DEVICE_BUFFER = 40 # 设备端最大缓冲帧数
DEFAULT_REFILL_RATE = 5 # 默认令牌补充速率（帧/秒）
DEFAULT_REFILL_RATE = OPUS_FRAMES_PER_SECOND # 默认令牌补充速率（帧/秒）
DEFAULT_MAX_WAIT_TIME = 5.0 # 流控最大等待时间（秒）
DEFAULT_RETRY_INTERVAL = 0.1 # 流控重试间隔（秒）
DEFAULT_RETRY_INTERVAL = 0.06 # 流控重试间隔（秒）
# 预缓冲参数
PRE_BUFFER_FRAMES = 3 # 预缓冲帧数