microsoft/semantic-kernel
1
0
Fork 0
mirror of https://github.com/microsoft/semantic-kernel.git synced 2026-10-12 11:19:56 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
8b16af69c0effc76e8c69bcea741e650c828c68d
semantic-kernel/python/semantic_kernel/functions/kernel_function.py
T
Evan Mattson 8b16af69c0 Python: Remove the memory tied to the kernel. Update kernel examples and notebooks. (#5143) 
### Motivation and Context

As we move toward v1, it is necessary to remove the single kernel memory
attribute so that memory connectors can be handled via plugins. This
change also refactors the `import_plugin` method name to be similar with
dotnet as `import_plugin_from_object`, and introduces a simple
`kernel.invoke_prompt` method.
<!-- Thank you for your contribution to the semantic-kernel repo!
Please help reviewers and future users, providing the following
information:
  1. Why is this change required?
  2. What problem does it solve?
  3. What scenario does it contribute to?
  4. If it fixes an open issue, please link to the issue here.
-->

### Description

Updates are:
- The memory connector is now passed to the `SemanticTextMemory` class,
along with the embedding generator. This allows the user to add memory
to multiple plugins/services as needed. Closes #4632
- Kernel Examples and Notebooks were updated to reflect these changes.
Now using Service Enum for all notebooks. Closes #4557
- There were planners using memory related to the kernel, to store the
functions, and many of these planners are non-existent in dotnet's v1
SDK. Python has pending work to update the planner/remove deprecated
planners before getting to v1.
- A Kernel method called `invoke_prompt` was introduced to invoke a
simple prompt, and return the `FunctionResult` when complete.
- All unit/integration tests are passing.

<!-- Describe your changes, the overall approach, the underlying design.
These notes will help understanding how your code works. Thanks! -->

### Contribution Checklist

<!-- Before submitting this PR, please make sure: -->

- [X] The code builds clean without any errors or warnings
- [X] The PR follows the [SK Contribution
Guidelines](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md)
and the [pre-submission formatting
script](https://github.com/microsoft/semantic-kernel/blob/main/CONTRIBUTING.md#development-scripts)
raises no violations
- [X] All unit tests pass, and I have added new tests where possible
- [ ] I didn't break anyone 😄
2024-02-26 17:50:23 +00:00

582 lines
25 KiB
Python
Raw Blame History

# Copyright (c) Microsoft. All rights reserved.
import asyncio
import logging
import platform
import sys
from functools import wraps
from inspect import isawaitable
from typing import TYPE_CHECKING, Any, AsyncIterable, Callable, Dict, List, Optional, Union
from pydantic import Field, ValidationError, field_validator
from semantic_kernel.kernel_pydantic import KernelBaseModel
if sys.version_info >= (3, 9):
pass
else:
pass
from semantic_kernel.connectors.ai.chat_completion_client_base import (
ChatCompletionClientBase,
)
from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
from semantic_kernel.connectors.ai.text_completion_client_base import (
TextCompletionClientBase,
)
from semantic_kernel.contents.chat_history import ChatHistory
from semantic_kernel.contents.streaming_kernel_content import StreamingKernelContent
from semantic_kernel.functions.function_result import FunctionResult
from semantic_kernel.functions.kernel_arguments import KernelArguments
from semantic_kernel.functions.kernel_function_metadata import KernelFunctionMetadata
from semantic_kernel.functions.kernel_parameter_metadata import KernelParameterMetadata
from semantic_kernel.prompt_template.kernel_prompt_template import KernelPromptTemplate
from semantic_kernel.prompt_template.prompt_template_base import PromptTemplateBase
from semantic_kernel.prompt_template.prompt_template_config import PromptTemplateConfig
from semantic_kernel.utils.naming import generate_random_ascii_name
if TYPE_CHECKING:
from semantic_kernel.kernel import Kernel
# TODO: is this needed anymore after sync code removal?
if platform.system() == "Windows" and sys.version_info >= (3, 8, 0):
asyncio.set_event_loop_policy(asyncio.WindowsSelectorEventLoopPolicy())
logger: logging.Logger = logging.getLogger(__name__)
class KernelFunction(KernelBaseModel):
"""
Semantic Kernel function.
Attributes:
name (str): The name of the function. Must be upper/lower case letters and
underscores with a minimum length of 1.
plugin_name (str): The name of the plugin that contains this function. Must be upper/lower
case letters and underscores with a minimum length of 1.
description (Optional[str]): The description of the function.
is_prompt (bool): Whether the function is semantic.
stream_function (Optional[Callable[..., Any]]): The stream function for the function.
parameters (List[KernelParameterMetadata]): The parameters for the function.
return_parameter (Optional[KernelParameterMetadata]): The return parameter for the function.
function (Callable[..., Any]): The function to call.
prompt_execution_settings (PromptExecutionSettings): The AI prompt execution settings.
prompt_template_config (PromptTemplateConfig): The prompt template configuration.
metadata (Optional[KernelFunctionMetadata]): The metadata for the function.
"""
# some attributes are now properties, still listed here for documentation purposes
metadata: KernelFunctionMetadata
function: Callable[..., Any]
stream_function: Optional[Callable[..., Any]] = None
prompt_template_config: Optional[PromptTemplateConfig] = Field(default_factory=PromptTemplateConfig)
prompt_execution_settings: Dict[str, PromptExecutionSettings] = Field(default_factory=dict)
def __init__(
self,
function_name: str,
plugin_name: str,
description: str,
function: Callable[..., Any],
is_prompt: bool,
parameters: List[KernelParameterMetadata],
return_parameter: Optional[KernelParameterMetadata] = None,
stream_function: Optional[Callable[..., Any]] = None,
prompt_execution_settings: Optional[
Union[PromptExecutionSettings, List[PromptExecutionSettings], Dict[str, PromptExecutionSettings]]
] = None,
prompt_template_config: Optional[PromptTemplateConfig] = None,
) -> None:
"""
Initializes a new instance of the KernelFunction class
Args:
function_name (str): The name of the function
plugin_name (str): The name of the plugin
description (str): The description for the function
function (Callable[..., Any]): The function
is_prompt (bool): Whether the function is semantic
parameters (List[KernelParameterMetadata]): The parameters for the function
return_parameter (Optional[KernelParameterMetadata]): The return parameter of the function
stream_function (Optional[Callable[..., Any]]): The stream function for the function
prompt_execution_settings (Optional): instance, list or dict of PromptExecutionSettings to be used by the
function, can also be supplied through prompt_template_config,
but the supplied one is used if both are present.
prompt_template_config (Optional[PromptTemplateConfig]): the prompt template config.
"""
try:
metadata = KernelFunctionMetadata(
name=function_name,
description=description,
parameters=parameters,
return_parameter=return_parameter,
is_prompt=is_prompt,
plugin_name=plugin_name,
)
except ValidationError as exc:
# reraise the exception to clarify it comes from KernelFunction init
raise exc
args: Dict[str, Any] = {
"metadata": metadata,
"function": function,
}
if stream_function:
args["stream_function"] = stream_function
if return_parameter:
args["return_parameter"] = return_parameter
if prompt_template_config:
args["prompt_template_config"] = prompt_template_config
if prompt_execution_settings:
args["prompt_execution_settings"] = prompt_execution_settings
elif prompt_template_config:
args["prompt_execution_settings"] = prompt_template_config.execution_settings
super().__init__(**args)
@field_validator("prompt_execution_settings", mode="before")
@classmethod
def rewrite_execution_settings(
cls,
prompt_execution_settings: Optional[
Union[PromptExecutionSettings, List[PromptExecutionSettings], Dict[str, PromptExecutionSettings]]
],
) -> Dict[str, PromptExecutionSettings]:
"""Rewrite execution settings to a dictionary."""
if not prompt_execution_settings:
return {}
if isinstance(prompt_execution_settings, PromptExecutionSettings):
return {prompt_execution_settings.service_id or "default": prompt_execution_settings}
if isinstance(prompt_execution_settings, list):
return {s.service_id or "default": s for s in prompt_execution_settings}
return prompt_execution_settings
@property
def name(self) -> str:
return self.metadata.name
@property
def plugin_name(self) -> str:
return self.metadata.plugin_name
@property
def description(self) -> Optional[str]:
return self.metadata.description
@property
def is_prompt(self) -> bool:
return self.metadata.is_prompt
@property
def parameters(self) -> List[KernelParameterMetadata]:
return self.metadata.parameters
@property
def return_parameter(self) -> Optional[KernelParameterMetadata]:
return self.metadata.return_parameter
@staticmethod
def from_native_method(method: Callable[..., Any], plugin_name: str) -> "KernelFunction":
"""
Create a KernelFunction from a method.
Args:
method (Callable[..., Any]): The method to create the function from
plugin_name (str): The name of the plugin
Returns:
KernelFunction: The kernel function
"""
if method is None:
raise ValueError("Method cannot be `None`")
if not hasattr(method, "__kernel_function__") or method.__kernel_function__ is None:
raise ValueError("Method is not a Kernel function")
parameters: List["KernelParameterMetadata"] = []
if hasattr(method, "__kernel_function_context_parameters__"):
for param in method.__kernel_function_context_parameters__:
assert "name" in param, "Parameter name is empty"
assert "description" in param, "Parameter description is empty"
assert "default_value" in param, "Parameter default value is empty"
parameters.append(
KernelParameterMetadata(
name=param["name"],
description=param["description"],
default_value=param["default_value"],
type=param.get("type", "str"),
required=param.get("required", False),
expose=True,
)
)
return_param = KernelParameterMetadata(
name="return",
description=method.__kernel_function_return_description__
if hasattr(method, "__kernel_function_return_description__")
else "",
default_value=None,
type=method.__kernel_function_return_type__
if hasattr(method, "__kernel_function_return_type__")
else "None",
required=method.__kernel_function_return_required__
if hasattr(method, "__kernel_function_return_required__")
else False,
)
function_name = method.__kernel_function_name__
description = method.__kernel_function_description__
if getattr(method, "__kernel_function_streaming__", False):
streaming_method = method
@wraps(method)
async def _non_streaming_function(*args, **kwargs):
return [x async for x in streaming_method(*args, **kwargs)]
method = _non_streaming_function
else:
streaming_method = None
return KernelFunction(
function_name=function_name or f"f_{generate_random_ascii_name()}",
plugin_name=plugin_name or f"p_{generate_random_ascii_name()}",
description=description,
function=method,
parameters=parameters,
return_parameter=return_param,
stream_function=streaming_method,
is_prompt=False,
)
@staticmethod
def from_prompt(
prompt: str,
function_name: Optional[str] = None,
plugin_name: Optional[str] = None,
description: Optional[str] = None,
template_format: Optional[str] = None,
prompt_template: Optional[PromptTemplateBase] = None,
prompt_execution_settings: Optional[PromptExecutionSettings] = None,
prompt_template_config: Optional[PromptTemplateConfig] = None,
) -> "KernelFunction":
"""
Create a Kernel Function from a prompt
Args:
prompt (str): The prompt
execution_settings (Optional[PromptExecutionSettings]): The execution settings
plugin_name (Optional[str]): The name of the plugin
function_name (Optional[str]): The name of the function
description (Optional[str]): The description of the function
template_format (Optional[str]): The template format
prompt_template (Optional[PromptTemplateBase]): The prompt template
prompt_template_config (Optional[PromptTemplateConfig]): The prompt template configuration
Returns:
KernelFunction: The kernel function
"""
if prompt_template and not template_format:
raise ValueError(f"Template format cannot be `None` when providing a {prompt_template}")
if not prompt_template_config:
prompt_template_config = PromptTemplateConfig(
name=function_name,
description=description if description else "Semantic function, unknown purpose",
template=prompt,
template_format=template_format if template_format else "semantic-kernel",
execution_settings=prompt_execution_settings
if prompt_execution_settings
else PromptExecutionSettings(),
)
if not prompt_template:
prompt_template = KernelPromptTemplate(prompt_template_config=prompt_template_config)
async def _local_func(
function: KernelFunctionMetadata,
kernel: "Kernel",
service: Union[TextCompletionClientBase, ChatCompletionClientBase],
execution_settings: PromptExecutionSettings,
arguments: KernelArguments,
) -> "FunctionResult":
if service is None:
raise ValueError("AI LLM service cannot be `None`")
prompt = await prompt_template.render(kernel, arguments)
if isinstance(service, ChatCompletionClientBase):
chat_history = ChatHistory.from_rendered_prompt(prompt, service.get_chat_message_content_class())
try:
completions = await service.complete_chat(chat_history, execution_settings)
# this can be expanded for auto-invoking function calls
# store_results function is in utils.chat
# chat_history = store_results(chat_history=chat_history, results=completions)
except Exception as exc:
logger.error(f"Error occurred while invoking function {function.name}: {exc}")
raise exc
return FunctionResult(
function=function,
value=completions,
metadata={
"messages": chat_history,
"arguments": arguments,
"metadata": [completion.metadata for completion in completions],
},
)
if isinstance(service, TextCompletionClientBase):
try:
completions = await service.complete(prompt, execution_settings)
except Exception as e:
logger.error(f"Error occurred while invoking function {function.name}: {e}")
raise e
return FunctionResult(
function=function,
value=completions,
metadata={
"prompt": prompt,
"arguments": arguments,
"metadata": [completion.metadata for completion in completions],
},
)
raise ValueError(f"Service `{type(service)}` is not a valid AI service")
async def _local_stream_func(
function: KernelFunctionMetadata,
kernel: "Kernel",
service: Union[TextCompletionClientBase, ChatCompletionClientBase],
execution_settings: PromptExecutionSettings,
arguments: KernelArguments,
) -> AsyncIterable[Union[FunctionResult, List[StreamingKernelContent]]]:
if service is None:
raise ValueError("AI LLM service cannot be `None`")
prompt = await prompt_template.render(kernel, arguments)
if isinstance(service, ChatCompletionClientBase):
chat_history = ChatHistory.from_rendered_prompt(prompt, service.get_chat_message_content_class())
try:
async for partial_content in service.complete_chat_stream(
chat_history=chat_history, settings=execution_settings
):
yield partial_content
return
except Exception as e:
logger.error(f"Error occurred while invoking function {function.name}: {e}")
yield FunctionResult(function=function, value=None, metadata={"exception": e})
if isinstance(service, TextCompletionClientBase):
try:
async for partial_content in service.complete_stream(prompt, execution_settings):
yield partial_content
return
except Exception as e:
logger.error(f"Error occurred while invoking function {function.name}: {e}")
yield FunctionResult(function=function, value=None, metadata={"exception": e})
semantic_function_params = [
KernelParameterMetadata(
name="function",
description="The function to execute",
default_value=None,
type="KernelFunctionMetadata",
required=True,
expose=False,
),
KernelParameterMetadata(
name="kernel",
description="The kernel",
default_value=None,
type="Kernel",
required=True,
expose=False,
),
KernelParameterMetadata(
name="service",
description="The AI service client",
default_value=None,
type="AIServiceClientBase",
required=True,
expose=False,
),
KernelParameterMetadata(
name="execution_settings",
description="The execution settings",
default_value=None,
type="PromptExecutionSettings",
required=True,
expose=False,
),
KernelParameterMetadata(
name="arguments",
description="The kernel arguments",
default_value=None,
type="KernelArguments",
required=True,
expose=False,
),
]
semantic_function_params.extend(prompt_template_config.get_kernel_parameter_metadata())
return KernelFunction(
function_name=function_name or prompt_template_config.name or f"f_{generate_random_ascii_name()}",
plugin_name=plugin_name or f"p_{generate_random_ascii_name()}",
description=description or prompt_template_config.description or "Semantic function, unknown purpose",
function=_local_func,
parameters=semantic_function_params,
return_parameter=KernelParameterMetadata(
name="return",
description="The completion result",
default_value=None,
type="FunctionResult",
required=True,
),
stream_function=_local_stream_func,
is_prompt=True,
prompt_template_config=prompt_template_config,
prompt_execution_settings=prompt_execution_settings,
)
async def __call__(
self,
kernel: "Kernel",
arguments: Optional[KernelArguments] = None,
**kwargs: Dict[str, Any],
) -> "FunctionResult":
"""Invoke the function with the given arguments.
Args:
kernel (Kernel): The kernel
arguments (Optional[KernelArguments]): The Kernel arguments.
Optional, defaults to None.
kwargs (Dict[str, Any]): Additional keyword arguments that will be
Returns:
FunctionResult: The result of the function
"""
return await self.invoke(kernel, arguments, **kwargs)
async def invoke(
self,
kernel: "Kernel",
arguments: Optional[KernelArguments] = None,
**kwargs: Dict[str, Any],
) -> "FunctionResult":
"""Invoke the function with the given arguments.
Args:
kernel (Kernel): The kernel
arguments (KernelArguments): The Kernel arguments
kwargs (Dict[str, Any]): Additional keyword arguments that will be
added to the KernelArguments.
Returns:
FunctionResult: The result of the function
"""
if not arguments:
arguments = KernelArguments(**kwargs)
function_arguments = self.gather_function_parameters(kernel, arguments)
logger.debug("Invoking %s with arguments: %s", self.name, function_arguments)
try:
result = self.function(**function_arguments)
if isawaitable(result):
result = await result
except Exception as exc:
logger.error(f"Error occurred while invoking function {self.name}: {exc}")
return FunctionResult(
function=self.metadata, value=None, metadata={"error": exc, "arguments": function_arguments}
)
logger.debug("Function result: %s", result)
logger.debug("Function result type %s", type(result))
if self.return_parameter and self.return_parameter.type_ and "FunctionResult" in self.return_parameter.type_:
return result
return FunctionResult(function=self.metadata, value=result, metadata={"arguments": function_arguments})
async def invoke_stream(
self,
kernel: "Kernel",
arguments: Optional[KernelArguments] = None,
**kwargs: Dict[str, Any],
) -> AsyncIterable[Union[FunctionResult, List[Union[StreamingKernelContent, Any]]]]:
"""
Invoke a stream async function with the given arguments.
Args:
kernel (Kernel): The kernel
arguments (KernelArguments): The Kernel arguments
kwargs (Dict[str, Any]): Additional keyword arguments that will be
added to the KernelArguments.
Yields:
StreamingKernelContent or FunctionResult -- The results of the function,
if there is an error a FunctionResult is yielded.
"""
if not arguments:
arguments = KernelArguments(**kwargs)
if not self.stream_function:
raise ValueError("Function does not support streaming")
function_arguments = self.gather_function_parameters(kernel, arguments)
logger.debug("Invoking %s with arguments: %s", self.name, function_arguments)
try:
async for stream_msg in self.stream_function(**function_arguments):
yield stream_msg
except Exception as e:
logger.error(f"Error occurred while invoking function {self.name}: {e}")
yield FunctionResult(
function=self.metadata, value=None, metadata={"error": e, "arguments": function_arguments}
)
def gather_function_parameters(self, kernel: "Kernel", arguments: "KernelArguments") -> Dict[str, Any]:
"""Gathers the function parameters from the arguments."""
if self.prompt_template_config:
arguments = self.add_default_values(arguments, self.prompt_template_config)
function_arguments: Dict[str, Any] = {}
for param in self.parameters:
if param.name == "function":
function_arguments[param.name] = self.metadata
continue
if param.name == "kernel":
function_arguments[param.name] = kernel
continue
if param.name == "service":
function_arguments[param.name] = kernel.select_ai_service(self, arguments)[0]
continue
if param.name == "execution_settings":
function_arguments[param.name] = kernel.select_ai_service(self, arguments)[1]
continue
if param.name == "arguments":
function_arguments[param.name] = arguments
continue
if param.name == "prompt_template_config":
function_arguments[param.name] = self.prompt_template_config
continue
if param.type_ == "ChatHistory":
if param.name in arguments:
function_arguments[param.name] = arguments[param.name]
else:
function_arguments[param.name] = ChatHistory()
continue
if self.is_prompt:
# a semantic function will use the arguments (KernelArguments) instead of named arguments
continue
if param.name in arguments:
function_arguments[param.name] = arguments[param.name]
continue
if param.required:
raise ValueError(f"Parameter {param.name} is required but not provided in the arguments.")
logger.debug(f"Parameter {param.name} is not provided, using default value {param.default_value}")
return function_arguments
def add_default_values(
self, arguments: KernelArguments, prompt_template_config: PromptTemplateConfig
) -> KernelArguments:
"""Adds default values to the arguments."""
for parameter in prompt_template_config.input_variables:
if parameter.name not in arguments and parameter.default not in {None, "", False, 0}:
arguments[parameter.name] = parameter.default
return arguments
Reference in New Issue View Git Blame Copy Permalink