Task 01 - Use the built-in embedding service of Semantic Kernel to perform vector embedding (20 minutes)

Update the Program.cs file to add the built-in TextEmbeddingGeneration service:

1. Add AddAzureOpenAITextEmbeddingGeneration(embeddingDeploymentName, endpoint, key) to the Kernel singleton builder call added in Exercise 2 Task 2 (line 224 of the 0202.MD file).

    kernelBuilder.AddAzureOpenAITextEmbeddingGeneration(
        deploymentName: builder.Configuration["AzureOpenAI:EmbeddingDeploymentName"]!,
        endpoint: builder.Configuration["AzureOpenAI:Endpoint"]!,
        apiKey: builder.Configuration["AzureOpenAI:ApiKey"]!
    );
   

   The code will be underlined by a red squiggly line, indicating an issue.

   The AddAzureOpenAITextEmbeddingGeneration feature of Semantic Kernel is flagged for evaluation purposes only at this time, so you must suppress that issue to be able to use it in the project.

2. To suppress the issue, hover your mouse cursor over the AddAzureOpenAITextEmbeddingGeneration function call and select the dropdown in the tooltip icon that appears.

   1. In the flyout menu, select Suppress of configure issues -> Suppress SKEXP0010 and then select in Source.

      

   2. This will wrap the code in #pragma warning disable and #pragma warning restore directives, and the final code will look like this:

      #pragma warning disable SKEXP0010 // Type is for evaluation purposes only and is subject to change or removal in future updates. Suppress this diagnostic to proceed.
          kernelBuilder.AddAzureOpenAITextEmbeddingGeneration(
              deploymentName: builder.Configuration["AzureOpenAI:EmbeddingDeploymentName"]!,
              endpoint: builder.Configuration["AzureOpenAI:Endpoint"]!,
              apiKey: builder.Configuration["AzureOpenAI:ApiKey"]!
          );
      #pragma warning restore SKEXP0010 // Type is for evaluation purposes only and is subject to change or removal in future updates. Suppress this diagnostic to proceed.
   

To remove the AzureOpenAIClient singleton service from Program.cs, locate the line that starts with builder.Services.AddSingleton<AzureOpenAIClient> and delete that line and the code block it encapsulates.

To update the VectorizationService primary constructor to remove the AzureOpenAIClient and inject the Kernel service, perform the following steps:

1. Remove the Azure.AI.OpenAI using statement.

2. Add the following Semantic Kernel references:

    using Microsoft.SemanticKernel;
    using Microsoft.SemanticKernel.Embeddings;
   

3. Change the primary constructor for the VectorizationService to swap out the use of AzureOpenAIClient for Semantic Kernel, deleting the injected AzureOpenAIClient and replacing it with Kernel. The new primary constructor for the class will look like:

    public class VectorizationService(Kernel kernel, CosmosClient cosmosClient, IConfiguration configuration) : IVectorizationService
   

4. Delete the _client class variable and create a new one named _kernel with a type of Kernel.

    private readonly Kernel _kernel = kernel;
   

The steps to update the VectorizationService to use the built-in vector embedding capabilities of Semantic Kernel are:

1. In the GetEmbeddings function, remove the embeddingClient variable, as it is no longer necessary.

2. Replace the call to the embedding client for the built-in GenerateEmbeddingAsync() method of Semantic Kernel.

    // Generate a vector for the provided text.
    var embeddings = await _kernel.GetRequiredService<ITextEmbeddingGenerationService>().GenerateEmbeddingAsync(text);
   

3. As you did in the Program.cs class, you must suppress the issue with the ITextEmbeddingGenerationService feature using # pragma warning directives to use it in this class. The final code will look similar to the following:

    #pragma warning disable SKEXP0001 // Type is for evaluation purposes only and is subject to change or removal in future updates. Suppress this diagnostic to proceed.
    // Generate a vector for the provided text.
    var embeddings = await _kernel.GetRequiredService<ITextEmbeddingGenerationService>().GenerateEmbeddingAsync(text);
    #pragma warning restore SKEXP0001
   

4. Update the vector variable to handle the different type returned by the GenerateEmbeddingsAsync method.

    var vector = embeddings.ToArray();
   

5. Save the VectorizationService file.

1. To test the updated VectorizationService, start the web API and run the Steamlit dashboard.
   1. In Visual Studio Code, open a new terminal window and change the directory to scr\ContotoSuitesWebAPI.

   2. At the terminal prompt, enter the following command to run the API locally:

        dotnet run
      

2. Once the API has started, as indicated by output in the terminal stating Now listening on: http://localhost:5292, open a new terminal window in Visual Studio Code, navigate to the src\ContosoSuitesDashboard folder, and run the following command to start the Streamlit dashboard:

    python -m streamlit run Index.py
   

3. Navigate to the Vector Search page in the Streamlit dashboard, and then submit the following query for maintenance requests:
   1. Enter a search query of “Show me requests relating to elevator noises” with max results of ten and a minimum similarity score of 0.78.
   2. You should see several results about elevators and noise, indicating the updates to the API and VectorizationService are working correctly.