Understanding the kernel in Semantic Kernel

In this article

Similar to operating system, the kernel is responsible for managing resources that are necessary to run "code" in an AI application. This includes managing the AI models, services, and plugins that are necessary for both native code and AI services to run together.

If you want to see the code demonstrated in this article in a complete solution, check out the following samples in the public documentation repository.

Because the kernel has all of the services and plugins necessary to run both native code and AI services, it is used by nearly every component within the Semantic Kernel SDK. This means that if run any prompt or code in Semantic Kernel, it will always go through a kernel.

This is extremely powerful, because it means you as a developer have a single place where you can configure, and most importantly monitor, your AI application. Take for example, when you invoke a prompt from the kernel. When you do so, the kernel will...

1. Select the best AI service to run the prompt.
2. Build the prompt using the provided prompt template.
3. Send the prompt to the AI service.
4. Receive and parse the response.
5. Before finally returning the response to your application.

Throughout this entire process, you can create events and middleware that are triggered at each of these steps. This means you can perform actions like logging, provide status updates to users, and most importantly responsible AI. All from a single place.

Before building a kernel, you should first understand the two types of components that exist within a kernel: services and plugins. Services consist of both AI services and other services that are necessary to run your application (e.g., logging, telemetry, etc.). Plugins, meanwhile, are any code you want AI to call or leverage within a prompt.

In the following examples, you can see how to add a logger, chat completion service, and plugin to the kernel.

using Microsoft.Extensions.DependencyInjection;
using Microsoft.Extensions.Logging;
using Microsoft.SemanticKernel;
using Plugins;

var builder = Kernel.CreateBuilder(); ;
builder.Services.AddLogging(c => c.AddDebug().SetMinimumLevel(LogLevel.Trace));
builder.Services.AddAzureOpenAIChatCompletion(
    AzureOpenAIDeploymentName,   // The name of your deployment (e.g., "gpt-35-turbo")
    AzureOpenAIEndpoint,         // The endpoint of your Azure OpenAI service
    AzureOpenAIApiKey,           // The API key of your Azure OpenAI service
    modelId: AzureOpenAIModelId  // The model ID of your Azure OpenAI service
);
builder.Plugins.AddFromType<TimePlugin>();
builder.Plugins.AddFromPromptDirectory("./plugins/WriterPlugin");

var kernel = builder.Build();

var builder = Kernel.CreateBuilder();
builder.Services.AddLogging(c => c.AddDebug().SetMinimumLevel(LogLevel.Trace));
builder.Services.AddOpenAIChatCompletion(
    OpenAIModelId,       // The model ID of your OpenAI service
    OpenAIApiKey,        // The API key of your OpenAI service
    OpenAIOrgId          // The organization ID of your OpenAI service
);
builder.Plugins.AddFromType<TimePlugin>();
builder.Plugins.AddFromPromptDirectory("./plugins/WriterPlugin");

var kernel = builder.Build();

Semantic Kernel makes it easy to run prompts alongside native code because they are both expressed as KernelFunctions. This means you can invoke them in exactly same way.

To run KernelFunctions, Semantic Kernel provides the InvokeAsync method. Simply pass in the function you want to run, its arguments, and the kernel will handle the rest.

var currentTime = await kernel.InvokeAsync("TimePlugin", "GetCurrentUtcTime");
Console.WriteLine(currentTime);

var poemResult = await kernel.InvokeAsync("WriterPlugin", "ShortPoem", new() {
    { "input", currentTime }
});
Console.WriteLine(poemResult);

There once was a sun in the sky
That shone so bright, it caught the eye
But on December tenth
It decided to vent
And took a break, said "Bye bye!"

For more details on how to configure and leverage these properties, please refer to the following articles:

Once you're done configuring the kernel, you can learn how to create prompts to run AI services from the kernel.