Support for passing in a ToolsSchema in lieu of a list of provider-specific dicts when updating OpenAIRealtimeLLMService using LLMUpdateSettingsFrame.
This commit is contained in:
@@ -10,7 +10,8 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
|
|||||||
### Added
|
### Added
|
||||||
|
|
||||||
- Added support for passing in a `ToolsSchem` in lieu of a list of provider-
|
- Added support for passing in a `ToolsSchem` in lieu of a list of provider-
|
||||||
specific dicts when initializing `OpenAIRealtimeLLMService`.
|
specific dicts when initializing `OpenAIRealtimeLLMService` or when updating
|
||||||
|
it using `LLMUpdateSettingsFrame`.
|
||||||
|
|
||||||
- Added `TransportParams.audio_out_silence_secs`, which specifies how many
|
- Added `TransportParams.audio_out_silence_secs`, which specifies how many
|
||||||
seconds of silence to output when an `EndFrame` reaches the output
|
seconds of silence to output when an `EndFrame` reaches the output
|
||||||
|
|||||||
@@ -14,8 +14,14 @@ from loguru import logger
|
|||||||
|
|
||||||
from pipecat.adapters.schemas.function_schema import FunctionSchema
|
from pipecat.adapters.schemas.function_schema import FunctionSchema
|
||||||
from pipecat.adapters.schemas.tools_schema import ToolsSchema
|
from pipecat.adapters.schemas.tools_schema import ToolsSchema
|
||||||
|
from pipecat.adapters.services.open_ai_realtime_adapter import OpenAIRealtimeLLMAdapter
|
||||||
from pipecat.audio.vad.silero import SileroVADAnalyzer
|
from pipecat.audio.vad.silero import SileroVADAnalyzer
|
||||||
from pipecat.frames.frames import LLMRunFrame, LLMSetToolsFrame, TranscriptionMessage
|
from pipecat.frames.frames import (
|
||||||
|
LLMRunFrame,
|
||||||
|
LLMSetToolsFrame,
|
||||||
|
LLMUpdateSettingsFrame,
|
||||||
|
TranscriptionMessage,
|
||||||
|
)
|
||||||
from pipecat.observers.loggers.transcription_log_observer import TranscriptionLogObserver
|
from pipecat.observers.loggers.transcription_log_observer import TranscriptionLogObserver
|
||||||
from pipecat.pipeline.pipeline import Pipeline
|
from pipecat.pipeline.pipeline import Pipeline
|
||||||
from pipecat.pipeline.runner import PipelineRunner
|
from pipecat.pipeline.runner import PipelineRunner
|
||||||
@@ -148,6 +154,8 @@ async def run_bot(transport: BaseTransport, runner_args: RunnerArguments):
|
|||||||
noise_reduction=InputAudioNoiseReduction(type="near_field"),
|
noise_reduction=InputAudioNoiseReduction(type="near_field"),
|
||||||
)
|
)
|
||||||
),
|
),
|
||||||
|
# In this example we provide tools through the context, but you could
|
||||||
|
# alternatively provide them here.
|
||||||
# tools=tools,
|
# tools=tools,
|
||||||
instructions="""You are a helpful and friendly AI.
|
instructions="""You are a helpful and friendly AI.
|
||||||
|
|
||||||
@@ -223,6 +231,10 @@ Remember, your responses should be short. Just one or two sentences, usually. Re
|
|||||||
standard_tools=[weather_function, restaurant_function, get_news_function]
|
standard_tools=[weather_function, restaurant_function, get_news_function]
|
||||||
)
|
)
|
||||||
await task.queue_frames([LLMSetToolsFrame(tools=new_tools)])
|
await task.queue_frames([LLMSetToolsFrame(tools=new_tools)])
|
||||||
|
# Alternative pattern, useful if you're changing other session properties too
|
||||||
|
# await task.queue_frames(
|
||||||
|
# [LLMUpdateSettingsFrame(settings=SessionProperties(tools=new_tools).model_dump())]
|
||||||
|
# )
|
||||||
|
|
||||||
@transport.event_handler("on_client_disconnected")
|
@transport.event_handler("on_client_disconnected")
|
||||||
async def on_client_disconnected(transport, client):
|
async def on_client_disconnected(transport, client):
|
||||||
|
|||||||
@@ -198,9 +198,9 @@ class SessionProperties(BaseModel):
|
|||||||
output_modalities: Optional[List[Literal["text", "audio"]]] = None
|
output_modalities: Optional[List[Literal["text", "audio"]]] = None
|
||||||
instructions: Optional[str] = None
|
instructions: Optional[str] = None
|
||||||
audio: Optional[AudioConfiguration] = None
|
audio: Optional[AudioConfiguration] = None
|
||||||
# Tools can only be ToolsSchema when provided by user in
|
# Tools can only be ToolsSchema when provided by the user, in either the
|
||||||
# OpenAIRealtimeLLMService constructor. We'll never serialize/deserialize
|
# OpenAIRealtimeLLMService constructor or through LLMUpdateSettingsFrame.
|
||||||
# ToolsSchema.
|
# We'll never serialize/deserialize ToolsSchema when talking to the server.
|
||||||
tools: Optional[ToolsSchema | List[Dict]] = None
|
tools: Optional[ToolsSchema | List[Dict]] = None
|
||||||
tool_choice: Optional[Literal["auto", "none", "required"]] = None
|
tool_choice: Optional[Literal["auto", "none", "required"]] = None
|
||||||
max_output_tokens: Optional[Union[int, Literal["inf"]]] = None
|
max_output_tokens: Optional[Union[int, Literal["inf"]]] = None
|
||||||
|
|||||||
@@ -156,16 +156,6 @@ class OpenAIRealtimeLLMService(LLMService):
|
|||||||
self._session_properties: events.SessionProperties = (
|
self._session_properties: events.SessionProperties = (
|
||||||
session_properties or events.SessionProperties()
|
session_properties or events.SessionProperties()
|
||||||
)
|
)
|
||||||
# If needed, map session_properties.tools from ToolsSchema to list of
|
|
||||||
# dicts, which remote server expects
|
|
||||||
if self._session_properties.tools and isinstance(
|
|
||||||
self._session_properties.tools, ToolsSchema
|
|
||||||
):
|
|
||||||
adapter = self.get_llm_adapter()
|
|
||||||
self._session_properties.tools = adapter.from_standard_tools(
|
|
||||||
self._session_properties.tools
|
|
||||||
)
|
|
||||||
|
|
||||||
self._audio_input_paused = start_audio_paused
|
self._audio_input_paused = start_audio_paused
|
||||||
self._websocket = None
|
self._websocket = None
|
||||||
self._receive_task = None
|
self._receive_task = None
|
||||||
@@ -492,9 +482,9 @@ class OpenAIRealtimeLLMService(LLMService):
|
|||||||
|
|
||||||
async def _update_settings(self):
|
async def _update_settings(self):
|
||||||
settings = self._session_properties
|
settings = self._session_properties
|
||||||
|
adapter: OpenAIRealtimeLLMAdapter = self.get_llm_adapter()
|
||||||
|
|
||||||
if self._context:
|
if self._context:
|
||||||
adapter: OpenAIRealtimeLLMAdapter = self.get_llm_adapter()
|
|
||||||
llm_invocation_params = adapter.get_llm_invocation_params(self._context)
|
llm_invocation_params = adapter.get_llm_invocation_params(self._context)
|
||||||
|
|
||||||
# tools given in the context override the tools in the session properties
|
# tools given in the context override the tools in the session properties
|
||||||
@@ -506,6 +496,12 @@ class OpenAIRealtimeLLMService(LLMService):
|
|||||||
if llm_invocation_params["system_instruction"]:
|
if llm_invocation_params["system_instruction"]:
|
||||||
settings.instructions = llm_invocation_params["system_instruction"]
|
settings.instructions = llm_invocation_params["system_instruction"]
|
||||||
|
|
||||||
|
# If needed, map settings.tools from ToolsSchema to list of dicts,
|
||||||
|
# which remote server expects. It would only be a ToolsSchema if that's
|
||||||
|
# how it was provided in the constructor or via LLMUpdateSettingsFrame.
|
||||||
|
if settings.tools and isinstance(settings.tools, ToolsSchema):
|
||||||
|
settings.tools = adapter.from_standard_tools(settings.tools)
|
||||||
|
|
||||||
await self.send_client_event(events.SessionUpdateEvent(session=settings))
|
await self.send_client_event(events.SessionUpdateEvent(session=settings))
|
||||||
|
|
||||||
#
|
#
|
||||||
|
|||||||
Reference in New Issue
Block a user