wx44wx/pipecat
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Update OpenAIRealtimeLLMService to work with LLMContext and LLMContextAggregatorPair

Browse Source
This commit is contained in:
Paul Kompfner
2025-09-30 14:47:06 -04:00
parent ae22673595
commit 8962263329
1 changed files with 17 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
src/pipecat/services/openai_realtime/openai.py
View File

@@ -41,6 +41,7 @@ from pipecat.frames.frames import (
UserStoppedSpeakingFrame, UserStoppedSpeakingFrame,
) )
from pipecat.metrics.metrics import LLMTokenUsage from pipecat.metrics.metrics import LLMTokenUsage
from pipecat.processors.aggregators.llm_context import LLMContext
from pipecat.processors.aggregators.llm_response import ( from pipecat.processors.aggregators.llm_response import (
LLMAssistantAggregatorParams, LLMAssistantAggregatorParams,
LLMUserAggregatorParams, LLMUserAggregatorParams,
@@ -138,7 +139,15 @@ class OpenAIRealtimeLLMService(LLMService):
self._send_transcription_frames = send_transcription_frames self._send_transcription_frames = send_transcription_frames
self._websocket = None self._websocket = None
self._receive_task = None self._receive_task = None
self._context = None # "Last received context" is only needed while we still support
# OpenAILLMContextFrame. The "last received context" is the context received
# in the most recent OpenAILLMContextFrame or LLMContextFrame, before
# we convert it to an LLMContext if needed. Storing the "last received
# context" lets us determine whether the context has changed. (We can't
# compare contexts after conversion because conversion creates a new
# object.)
self._context: LLMContext = None
self._last_received_context: OpenAILLMContext | LLMContext = None
self._disconnecting = False self._disconnecting = False
self._api_session_ready = False self._api_session_ready = False
@@ -347,22 +356,22 @@ class OpenAIRealtimeLLMService(LLMService):
if isinstance(frame, TranscriptionFrame): if isinstance(frame, TranscriptionFrame):
pass pass
elif isinstance(frame, OpenAILLMContextFrame): elif isinstance(frame, (LLMContextFrame, OpenAILLMContextFrame)):
context: OpenAIRealtimeLLMContext = OpenAIRealtimeLLMContext.upgrade_to_realtime( context = (
frame.context frame.context
if isinstance(frame, LLMContextFrame)
else LLMContext.from_openai_context(frame.context)
) )
if not self._context: if not self._context:
self._last_received_context = frame.context
self._context = context self._context = context
elif frame.context is not self._context: elif frame.context is not self._last_received_context:
# If the context has changed, reset the conversation # If the context has changed, reset the conversation
self._last_received_context = frame.context
self._context = context self._context = context
await self.reset_conversation() await self.reset_conversation()
# Run the LLM at next opportunity # Run the LLM at next opportunity
await self._create_response() await self._create_response()
elif isinstance(frame, LLMContextFrame):
raise NotImplementedError(
"Universal LLMContext is not yet supported for OpenAI Realtime."
)
elif isinstance(frame, InputAudioRawFrame): elif isinstance(frame, InputAudioRawFrame):
if not self._audio_input_paused: if not self._audio_input_paused:
await self._send_user_audio(frame) await self._send_user_audio(frame)