dv-pipecat-ai 0.0.82.dev69__py3-none-any.whl → 0.0.82.dev759__py3-none-any.whl

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.

Potentially problematic release.


This version of dv-pipecat-ai might be problematic. Click here for more details.

Files changed (157) hide show
  1. {dv_pipecat_ai-0.0.82.dev69.dist-info → dv_pipecat_ai-0.0.82.dev759.dist-info}/METADATA +78 -117
  2. {dv_pipecat_ai-0.0.82.dev69.dist-info → dv_pipecat_ai-0.0.82.dev759.dist-info}/RECORD +157 -123
  3. pipecat/adapters/base_llm_adapter.py +38 -1
  4. pipecat/adapters/services/anthropic_adapter.py +9 -14
  5. pipecat/adapters/services/aws_nova_sonic_adapter.py +5 -0
  6. pipecat/adapters/services/bedrock_adapter.py +236 -13
  7. pipecat/adapters/services/gemini_adapter.py +12 -8
  8. pipecat/adapters/services/open_ai_adapter.py +19 -7
  9. pipecat/adapters/services/open_ai_realtime_adapter.py +5 -0
  10. pipecat/audio/filters/krisp_viva_filter.py +193 -0
  11. pipecat/audio/filters/noisereduce_filter.py +15 -0
  12. pipecat/audio/turn/base_turn_analyzer.py +9 -1
  13. pipecat/audio/turn/smart_turn/base_smart_turn.py +14 -8
  14. pipecat/audio/turn/smart_turn/data/__init__.py +0 -0
  15. pipecat/audio/turn/smart_turn/data/smart-turn-v3.0.onnx +0 -0
  16. pipecat/audio/turn/smart_turn/http_smart_turn.py +6 -2
  17. pipecat/audio/turn/smart_turn/local_smart_turn.py +1 -1
  18. pipecat/audio/turn/smart_turn/local_smart_turn_v2.py +1 -1
  19. pipecat/audio/turn/smart_turn/local_smart_turn_v3.py +124 -0
  20. pipecat/audio/vad/data/README.md +10 -0
  21. pipecat/audio/vad/vad_analyzer.py +13 -1
  22. pipecat/extensions/voicemail/voicemail_detector.py +5 -5
  23. pipecat/frames/frames.py +232 -88
  24. pipecat/observers/loggers/debug_log_observer.py +3 -3
  25. pipecat/observers/loggers/llm_log_observer.py +7 -3
  26. pipecat/observers/loggers/user_bot_latency_log_observer.py +22 -10
  27. pipecat/pipeline/runner.py +12 -4
  28. pipecat/pipeline/service_switcher.py +64 -36
  29. pipecat/pipeline/task.py +85 -24
  30. pipecat/processors/aggregators/dtmf_aggregator.py +28 -22
  31. pipecat/processors/aggregators/{gated_openai_llm_context.py → gated_llm_context.py} +9 -9
  32. pipecat/processors/aggregators/gated_open_ai_llm_context.py +12 -0
  33. pipecat/processors/aggregators/llm_response.py +6 -7
  34. pipecat/processors/aggregators/llm_response_universal.py +19 -15
  35. pipecat/processors/aggregators/user_response.py +6 -6
  36. pipecat/processors/aggregators/vision_image_frame.py +24 -2
  37. pipecat/processors/audio/audio_buffer_processor.py +43 -8
  38. pipecat/processors/dtmf_aggregator.py +128 -87
  39. pipecat/processors/filters/stt_mute_filter.py +17 -0
  40. pipecat/processors/frame_processor.py +103 -17
  41. pipecat/processors/frameworks/langchain.py +8 -2
  42. pipecat/processors/frameworks/rtvi.py +209 -68
  43. pipecat/processors/frameworks/strands_agents.py +170 -0
  44. pipecat/processors/logger.py +2 -2
  45. pipecat/processors/transcript_processor.py +4 -4
  46. pipecat/processors/user_idle_processor.py +18 -10
  47. pipecat/runner/run.py +270 -50
  48. pipecat/runner/types.py +2 -0
  49. pipecat/runner/utils.py +51 -10
  50. pipecat/serializers/exotel.py +5 -5
  51. pipecat/serializers/livekit.py +20 -0
  52. pipecat/serializers/plivo.py +6 -9
  53. pipecat/serializers/protobuf.py +6 -5
  54. pipecat/serializers/telnyx.py +2 -2
  55. pipecat/serializers/twilio.py +43 -23
  56. pipecat/services/ai_service.py +2 -6
  57. pipecat/services/anthropic/llm.py +2 -25
  58. pipecat/services/asyncai/tts.py +2 -3
  59. pipecat/services/aws/__init__.py +1 -0
  60. pipecat/services/aws/llm.py +122 -97
  61. pipecat/services/aws/nova_sonic/__init__.py +0 -0
  62. pipecat/services/aws/nova_sonic/context.py +367 -0
  63. pipecat/services/aws/nova_sonic/frames.py +25 -0
  64. pipecat/services/aws/nova_sonic/llm.py +1155 -0
  65. pipecat/services/aws/stt.py +1 -3
  66. pipecat/services/aws_nova_sonic/__init__.py +19 -1
  67. pipecat/services/aws_nova_sonic/aws.py +11 -1151
  68. pipecat/services/aws_nova_sonic/context.py +13 -355
  69. pipecat/services/aws_nova_sonic/frames.py +13 -17
  70. pipecat/services/azure/realtime/__init__.py +0 -0
  71. pipecat/services/azure/realtime/llm.py +65 -0
  72. pipecat/services/azure/stt.py +15 -0
  73. pipecat/services/cartesia/tts.py +2 -2
  74. pipecat/services/deepgram/__init__.py +1 -0
  75. pipecat/services/deepgram/flux/__init__.py +0 -0
  76. pipecat/services/deepgram/flux/stt.py +636 -0
  77. pipecat/services/elevenlabs/__init__.py +2 -1
  78. pipecat/services/elevenlabs/stt.py +254 -276
  79. pipecat/services/elevenlabs/tts.py +5 -5
  80. pipecat/services/fish/tts.py +2 -2
  81. pipecat/services/gemini_multimodal_live/events.py +38 -524
  82. pipecat/services/gemini_multimodal_live/file_api.py +23 -173
  83. pipecat/services/gemini_multimodal_live/gemini.py +41 -1403
  84. pipecat/services/gladia/stt.py +56 -72
  85. pipecat/services/google/__init__.py +1 -0
  86. pipecat/services/google/gemini_live/__init__.py +3 -0
  87. pipecat/services/google/gemini_live/file_api.py +189 -0
  88. pipecat/services/google/gemini_live/llm.py +1582 -0
  89. pipecat/services/google/gemini_live/llm_vertex.py +184 -0
  90. pipecat/services/google/llm.py +15 -11
  91. pipecat/services/google/llm_openai.py +3 -3
  92. pipecat/services/google/llm_vertex.py +86 -16
  93. pipecat/services/google/tts.py +7 -3
  94. pipecat/services/heygen/api.py +2 -0
  95. pipecat/services/heygen/client.py +8 -4
  96. pipecat/services/heygen/video.py +2 -0
  97. pipecat/services/hume/__init__.py +5 -0
  98. pipecat/services/hume/tts.py +220 -0
  99. pipecat/services/inworld/tts.py +6 -6
  100. pipecat/services/llm_service.py +15 -5
  101. pipecat/services/lmnt/tts.py +2 -2
  102. pipecat/services/mcp_service.py +4 -2
  103. pipecat/services/mem0/memory.py +6 -5
  104. pipecat/services/mistral/llm.py +29 -8
  105. pipecat/services/moondream/vision.py +42 -16
  106. pipecat/services/neuphonic/tts.py +2 -2
  107. pipecat/services/openai/__init__.py +1 -0
  108. pipecat/services/openai/base_llm.py +27 -20
  109. pipecat/services/openai/realtime/__init__.py +0 -0
  110. pipecat/services/openai/realtime/context.py +272 -0
  111. pipecat/services/openai/realtime/events.py +1106 -0
  112. pipecat/services/openai/realtime/frames.py +37 -0
  113. pipecat/services/openai/realtime/llm.py +829 -0
  114. pipecat/services/openai/tts.py +16 -8
  115. pipecat/services/openai_realtime/__init__.py +27 -0
  116. pipecat/services/openai_realtime/azure.py +21 -0
  117. pipecat/services/openai_realtime/context.py +21 -0
  118. pipecat/services/openai_realtime/events.py +21 -0
  119. pipecat/services/openai_realtime/frames.py +21 -0
  120. pipecat/services/openai_realtime_beta/azure.py +16 -0
  121. pipecat/services/openai_realtime_beta/openai.py +17 -5
  122. pipecat/services/playht/tts.py +31 -4
  123. pipecat/services/rime/tts.py +3 -4
  124. pipecat/services/sarvam/tts.py +2 -6
  125. pipecat/services/simli/video.py +2 -2
  126. pipecat/services/speechmatics/stt.py +1 -7
  127. pipecat/services/stt_service.py +34 -0
  128. pipecat/services/tavus/video.py +2 -2
  129. pipecat/services/tts_service.py +9 -9
  130. pipecat/services/vision_service.py +7 -6
  131. pipecat/tests/utils.py +4 -4
  132. pipecat/transcriptions/language.py +41 -1
  133. pipecat/transports/base_input.py +17 -42
  134. pipecat/transports/base_output.py +42 -26
  135. pipecat/transports/daily/transport.py +199 -26
  136. pipecat/transports/heygen/__init__.py +0 -0
  137. pipecat/transports/heygen/transport.py +381 -0
  138. pipecat/transports/livekit/transport.py +228 -63
  139. pipecat/transports/local/audio.py +6 -1
  140. pipecat/transports/local/tk.py +11 -2
  141. pipecat/transports/network/fastapi_websocket.py +1 -1
  142. pipecat/transports/smallwebrtc/connection.py +98 -19
  143. pipecat/transports/smallwebrtc/request_handler.py +204 -0
  144. pipecat/transports/smallwebrtc/transport.py +65 -23
  145. pipecat/transports/tavus/transport.py +23 -12
  146. pipecat/transports/websocket/client.py +41 -5
  147. pipecat/transports/websocket/fastapi.py +21 -11
  148. pipecat/transports/websocket/server.py +14 -7
  149. pipecat/transports/whatsapp/api.py +8 -0
  150. pipecat/transports/whatsapp/client.py +47 -0
  151. pipecat/utils/base_object.py +54 -22
  152. pipecat/utils/string.py +12 -1
  153. pipecat/utils/tracing/service_decorators.py +21 -21
  154. {dv_pipecat_ai-0.0.82.dev69.dist-info → dv_pipecat_ai-0.0.82.dev759.dist-info}/WHEEL +0 -0
  155. {dv_pipecat_ai-0.0.82.dev69.dist-info → dv_pipecat_ai-0.0.82.dev759.dist-info}/licenses/LICENSE +0 -0
  156. {dv_pipecat_ai-0.0.82.dev69.dist-info → dv_pipecat_ai-0.0.82.dev759.dist-info}/top_level.txt +0 -0
  157. /pipecat/services/{aws_nova_sonic → aws/nova_sonic}/ready.wav +0 -0
@@ -29,8 +29,9 @@ from pipecat.frames.frames import (
29
29
  FrameProcessorPauseUrgentFrame,
30
30
  FrameProcessorResumeFrame,
31
31
  FrameProcessorResumeUrgentFrame,
32
+ InterruptionFrame,
33
+ InterruptionTaskFrame,
32
34
  StartFrame,
33
- StartInterruptionFrame,
34
35
  SystemFrame,
35
36
  )
36
37
  from pipecat.metrics.metrics import LLMTokenUsage, MetricsData
@@ -141,6 +142,12 @@ class FrameProcessor(BaseObject):
141
142
  task. System frames are also processed in a separate task which guarantees
142
143
  frame priority.
143
144
 
145
+ Event handlers available:
146
+
147
+ - on_before_process_frame: Called before a frame is processed
148
+ - on_after_process_frame: Called after a frame is processed
149
+ - on_before_push_frame: Called before a frame is pushed
150
+ - on_after_push_frame: Called after a frame is pushed
144
151
  """
145
152
 
146
153
  def __init__(
@@ -221,6 +228,20 @@ class FrameProcessor(BaseObject):
221
228
  self.__process_frame_task: Optional[asyncio.Task] = None
222
229
  self.logger = logger # Will later be replaced with a bound logger
223
230
 
231
+ # To interrupt a pipeline, we push an `InterruptionTaskFrame` upstream.
232
+ # Then we wait for the corresponding `InterruptionFrame` to travel from
233
+ # the start of the pipeline back to the processor that sent the
234
+ # `InterruptionTaskFrame`. This wait is handled using the following
235
+ # event.
236
+ self._wait_for_interruption = False
237
+ self._wait_interruption_event = asyncio.Event()
238
+
239
+ # Frame processor events.
240
+ self._register_event_handler("on_before_process_frame", sync=True)
241
+ self._register_event_handler("on_after_process_frame", sync=True)
242
+ self._register_event_handler("on_before_push_frame", sync=True)
243
+ self._register_event_handler("on_after_push_frame", sync=True)
244
+
224
245
  @property
225
246
  def id(self) -> int:
226
247
  """Get the unique identifier for this processor.
@@ -436,9 +457,13 @@ class FrameProcessor(BaseObject):
436
457
  name = f"{self}::{coroutine.cr_code.co_name}"
437
458
  return self.task_manager.create_task(coroutine, name)
438
459
 
439
- async def cancel_task(self, task: asyncio.Task, timeout: Optional[float] = None):
460
+ async def cancel_task(self, task: asyncio.Task, timeout: Optional[float] = 1.0):
440
461
  """Cancel a task managed by this processor.
441
462
 
463
+ A default timeout if 1 second is used in order to avoid potential
464
+ freezes caused by certain libraries that swallow
465
+ `asyncio.CancelledError`.
466
+
442
467
  Args:
443
468
  task: The task to cancel.
444
469
  timeout: Optional timeout for task cancellation.
@@ -544,6 +569,14 @@ class FrameProcessor(BaseObject):
544
569
  if self._cancelling:
545
570
  return
546
571
 
572
+ # If we are waiting for an interruption we will bypass all queued system
573
+ # frames and we will process the frame right away. This is because a
574
+ # previous system frame might be waiting for the interruption frame and
575
+ # it's blocking the input task.
576
+ if self._wait_for_interruption and isinstance(frame, InterruptionFrame):
577
+ await self.__process_frame(frame, direction, callback)
578
+ return
579
+
547
580
  if self._enable_direct_mode:
548
581
  await self.__process_frame(frame, direction, callback)
549
582
  else:
@@ -553,11 +586,15 @@ class FrameProcessor(BaseObject):
553
586
  """Pause processing of queued frames."""
554
587
  self.logger.trace(f"{self}: pausing frame processing")
555
588
  self.__should_block_frames = True
589
+ if self.__process_event:
590
+ self.__process_event.clear()
556
591
 
557
592
  async def pause_processing_system_frames(self):
558
593
  """Pause processing of queued system frames."""
559
594
  logger.trace(f"{self}: pausing system frame processing")
560
595
  self.__should_block_system_frames = True
596
+ if self.__input_event:
597
+ self.__input_event.clear()
561
598
 
562
599
  async def resume_processing_frames(self):
563
600
  """Resume processing of queued frames."""
@@ -590,7 +627,7 @@ class FrameProcessor(BaseObject):
590
627
 
591
628
  if isinstance(frame, StartFrame):
592
629
  await self.__start(frame)
593
- elif isinstance(frame, StartInterruptionFrame):
630
+ elif isinstance(frame, InterruptionFrame):
594
631
  await self._start_interruption()
595
632
  await self.stop_all_metrics()
596
633
  elif isinstance(frame, CancelFrame):
@@ -620,8 +657,40 @@ class FrameProcessor(BaseObject):
620
657
  if not self._check_started(frame):
621
658
  return
622
659
 
660
+ await self._call_event_handler("on_before_push_frame", frame)
661
+
623
662
  await self.__internal_push_frame(frame, direction)
624
663
 
664
+ await self._call_event_handler("on_after_push_frame", frame)
665
+
666
+ # If we are waiting for an interruption and we get an interruption, then
667
+ # we can unblock `push_interruption_task_frame_and_wait()`.
668
+ if self._wait_for_interruption and isinstance(frame, InterruptionFrame):
669
+ self._wait_interruption_event.set()
670
+
671
+ async def push_interruption_task_frame_and_wait(self):
672
+ """Push an interruption task frame upstream and wait for the interruption.
673
+
674
+ This function sends an `InterruptionTaskFrame` upstream to the pipeline
675
+ task and waits to receive the corresponding `InterruptionFrame`. When
676
+ the function finishes it is guaranteed that the `InterruptionFrame` has
677
+ been pushed downstream.
678
+ """
679
+ self._wait_for_interruption = True
680
+
681
+ await self.push_frame(InterruptionTaskFrame(), FrameDirection.UPSTREAM)
682
+
683
+ # Wait for an `InterruptionFrame` to come to this processor and be
684
+ # pushed. Take a look at `push_frame()` to see how we first push the
685
+ # `InterruptionFrame` and then we set the event in order to maintain
686
+ # frame ordering.
687
+ await self._wait_interruption_event.wait()
688
+
689
+ # Clean the event.
690
+ self._wait_interruption_event.clear()
691
+
692
+ self._wait_for_interruption = False
693
+
625
694
  async def __start(self, frame: StartFrame):
626
695
  """Handle the start frame to initialize processor state.
627
696
 
@@ -674,22 +743,24 @@ class FrameProcessor(BaseObject):
674
743
  async def _start_interruption(self):
675
744
  """Start handling an interruption by cancelling current tasks."""
676
745
  try:
677
- # Cancel the process task. This will stop processing queued frames.
678
- await self.__cancel_process_task()
746
+ if self._wait_for_interruption:
747
+ # If we get here we know the process task was just waiting for
748
+ # an interruption (push_interruption_task_frame_and_wait()), so
749
+ # we can't cancel the task because it might still need to do
750
+ # more things (e.g. pushing a frame after the
751
+ # interruption). Instead we just drain the queue because this is
752
+ # an interruption.
753
+ self.__reset_process_task()
754
+ else:
755
+ # Cancel and re-create the process task including the queue.
756
+ await self.__cancel_process_task()
757
+ self.__create_process_task()
679
758
  except Exception as e:
680
759
  self.logger.exception(
681
760
  f"Uncaught exception in {self} when handling _start_interruption: {e}"
682
761
  )
683
762
  await self.push_error(ErrorFrame(str(e)))
684
763
 
685
- # Create a new process queue and task.
686
- self.__create_process_task()
687
-
688
- async def _stop_interruption(self):
689
- """Stop handling an interruption."""
690
- # Nothing to do right now.
691
- pass
692
-
693
764
  async def __internal_push_frame(self, frame: Frame, direction: FrameDirection):
694
765
  """Internal method to push frames to adjacent processors.
695
766
 
@@ -774,6 +845,17 @@ class FrameProcessor(BaseObject):
774
845
  self.__process_queue = asyncio.Queue()
775
846
  self.__process_frame_task = self.create_task(self.__process_frame_task_handler())
776
847
 
848
+ def __reset_process_task(self):
849
+ """Reset non-system frame processing task."""
850
+ if self._enable_direct_mode:
851
+ return
852
+
853
+ self.__should_block_frames = False
854
+ self.__process_event = asyncio.Event()
855
+ while not self.__process_queue.empty():
856
+ self.__process_queue.get_nowait()
857
+ self.__process_queue.task_done()
858
+
777
859
  async def __cancel_process_task(self):
778
860
  """Cancel the non-system frame processing task."""
779
861
  if self.__process_frame_task:
@@ -784,11 +866,15 @@ class FrameProcessor(BaseObject):
784
866
  self, frame: Frame, direction: FrameDirection, callback: Optional[FrameCallback]
785
867
  ):
786
868
  try:
869
+ await self._call_event_handler("on_before_process_frame", frame)
870
+
787
871
  # Process the frame.
788
872
  await self.process_frame(frame, direction)
789
873
  # If this frame has an associated callback, call it now.
790
874
  if callback:
791
875
  await callback(self, frame, direction)
876
+
877
+ await self._call_event_handler("on_after_process_frame", frame)
792
878
  except Exception as e:
793
879
  logger.exception(f"{self}: error processing frame: {e}")
794
880
  await self.push_error(ErrorFrame(str(e)))
@@ -801,6 +887,8 @@ class FrameProcessor(BaseObject):
801
887
 
802
888
  """
803
889
  while True:
890
+ (frame, direction, callback) = await self.__input_queue.get()
891
+
804
892
  if self.__should_block_system_frames and self.__input_event:
805
893
  logger.trace(f"{self}: system frame processing paused")
806
894
  await self.__input_event.wait()
@@ -808,8 +896,6 @@ class FrameProcessor(BaseObject):
808
896
  self.__should_block_system_frames = False
809
897
  logger.trace(f"{self}: system frame processing resumed")
810
898
 
811
- (frame, direction, callback) = await self.__input_queue.get()
812
-
813
899
  if isinstance(frame, SystemFrame):
814
900
  await self.__process_frame(frame, direction, callback)
815
901
  elif self.__process_queue:
@@ -824,6 +910,8 @@ class FrameProcessor(BaseObject):
824
910
  async def __process_frame_task_handler(self):
825
911
  """Handle non-system frames from the process queue."""
826
912
  while True:
913
+ (frame, direction, callback) = await self.__process_queue.get()
914
+
827
915
  if self.__should_block_frames and self.__process_event:
828
916
  logger.trace(f"{self}: frame processing paused")
829
917
  await self.__process_event.wait()
@@ -831,8 +919,6 @@ class FrameProcessor(BaseObject):
831
919
  self.__should_block_frames = False
832
920
  logger.trace(f"{self}: frame processing resumed")
833
921
 
834
- (frame, direction, callback) = await self.__process_queue.get()
835
-
836
922
  await self.__process_frame(frame, direction, callback)
837
923
 
838
924
  self.__process_queue.task_done()
@@ -12,6 +12,7 @@ from loguru import logger
12
12
 
13
13
  from pipecat.frames.frames import (
14
14
  Frame,
15
+ LLMContextFrame,
15
16
  LLMFullResponseEndFrame,
16
17
  LLMFullResponseStartFrame,
17
18
  TextFrame,
@@ -64,11 +65,16 @@ class LangchainProcessor(FrameProcessor):
64
65
  """
65
66
  await super().process_frame(frame, direction)
66
67
 
67
- if isinstance(frame, OpenAILLMContextFrame):
68
+ if isinstance(frame, (LLMContextFrame, OpenAILLMContextFrame)):
68
69
  # Messages are accumulated on the context as a list of messages.
69
70
  # The last one by the human is the one we want to send to the LLM.
70
71
  logger.debug(f"Got transcription frame {frame}")
71
- text: str = frame.context.messages[-1]["content"]
72
+ messages = (
73
+ frame.context.messages
74
+ if isinstance(frame, OpenAILLMContextFrame)
75
+ else frame.context.get_messages()
76
+ )
77
+ text: str = messages[-1]["content"]
72
78
 
73
79
  await self._ainvoke(text.strip())
74
80
  else: