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

Remove unnecessary system_instruction argument from run_inference() methods

Browse Source
This commit is contained in:
Paul Kompfner
2025-09-04 12:31:33 -04:00
parent 0826d72f74
commit f72ca2fd7d
6 changed files with 10 additions and 46 deletions
Download Patch File Download Diff File Expand all files Collapse all files

12
src/pipecat/pipeline/llm_switcher.py
View File

@@ -30,25 +30,17 @@ class LLMSwitcher(ServiceSwitcher[StrategyType]):
"""Get the currently active LLM, if any.""" """Get the currently active LLM, if any."""
return self.strategy.active_service return self.strategy.active_service
async def run_inference( async def run_inference(self, context: LLMContext) -> Optional[str]:
self, context: LLMContext, system_instruction: Optional[str] = None
) -> Optional[str]:
"""Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context, using the currently active LLM. """Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context, using the currently active LLM.
Args: Args:
context: The LLM context containing conversation history. context: The LLM context containing conversation history.
system_instruction: Optional system instruction to guide the LLM's
behavior. You could also (again, optionally) provide a system
instruction directly in the context. If both are provided, the
one in the context takes precedence.
Returns: Returns:
The LLM's response as a string, or None if no response is generated. The LLM's response as a string, or None if no response is generated.
""" """
if self.active_llm: if self.active_llm:
return await self.active_llm.run_inference( return await self.active_llm.run_inference(context=context)
context=context, system_instruction=system_instruction
)
return None return None
def register_function( def register_function(

10
src/pipecat/services/anthropic/llm.py
View File

@@ -228,17 +228,11 @@ class AnthropicLLMService(LLMService):
response = await api_call(**params) response = await api_call(**params)
return response return response
async def run_inference( async def run_inference(self, context: LLMContext | OpenAILLMContext) -> Optional[str]:
self, context: LLMContext | OpenAILLMContext, system_instruction: Optional[str] = None
) -> Optional[str]:
"""Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context. """Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context.
Args: Args:
context: The LLM context containing conversation history. context: The LLM context containing conversation history.
system_instruction: Optional system instruction to guide the LLM's
behavior. You could also (again, optionally) provide a system
instruction directly in the context. If both are provided, the
one in the context takes precedence.
Returns: Returns:
The LLM's response as a string, or None if no response is generated. The LLM's response as a string, or None if no response is generated.
@@ -255,7 +249,7 @@ class AnthropicLLMService(LLMService):
else: else:
context = AnthropicLLMContext.upgrade_to_anthropic(context) context = AnthropicLLMContext.upgrade_to_anthropic(context)
messages = context.messages messages = context.messages
system = getattr(context, "system", None) or system_instruction or NOT_GIVEN system = getattr(context, "system", NOT_GIVEN)
# LLM completion # LLM completion
response = await self._client.messages.create( response = await self._client.messages.create(

10
src/pipecat/services/aws/llm.py
View File

@@ -792,17 +792,11 @@ class AWSBedrockLLMService(LLMService):
""" """
return True return True
async def run_inference( async def run_inference(self, context: LLMContext | OpenAILLMContext) -> Optional[str]:
self, context: LLMContext | OpenAILLMContext, system_instruction: Optional[str] = None
) -> Optional[str]:
"""Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context. """Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context.
Args: Args:
context: The LLM context containing conversation history. context: The LLM context containing conversation history.
system_instruction: Optional system instruction to guide the LLM's
behavior. You could also (again, optionally) provide a system
instruction directly in the context. If both are provided, the
one in the context takes precedence.
Returns: Returns:
The LLM's response as a string, or None if no response is generated. The LLM's response as a string, or None if no response is generated.
@@ -822,7 +816,7 @@ class AWSBedrockLLMService(LLMService):
else: else:
context = AWSBedrockLLMContext.upgrade_to_bedrock(context) context = AWSBedrockLLMContext.upgrade_to_bedrock(context)
messages = context.messages messages = context.messages
system = getattr(context, "system", None) or system_instruction system = getattr(context, "system", None)
# Determine if we're using Claude or Nova based on model ID # Determine if we're using Claude or Nova based on model ID
model_id = self.model_name model_id = self.model_name

10
src/pipecat/services/google/llm.py
View File

@@ -733,17 +733,11 @@ class GoogleLLMService(LLMService):
def _create_client(self, api_key: str, http_options: Optional[HttpOptions] = None): def _create_client(self, api_key: str, http_options: Optional[HttpOptions] = None):
self._client = genai.Client(api_key=api_key, http_options=http_options) self._client = genai.Client(api_key=api_key, http_options=http_options)
async def run_inference( async def run_inference(self, context: LLMContext | OpenAILLMContext) -> Optional[str]:
self, context: LLMContext | OpenAILLMContext, system_instruction: Optional[str] = None
) -> Optional[str]:
"""Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context. """Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context.
Args: Args:
context: The LLM context containing conversation history. context: The LLM context containing conversation history.
system_instruction: Optional system instruction to guide the LLM's
behavior. You could also (again, optionally) provide a system
instruction directly in the context. If both are provided, the
one in the context takes precedence.
Returns: Returns:
The LLM's response as a string, or None if no response is generated. The LLM's response as a string, or None if no response is generated.
@@ -758,7 +752,7 @@ class GoogleLLMService(LLMService):
else: else:
context = GoogleLLMContext.upgrade_to_google(context) context = GoogleLLMContext.upgrade_to_google(context)
messages = context.messages messages = context.messages
system = getattr(context, "system_message", None) or system_instruction system = getattr(context, "system_message", None)
generation_config = GenerateContentConfig(system_instruction=system) generation_config = GenerateContentConfig(system_instruction=system)

7
src/pipecat/services/llm_service.py
View File

@@ -195,18 +195,13 @@ class LLMService(AIService):
""" """
return self._adapter return self._adapter
async def run_inference( async def run_inference(self, context: LLMContext | OpenAILLMContext) -> Optional[str]:
self, context: LLMContext | OpenAILLMContext, system_instruction: Optional[str] = None
) -> Optional[str]:
"""Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context. """Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context.
Must be implemented by subclasses. Must be implemented by subclasses.
Args: Args:
context: The LLM context containing conversation history. context: The LLM context containing conversation history.
system_instruction: Optional system instruction to guide the LLM's
behavior. You could also (again, optionally) provide a system
instruction directly in the context.
Returns: Returns:
The LLM's response as a string, or None if no response is generated. The LLM's response as a string, or None if no response is generated.

7
src/pipecat/services/openai/base_llm.py
View File

@@ -245,16 +245,11 @@ class BaseOpenAILLMService(LLMService):
params.update(self._settings["extra"]) params.update(self._settings["extra"])
return params return params
async def run_inference( async def run_inference(self, context: LLMContext | OpenAILLMContext) -> Optional[str]:
self, context: LLMContext | OpenAILLMContext, system_instruction: Optional[str] = None
) -> Optional[str]:
"""Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context. """Run a one-shot, out-of-band (i.e. out-of-pipeline) inference with the given LLM context.
Args: Args:
context: The LLM context containing conversation history. context: The LLM context containing conversation history.
system_instruction: Optional system instruction to guide the LLM's
behavior. You could also (again, optionally) provide a system
instruction directly in the context.
Returns: Returns:
The LLM's response as a string, or None if no response is generated. The LLM's response as a string, or None if no response is generated.