TranscriptProcessor: send TranscriptionUpdateFrame after interruption

CHANGELOG.md

@@ -73,6 +73,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- Fixed an issue that could cause the `TranscriptionUpdateFrame` being pushed
+  because of an interruption to be discarded.
+
 - Fixed an issue that would cause `SegmentedSTTService` based services
   (e.g. `OpenAISTTService`) to try to transcribe non-spoken audio, causing
   invalid transcriptions.

src/pipecat/processors/transcript_processor.py

@@ -175,22 +175,28 @@ class AssistantTranscriptProcessor(BaseTranscriptProcessor):
         """
         await super().process_frame(frame, direction)
 
-        if isinstance(frame, TTSTextFrame):
+        if isinstance(frame, (StartInterruptionFrame, CancelFrame)):
+            # Push frame first otherwise our emitted transcription update frame
+            # might get cleaned up.
+            await self.push_frame(frame, direction)
+            # Emit accumulated text with interruptions
+            await self._emit_aggregated_text()
+        elif isinstance(frame, TTSTextFrame):
             # Start timestamp on first text part
             if not self._aggregation_start_time:
                 self._aggregation_start_time = time_now_iso8601()
 
             self._current_text_parts.append(frame.text)
 
-        elif isinstance(frame, (BotStoppedSpeakingFrame, StartInterruptionFrame, CancelFrame)):
+            # Push frame.
-            # Emit accumulated text when bot finishes speaking or is interrupted
+            await self.push_frame(frame, direction)
+        elif isinstance(frame, (BotStoppedSpeakingFrame, EndFrame)):
+            # Emit accumulated text when bot finishes speaking or pipeline ends.
             await self._emit_aggregated_text()
-
+            # Push frame.
-        elif isinstance(frame, EndFrame):
+            await self.push_frame(frame, direction)
-            # Emit any remaining text when pipeline ends
+        else:
-            await self._emit_aggregated_text()
+            await self.push_frame(frame, direction)
-
-        await self.push_frame(frame, direction)
 
 
 class TranscriptProcessor:

tests/test_context_aggregators.py

@@ -41,7 +41,10 @@ from pipecat.services.google.llm import (
     GoogleLLMContext,
     GoogleUserContextAggregator,
 )
-from pipecat.services.openai import OpenAIAssistantContextAggregator, OpenAIUserContextAggregator
+from pipecat.services.openai.llm import (
+    OpenAIAssistantContextAggregator,
+    OpenAIUserContextAggregator,
+)
 from pipecat.tests.utils import SleepFrame, run_test
 
 AGGREGATION_TIMEOUT = 0.1

tests/test_transcript_processor.py

@@ -238,8 +238,8 @@ class TestUserTranscriptProcessor(unittest.IsolatedAsyncioTestCase):
             TTSTextFrame(text="world!"),
             SleepFrame(sleep=0.1),
             StartInterruptionFrame(),  # User interrupts here
-            BotStartedSpeakingFrame(),
             SleepFrame(sleep=0.1),
+            BotStartedSpeakingFrame(),
             TTSTextFrame(text="New"),
             TTSTextFrame(text="response"),
             SleepFrame(sleep=0.1),
@@ -251,8 +251,8 @@ class TestUserTranscriptProcessor(unittest.IsolatedAsyncioTestCase):
             BotStartedSpeakingFrame,
             TTSTextFrame,  # "Hello"
             TTSTextFrame,  # "world!"
+            StartInterruptionFrame,
             TranscriptionUpdateFrame,  # First message (emitted due to interruption)
-            StartInterruptionFrame,  # Interruption frame comes after the update
             BotStartedSpeakingFrame,
             TTSTextFrame,  # "New"
             TTSTextFrame,  # "response"