microsoft/semantic-kernel

Fork 0

mirror of https://github.com/microsoft/semantic-kernel.git synced 2026-10-12 02:19:06 +00:00

Files

T

5817c2b927 Python: major features for the Kernel Arguments, Function Result and new Prompt Templating (#5077 )

### Motivation and Context

<!-- 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. -->
Major work to replace the context with Kernel Arguments on the front-end
and Function Results on the back.

Updated the function decorator to a new approach, in line with dotnet.

Revamps the way function are called, allowing native functions to be
completely ignorant of SK, other then the decorator.

This also moves things into the new folder structure in sync with
dotnet.

### Description

<!-- Describe your changes, the overall approach, the underlying design.
These notes will help understanding how your code works. Thanks! -->
Adds:
- KernelArguments, a dict-like class that replaces the variables in
KernelContext. Closes #4565
- FunctionResult, a class to hold the results of function executions,
includes the function, the value and metadata, as well as two convenient
function to get the value out of it (str and get_inner_content) the
first is generic, the second specifically for KernelContent responses
(from AI services).
- AI Service Selector class, has a default, which is based on the order
in the arguments followed by the order in the functions, can be
overridden to implement your own strategy. Closes #4631
- Introduces ChatHistory and refactors the PromptTemplateConfig. Closes
#4856, #4630
- Improves parsing of templates, will now all validate during creation
and throw an error then, instead of some that do not check for validaty
until used.
- Introduces named_args block and thereby the ability to have multiple
arguments for a function call in a template. Closes #5003

Updates:
- kernel_function decorators, the parameter decorator was removed and
instead we now use Annotated to add a description of a field and we get
the type and required from the function definition.
- core plugins, use the new approach to kernel_function decorators.
- planners, template engines have all been updated to use the kernel and
kernelarguments instead of Context.
- Events have been updated, now use kernelarguments and function_result
- Tokenizers support for named_args and improvements on parsing and
checking.
- Kernel examples and notebooks to use the latest code.
- All unit and integration tests. There is more code coverage now than
before.

Removed:
- kernelContext
- kernel_function_parameter_decorator
- delegate handling code for native functions
- file_io_plugin and tests
- SemanticFunctionConfig
- ChatPromptTemplate

### 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 😄

---------

Co-authored-by: Evan Mattson <[email protected]>
Co-authored-by: Evan Mattson <[email protected]>
Co-authored-by: Evan Mattson <[email protected]>

2024-02-24 20:05:40 +00:00

3.4 KiB

Raw Blame History

Basic Loading of the Kernel

To run the notebooks we recommend using Poetry and starting a shell with a virtual environment prepared to use SK.

See DEV_SETUP.md for more information.

In [ ]:

!python -m pip install semantic-kernel==0.5.1.dev0

In [1]:

import semantic_kernel as sk

When using the kernel for AI requests, the kernel needs some settings like URL and credentials to the AI models.

The SDK currently supports OpenAI and Azure OpenAI, among other connectors.

If you need an Azure OpenAI key, go here.

In [ ]:

from services import Service

# Select a service to use for this notebook (available services: OpenAI, AzureOpenAI, HuggingFace)
selectedService = Service.OpenAI

In [ ]:

kernel = sk.Kernel()

service_id = None
if selectedService == Service.OpenAI:
    from semantic_kernel.connectors.ai.open_ai import OpenAIChatCompletion

    api_key, org_id = sk.openai_settings_from_dot_env()
    service_id = "oai_chat_gpt"
    kernel.add_service(
        OpenAIChatCompletion(service_id=service_id, ai_model_id="gpt-3.5-turbo-1106", api_key=api_key, org_id=org_id),
    )
elif selectedService == Service.AzureOpenAI:
    from semantic_kernel.connectors.ai.open_ai import AzureChatCompletion

    deployment, api_key, endpoint = sk.azure_openai_settings_from_dot_env()
    service_id = "aoai_chat_completion"
    kernel.add_service(
        AzureChatCompletion(service_id=service_id, deployment_name=deployment, endpoint=endpoint, api_key=api_key),
    )

Great, now that you're familiar with setting up the Semantic Kernel, let's see how we can use it to run prompts.

3.4 KiB Raw Blame History

Basic Loading of the Kernel

3.4 KiB

Raw Blame History