transport: create input transports push frame task

CHANGELOG.md

@@ -11,6 +11,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 - `DailyTransport`: don't publish camera and audio tracks if not enabled.
 
+- Fixed an issue in `BaseInputTransport` that was causing frames pushed
+  downstream not pushed in the right order.
+
 ## [0.0.15] - 2024-05-15
 
 ### Fixed

src/pipecat/transports/base_input.py

@@ -35,6 +35,13 @@ def __init__(self, params: TransportParams):
         if self._params.audio_in_enabled or self._params.vad_enabled:
             self._audio_in_queue = queue.Queue()
 
+        # Start push frame task. This is the task that will push frames in
+        # order. So, a transport guarantees that all frames are pushed in the
+        # same task.
+        loop = self.get_event_loop()
+        self._push_frame_task = loop.create_task(self._push_frame_task_handler())
+        self._push_queue = asyncio.Queue()
+
     async def start(self):
         if self._running:
             return
@@ -74,12 +81,30 @@ async def cleanup(self):
     async def process_frame(self, frame: Frame, direction: FrameDirection):
         if isinstance(frame, StartFrame):
             await self.start()
-            await self.push_frame(frame, direction)
+            await self._internal_push_frame(frame, direction)
         elif isinstance(frame, CancelFrame) or isinstance(frame, EndFrame):
             await self.stop()
-            await self.push_frame(frame, direction)
+            await self._internal_push_frame(frame, direction)
         else:
-            await self.push_frame(frame, direction)
+            await self._internal_push_frame(frame, direction)
+
+    #
+    # Push frames task
+    #
+
+    async def _internal_push_frame(
+            self,
+            frame: Frame,
+            direction: FrameDirection = FrameDirection.DOWNSTREAM):
+        await self._push_queue.put((frame, direction))
+
+    async def _push_frame_task_handler(self):
+        running = True
+        while running:
+            (frame, direction) = await self._push_queue.get()
+            if frame:
+                await self.push_frame(frame, direction)
+            running = frame is not None
 
     #
     # Audio input
@@ -95,7 +120,7 @@ def _handle_vad(self, audio_frames: bytes, vad_state: VADState):
                 frame = UserStoppedSpeakingFrame()
             if frame:
                 future = asyncio.run_coroutine_threadsafe(
-                    self.push_frame(frame), self.get_event_loop())
+                    self._internal_push_frame(frame), self.get_event_loop())
                 future.result()
                 vad_state = new_vad_state
         return vad_state
@@ -133,7 +158,7 @@ def _audio_out_thread_handler(self):
                 # Push audio downstream if passthrough.
                 if audio_passthrough:
                     future = asyncio.run_coroutine_threadsafe(
-                        self.push_frame(frame), self.get_event_loop())
+                        self._internal_push_frame(frame), self.get_event_loop())
                     future.result()
             except queue.Empty:
                 pass

src/pipecat/transports/local/tk.py

@@ -87,9 +87,7 @@ def write_raw_audio_frames(self, frames: bytes):
         self._out_stream.write(frames)
 
     def write_frame_to_camera(self, frame: ImageRawFrame):
-        future = asyncio.run_coroutine_threadsafe(
-            self._write_frame_to_tk(frame), self.get_event_loop())
-        future.result()
+        self.get_event_loop().call_soon(self._write_frame_to_tk, frame)
 
     async def start(self):
         await super().start()
@@ -107,7 +105,7 @@ async def cleanup(self):
 
         await super().cleanup()
 
-    async def _write_frame_to_tk(self, frame: ImageRawFrame):
+    def _write_frame_to_tk(self, frame: ImageRawFrame):
         width = frame.size[0]
         height = frame.size[1]
         data = f"P6 {width} {height} 255 ".encode() + frame.image

src/pipecat/transports/services/daily.py

@@ -478,12 +478,14 @@ async def process_frame(self, frame: Frame, direction: FrameDirection):
     #
 
     def push_transcription_frame(self, frame: TranscriptionFrame | InterimTranscriptionFrame):
-        future = asyncio.run_coroutine_threadsafe(self.push_frame(frame), self.get_event_loop())
+        future = asyncio.run_coroutine_threadsafe(
+            self._internal_push_frame(frame), self.get_event_loop())
         future.result()
 
     def push_app_message(self, message: Any, sender: str):
         frame = DailyTransportMessageFrame(message=message, participant_id=sender)
-        future = asyncio.run_coroutine_threadsafe(self.push_frame(frame), self.get_event_loop())
+        future = asyncio.run_coroutine_threadsafe(
+            self._internal_push_frame(frame), self.get_event_loop())
         future.result()
 
     #
@@ -543,7 +545,7 @@ def _camera_in_thread_handler(self):
             try:
                 frame = self._camera_in_queue.get(timeout=1)
                 future = asyncio.run_coroutine_threadsafe(
-                    self.push_frame(frame), self.get_event_loop())
+                    self._internal_push_frame(frame), self.get_event_loop())
                 future.result()
             except queue.Empty:
                 pass