Merge pull request #161 from pipecat-ai/only-interrupt-assistant

processors: only interrupt asssisstant

CHANGELOG.md

@@ -9,6 +9,11 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- Fixed an issue where `StartInterruptionFrame` would cause
+  `LLMUserResponseAggregator` to push the accumulated text causing the LLM
+  respond in the wrong task. The `StartInterruptionFrame` should not trigger any
+  new LLM response because that would be spoken in a different task.
+
 - Fixed an issue where tasks and threads could be paused because the executor
   didn't have more tasks available. This was causing issues when cancelling and
   recreating tasks during interruptions.

src/pipecat/processors/aggregators/llm_response.py

@@ -30,7 +30,8 @@ class LLMResponseAggregator(FrameProcessor):
         start_frame,
         end_frame,
         accumulator_frame: TextFrame,
-        interim_accumulator_frame: TextFrame | None = None
+        interim_accumulator_frame: TextFrame | None = None,
+        handle_interruptions: bool = False
     ):
         super().__init__()
 
@@ -40,6 +41,7 @@ class LLMResponseAggregator(FrameProcessor):
         self._end_frame = end_frame
         self._accumulator_frame = accumulator_frame
         self._interim_accumulator_frame = interim_accumulator_frame
+        self._handle_interruptions = handle_interruptions
 
         # Reset our accumulator state.
         self._reset()
@@ -101,7 +103,7 @@ class LLMResponseAggregator(FrameProcessor):
             self._seen_interim_results = False
         elif self._interim_accumulator_frame and isinstance(frame, self._interim_accumulator_frame):
             self._seen_interim_results = True
-        elif isinstance(frame, StartInterruptionFrame):
+        elif self._handle_interruptions and isinstance(frame, StartInterruptionFrame):
             await self._push_aggregation()
             # Reset anyways
             self._reset()
@@ -136,7 +138,8 @@ class LLMAssistantResponseAggregator(LLMResponseAggregator):
             role="assistant",
             start_frame=LLMFullResponseStartFrame,
             end_frame=LLMFullResponseEndFrame,
-            accumulator_frame=TextFrame
+            accumulator_frame=TextFrame,
+            handle_interruptions=True
         )
 
 

src/pipecat/processors/aggregators/user_response.py

@@ -110,9 +110,6 @@ class ResponseAggregator(FrameProcessor):
             self._seen_interim_results = False
         elif self._interim_accumulator_frame and isinstance(frame, self._interim_accumulator_frame):
             self._seen_interim_results = True
-        elif isinstance(frame, StartInterruptionFrame):
-            self._reset()
-            await self.push_frame(frame, direction)
         else:
             await self.push_frame(frame, direction)
 

src/pipecat/services/openai.py

@@ -5,7 +5,6 @@
 #
 
 import io
-import json
 import time
 import aiohttp
 import base64
@@ -36,7 +35,6 @@ try:
     from openai import AsyncOpenAI, AsyncStream
 
     from openai.types.chat import (
-        ChatCompletion,
         ChatCompletionChunk,
         ChatCompletionMessageParam,
     )
@@ -99,15 +97,6 @@ class BaseOpenAILLMService(LLMService):
 
         return chunks
 
-    async def _chat_completions(self, messages) -> str | None:
-        response: ChatCompletion = await self._client.chat.completions.create(
-            model=self._model, stream=False, messages=messages
-        )
-        if response and len(response.choices) > 0:
-            return response.choices[0].message.content
-        else:
-            return None
-
     async def _process_context(self, context: OpenAILLMContext):
         function_name = ""
         arguments = ""