From 7de2590d0a7324f2caa0dd29ea15082a690ade92 Mon Sep 17 00:00:00 2001
From: yumoqing <yumoqing@gmail.com>
Date: Fri, 13 Sep 2024 10:18:40 +0800
Subject: [PATCH] bugfix

---
 rtcllm/a2a.py |  4 ++--
 rtcllm/aav.py | 35 ++++++++++++++++++++++++++++-------
 rtcllm/rtc.py |  2 +-
 rtcllm/vad.py |  4 +++-
 4 files changed, 34 insertions(+), 11 deletions(-)

diff --git a/rtcllm/a2a.py b/rtcllm/a2a.py
index e37152d..32ca72e 100644
--- a/rtcllm/a2a.py
+++ b/rtcllm/a2a.py
@@ -74,7 +74,7 @@ class LLMAudioStreamTrack(AudioStreamTrack):
 		if audio_file is None:
 			print(f'*****{self.__class__.__name__}._feed(),{audio_file=}')
 			return
-		x = self.oi.run_AT_batch_stream(audio_file)
-		self.audio_iters.append(x)
+		abiters = self.oi.run_AT_batch_stream(audio_file)
+		self.audio_iters.append(abiters)
 		return x
 	
diff --git a/rtcllm/aav.py b/rtcllm/aav.py
index 38f8b3b..4f38b0e 100644
--- a/rtcllm/aav.py
+++ b/rtcllm/aav.py
@@ -7,8 +7,10 @@ from aiortc.mediastreams import MediaStreamError
 class MyMediaPlayer(MediaPlayer):
 	pass
 
-class MyTrackBase:
+class MyTrackBase(MediaStreamTrack):
 	def __init__(self, source=None):
+		super().__init__()
+		self.dumb = None
 		self.source = source
 		self.set_source_track()
 		print(f'{self.kind=}, {self.__class__.__name__}, {dir(self)}')
@@ -30,13 +32,31 @@ class MyTrackBase:
 	async def recv(self):
 		print(f'============{self.__class__.__name__}, {self.source.duration=}, {self.source.time=}')
 		if self.source is None:
-			return None
-		if self.track.readyState != 'live':
-			self.source_reload()
-		f = await self.track.recv()
+			return await self.dumb.recv()
+		try:
+			f = await self.track.recv()
+			return f
+		except MediaStreamError:
+			return await self.dumb.recv()
+		except Exception as e:
+			print_exc()
+			print(f'{e}')
+			return await self.dumb.recv()
 		return f
 
+class MyAudioStreamTrack(MyTrackBase):
+	kind = 'audio'
+	def __init__(self, source=None):
+		super().__init__()
+		self.dumb = AudioStreamTrack()
+		
+class MyVideoStreamTrack(MyTrackBase):
+	kind = 'video'
+	def __init__(self, source=None):
+		super().__init__()
+		self.dumb = AudioStreamTrack()
 
+"""
 class MyAudioStreamTrack(AudioStreamTrack):
 	def  __init__(self, source=None):
 		super().__init__()
@@ -116,9 +136,9 @@ class MyVideoStreamTrack(VideoStreamTrack):
 
 
 class LoopingVideoTrack(VideoStreamTrack):
-	"""
+	'''
 	A video track that loops a video file.
-	"""
+	'''
 	def __init__(self, filename):
 		super().__init__()
 		self.player = MyMediaPlayer(filename)
@@ -130,4 +150,5 @@ class LoopingVideoTrack(VideoStreamTrack):
 			self.player = MyMediaPlayer(self.player._file_path)
 			frame = await self.player.video.recv()
 		return frame
+"""
 
diff --git a/rtcllm/rtc.py b/rtcllm/rtc.py
index 685c3ee..c5ce407 100644
--- a/rtcllm/rtc.py
+++ b/rtcllm/rtc.py
@@ -138,7 +138,7 @@ class RTCLLM:
 		pc = peer.pc
 		if track.kind == 'audio':
 			f = partial(self.vad_voiceend, peer)
-			vadtrack = AudioTrackVad(track, stage=2, onvoiceend=f)
+			vadtrack = AudioTrackVad(track, stage=3, onvoiceend=f)
 			peer.vadtrack = vadtrack
 			vadtrack.start_vad()
 
diff --git a/rtcllm/vad.py b/rtcllm/vad.py
index e74178c..77c2ea0 100644
--- a/rtcllm/vad.py
+++ b/rtcllm/vad.py
@@ -20,7 +20,7 @@ class AudioTrackVad(MediaStreamTrack):
 		# self.sample_rate = self.track.getSettings().sampleRate
 		# frameSize = self.track.getSettings().frameSize
 		# self.frame_duration_ms = (1000 * frameSize) / self.sample_rate
-		self.frame_duration_ms = 0.00002
+		self.frame_duration_ms = 0.02
 		self.num_padding_frames = 20
 		self.ring_buffer = collections.deque(maxlen=self.num_padding_frames)
 		self.triggered = False
@@ -51,6 +51,8 @@ class AudioTrackVad(MediaStreamTrack):
 	async def recv(self):
 		oldf = await self.track.recv()
 		self.sample_rate = oldf.sample_rate
+		duration = (oldf.samples * 1000) / oldf.sample_rate
+		print(f'{duration=}, {oldf.samples=}, (oldf.sample_rate=}')
 		frames = self.resample(oldf)
 		for f in frames:
 			if self.debug: