Merge pull request #2709 from xinnan-tech/py_audio_change

优化
This commit is contained in:
hrz
2025-12-16 22:04:02 +08:00
committed by GitHub
2 changed files with 40 additions and 21 deletions
@@ -7,6 +7,10 @@ from core.providers.tts.dto.dto import SentenceType
from core.utils.audioRateController import AudioRateController from core.utils.audioRateController import AudioRateController
TAG = __name__ TAG = __name__
# 音频帧时长(毫秒)
AUDIO_FRAME_DURATION = 60
# 预缓冲包数量,直接发送以减少延迟
PRE_BUFFER_COUNT = 5
async def sendAudioMessage(conn, sentenceType, audios, text): async def sendAudioMessage(conn, sentenceType, audios, text):
@@ -45,7 +49,7 @@ async def sendAudioMessage(conn, sentenceType, audios, text):
async def _wait_for_audio_completion(conn): async def _wait_for_audio_completion(conn):
""" """
等待音频队列清空 等待音频队列清空并等待预缓冲包播放完成
Args: Args:
conn: 连接对象 conn: 连接对象
@@ -56,6 +60,13 @@ async def _wait_for_audio_completion(conn):
f"等待音频发送完成,队列中还有 {len(rate_controller.queue)} 个包" f"等待音频发送完成,队列中还有 {len(rate_controller.queue)} 个包"
) )
await rate_controller.queue_empty_event.wait() await rate_controller.queue_empty_event.wait()
# 等待预缓冲包播放完成
# 前N个包直接发送,增加2个网络抖动包,需要额外等待它们在客户端播放完成
frame_duration_ms = rate_controller.frame_duration
pre_buffer_playback_time = (PRE_BUFFER_COUNT + 2) * frame_duration_ms / 1000.0
await asyncio.sleep(pre_buffer_playback_time)
conn.logger.bind(tag=TAG).debug("音频发送完成") conn.logger.bind(tag=TAG).debug("音频发送完成")
@@ -81,14 +92,14 @@ async def _send_to_mqtt_gateway(conn, opus_packet, timestamp, sequence):
await conn.websocket.send(complete_packet) await conn.websocket.send(complete_packet)
async def sendAudio(conn, audios, frame_duration=60): async def sendAudio(conn, audios, frame_duration=AUDIO_FRAME_DURATION):
""" """
发送音频包,使用 AudioRateController 进行精确的流量控制 发送音频包,使用 AudioRateController 进行精确的流量控制
Args: Args:
conn: 连接对象 conn: 连接对象
audios: 单个opus包(bytes) 或 opus包列表 audios: 单个opus包(bytes) 或 opus包列表
frame_duration: 帧时长(毫秒),默认60ms frame_duration: 帧时长(毫秒),默认使用全局常量AUDIO_FRAME_DURATION
""" """
if audios is None or len(audios) == 0: if audios is None or len(audios) == 0:
return return
@@ -187,16 +198,14 @@ async def _send_audio_with_rate_control(
flow_control: 流控状态 flow_control: 流控状态
send_delay: 固定延迟(秒),-1表示使用动态流控 send_delay: 固定延迟(秒),-1表示使用动态流控
""" """
pre_buffer_count = 5
for packet in audio_list: for packet in audio_list:
if conn.client_abort: if conn.client_abort:
return return
conn.last_activity_time = time.time() * 1000 conn.last_activity_time = time.time() * 1000
# 预缓冲:前5个包直接发送 # 预缓冲:前N个包直接发送
if flow_control["packet_count"] < pre_buffer_count: if flow_control["packet_count"] < PRE_BUFFER_COUNT:
await _do_send_audio(conn, packet, flow_control) await _do_send_audio(conn, packet, flow_control)
conn.client_is_speaking = True conn.client_is_speaking = True
elif send_delay > 0: elif send_delay > 0:
@@ -1,5 +1,6 @@
import time import time
import asyncio import asyncio
from collections import deque
from config.logger import setup_logging from config.logger import setup_logging
TAG = __name__ TAG = __name__
@@ -18,13 +19,14 @@ class AudioRateController:
frame_duration: 单个音频帧时长(毫秒),默认60ms frame_duration: 单个音频帧时长(毫秒),默认60ms
""" """
self.frame_duration = frame_duration self.frame_duration = frame_duration
self.queue = [] self.queue = deque()
self.play_position = 0 # 虚拟播放位置(毫秒) self.play_position = 0 # 虚拟播放位置(毫秒)
self.start_timestamp = None # 开始时间戳(只读,不修改) self.start_timestamp = None # 开始时间戳(只读,不修改)
self.pending_send_task = None self.pending_send_task = None
self.logger = logger self.logger = logger
self.queue_empty_event = asyncio.Event() # 队列清空事件 self.queue_empty_event = asyncio.Event() # 队列清空事件
self.queue_empty_event.set() # 初始为空状态 self.queue_empty_event.set() # 初始为空状态
self.queue_has_data_event = asyncio.Event() # 队列数据事件
def reset(self): def reset(self):
"""重置控制器状态""" """重置控制器状态"""
@@ -34,13 +36,17 @@ class AudioRateController:
self.queue.clear() self.queue.clear()
self.play_position = 0 self.play_position = 0
self.start_timestamp = time.time() self.start_timestamp = None # 由首个音频包设置
self.queue_empty_event.set() # 队列已清空 # 相关事件处理
self.queue_empty_event.set()
self.queue_has_data_event.clear()
def add_audio(self, opus_packet): def add_audio(self, opus_packet):
"""添加音频包到队列""" """添加音频包到队列"""
self.queue.append(("audio", opus_packet)) self.queue.append(("audio", opus_packet))
self.queue_empty_event.clear() # 队列非空,清除事件 # 相关事件处理
self.queue_empty_event.clear()
self.queue_has_data_event.set()
def add_message(self, message_callback): def add_message(self, message_callback):
""" """
@@ -50,13 +56,15 @@ class AudioRateController:
message_callback: 消息发送回调函数 async def() message_callback: 消息发送回调函数 async def()
""" """
self.queue.append(("message", message_callback)) self.queue.append(("message", message_callback))
self.queue_empty_event.clear() # 队列非空,清除事件 # 相关事件处理
self.queue_empty_event.clear()
self.queue_has_data_event.set()
def _get_elapsed_ms(self): def _get_elapsed_ms(self):
"""获取已经过的时间(毫秒)""" """获取已经过的时间(毫秒)"""
if self.start_timestamp is None: if self.start_timestamp is None:
return 0 return 0
return (time.time() - self.start_timestamp) * 1000 return (time.monotonic() - self.start_timestamp) * 1000
async def check_queue(self, send_audio_callback): async def check_queue(self, send_audio_callback):
""" """
@@ -65,9 +73,6 @@ class AudioRateController:
Args: Args:
send_audio_callback: 发送音频的回调函数 async def(opus_packet) send_audio_callback: 发送音频的回调函数 async def(opus_packet)
""" """
if self.start_timestamp is None:
self.start_timestamp = time.time()
while self.queue: while self.queue:
item = self.queue[0] item = self.queue[0]
item_type = item[0] item_type = item[0]
@@ -75,7 +80,7 @@ class AudioRateController:
if item_type == "message": if item_type == "message":
# 消息类型:立即发送,不占用播放时间 # 消息类型:立即发送,不占用播放时间
_, message_callback = item _, message_callback = item
self.queue.pop(0) self.queue.popleft()
try: try:
await message_callback() await message_callback()
except Exception as e: except Exception as e:
@@ -83,6 +88,9 @@ class AudioRateController:
raise raise
elif item_type == "audio": elif item_type == "audio":
if self.start_timestamp is None:
self.start_timestamp = time.monotonic()
_, opus_packet = item _, opus_packet = item
# 循环等待直到时间到达 # 循环等待直到时间到达
@@ -107,16 +115,17 @@ class AudioRateController:
break break
# 时间已到,从队列移除并发送 # 时间已到,从队列移除并发送
self.queue.pop(0) self.queue.popleft()
self.play_position += self.frame_duration self.play_position += self.frame_duration
try: try:
await send_audio_callback(opus_packet) await send_audio_callback(opus_packet)
except Exception as e: except Exception as e:
self.logger.bind(tag=TAG).error(f"发送音频失败: {e}") self.logger.bind(tag=TAG).error(f"发送音频失败: {e}")
raise raise
# 队列处理完后清除事件
self.queue_empty_event.set() self.queue_empty_event.set()
self.queue_has_data_event.clear()
def start_sending(self, send_audio_callback): def start_sending(self, send_audio_callback):
""" """
@@ -132,9 +141,10 @@ class AudioRateController:
async def _send_loop(): async def _send_loop():
try: try:
while True: while True:
# 等待队列数据事件,不轮询等待占用CPU
await self.queue_has_data_event.wait()
await self.check_queue(send_audio_callback) await self.check_queue(send_audio_callback)
# 如果队列空了,短暂等待后再检查(避免 busy loop)
await asyncio.sleep(0.01)
except asyncio.CancelledError: except asyncio.CancelledError:
self.logger.bind(tag=TAG).debug("音频发送循环已停止") self.logger.bind(tag=TAG).debug("音频发送循环已停止")
except Exception as e: except Exception as e: