semantic-kernel/dotnet/notebooks/05-using-function-calling.ipynb

Introduction to the Function Calling

The most powerful feature of chat completion is the ability to call functions from the model. This allows you to create a chat bot that can interact with your existing code, making it possible to automate business processes, create code snippets, and more.

With Semantic Kernel, we simplify the process of using function calling by automatically describing your functions and their parameters to the model and then handling the back-and-forth communication between the model and your code.

Read more about it here.

In [ ]:

#r "nuget: Microsoft.SemanticKernel, 1.23.0"

#!import config/Settings.cs
#!import config/Utils.cs

using Microsoft.SemanticKernel;
using Microsoft.SemanticKernel.Connectors.OpenAI;
using Kernel = Microsoft.SemanticKernel.Kernel;

var builder = Kernel.CreateBuilder();

// Configure AI backend used by the kernel
var (useAzureOpenAI, model, azureEndpoint, apiKey, orgId) = Settings.LoadFromFile();

if (useAzureOpenAI)
    builder.AddAzureOpenAIChatCompletion(model, azureEndpoint, apiKey);
else
    builder.AddOpenAIChatCompletion(model, apiKey, orgId);

var kernel = builder.Build();

Setting Up Execution Settings

Using FunctionChoiceBehavior.Auto() will enable automatic function calling. There are also other options like Required or None which allow to control function calling behavior. More information about it can be found here.

In [10]:

#pragma warning disable SKEXP0001

OpenAIPromptExecutionSettings openAIPromptExecutionSettings = new() 
{
    FunctionChoiceBehavior = FunctionChoiceBehavior.Auto()
};

Providing plugins to the Kernel

Function calling needs an information about available plugins/functions. Here we'll import the SummarizePlugin and WriterPlugin we have defined on disk.

In [11]:

var pluginsDirectory = Path.Combine(System.IO.Directory.GetCurrentDirectory(), "..", "..", "prompt_template_samples");

kernel.ImportPluginFromPromptDirectory(Path.Combine(pluginsDirectory, "SummarizePlugin"));
kernel.ImportPluginFromPromptDirectory(Path.Combine(pluginsDirectory, "WriterPlugin"));

Define your ASK. What do you want the Kernel to do?

In [12]:

var ask = "Tomorrow is Valentine's day. I need to come up with a few date ideas. My significant other likes poems so write them in the form of a poem.";

Since we imported available plugins to Kernel and defined the ask, we can now invoke a prompt with all the provided information.

We can run function calling with Kernel, if we are interested in result only.

In [ ]:

var result = await kernel.InvokePromptAsync(ask, new(openAIPromptExecutionSettings));

Console.WriteLine(result);

But we can also run it with IChatCompletionService to have an access to ChatHistory object, which allows us to see which functions were called as part of a function calling process. Note that passing a Kernel as a parameter to GetChatMessageContentAsync method is required, since Kernel holds an information about available plugins.

In [ ]:

using Microsoft.SemanticKernel.ChatCompletion;

var chatCompletionService = kernel.GetRequiredService<IChatCompletionService>();

var chatHistory = new ChatHistory();

chatHistory.AddUserMessage(ask);

var chatCompletionResult = await chatCompletionService.GetChatMessageContentAsync(chatHistory, openAIPromptExecutionSettings, kernel);

Console.WriteLine($"Result: {chatCompletionResult}\n");
Console.WriteLine($"Chat history: {JsonSerializer.Serialize(chatHistory)}\n");