Workshop Semantic Kernel

Exercise 1 - Chat

We will first start with a basic chat, and then evolve it to a more complex chat.

A. Single message

• Print hardcoded answer to console
• Read the user input from console instead of hard coding it

Hint

Configure Kernel

var kernelBuilder = Kernel.CreateBuilder();
var config = new { modelID = "gpt-4", azureEndpoint = "", apiKey = "….." };
kernelBuilder.AddAzureOpenAIChatCompletion(config.modelID, config.azureEndpoint,config.apiKey);
var kernel = kernelBuilder.Build();

Hint 2

Invoke Service

var chatService = kernel.GetRequiredService<IChatCompletionService>();
var chatResponse = await chatService.GetChatMessageContentAsync("Hi! How are you?");
Console.WriteLine(chatResponse);

B. Full back and forth chat

• Now that we can chat, let's allow for us to keep asking until we kill our app!

C. Remember me!

• Add a history to the chat so the model remembers what we talked about!

Hint

Keep array of messages to pass to call

D. Bonus - streaming chat

If you would like to get that delayed output feel, you can use the GetStreamChatMessageContentAsync Method:

E. Bonus - Plugin

• Create a plugin get Chuck Norris jokes, pokemon, or any other API you like.

F. Bonus - RAG

• Initialize your light plugin with some data

Hint

URL Hint