Generating embeddings for Semantic Kernel Vector Store connectors

These services can be constructed directly or added to a dependency injection container and resolved from there.

Constructing an embedding generator

You can construct instances of the text embedding services provided by Semantic Kernel directly. They all implement the ITextEmbeddingGenerationService interface.

// Constructing an Azure Open AI embedding generation service directly.
ITextEmbeddingGenerationService azureOpenAITES = new AzureOpenAITextEmbeddingGenerationService(
    "text-embedding-ada-002",
    "https://{myservice}.openai.azure.com/",
    "apikey");

// Constructing an Olama embedding generation service directly.
ITextEmbeddingGenerationService olamaTES = new OllamaTextEmbeddingGenerationService(
    "mxbai-embed-large",
    new Uri("http://localhost:11434"));

You can also use helpers to register them with a dependency injection container.

// Registering Google AI embedding generation service with a service collection.
var services = new ServiceCollection();
services.AddGoogleAIEmbeddingGeneration("text-embedding-004", "apiKey");

// Registering Mistral AI embedding generation service with the dependency injection container on
// the kernel builder.
var kernelBuilder = Kernel
    .CreateBuilder()
    .AddMistralTextEmbeddingGeneration("mistral-embed", "apiKey");

Generating embeddings

To use the ITextEmbeddingGenerationService you created, just call the GenerateEmbeddingAsync method on it.

Here is an example of generating embeddings when uploading records.

public async Task GenerateEmbeddingsAndUpsertAsync(
    ITextEmbeddingGenerationService textEmbeddingGenerationService,
    IVectorStoreRecordCollection<ulong, Hotel> collection)
{
    // Upsert a record.
    string descriptionText = "A place where everyone can be happy.";
    ulong hotelId = 1;

    // Generate the embedding.
    ReadOnlyMemory<float> embedding =
        await textEmbeddingGenerationService.GenerateEmbeddingAsync(descriptionText);

    // Create a record and upsert with the already generated embedding.
    await collection.UpsertAsync(new Hotel
    {
        HotelId = hotelId,
        HotelName = "Hotel Happy",
        Description = descriptionText,
        DescriptionEmbedding = embedding,
        Tags = new[] { "luxury", "pool" }
    });
}

Here is an example of generating embeddings when searching.

public async Task GenerateEmbeddingsAndSearchAsync(
    ITextEmbeddingGenerationService textEmbeddingGenerationService,
    IVectorStoreRecordCollection<ulong, Hotel> collection)
{
    // Upsert a record.
    string descriptionText = "Find me a hotel with happiness in mind.";

    // Generate the embedding.
    ReadOnlyMemory<float> searchEmbedding =
        await textEmbeddingGenerationService.GenerateEmbeddingAsync(descriptionText);

    // Search using the already generated embedding.
    List<VectorSearchResult<Hotel>> searchResult = await collection.VectorizedSearchAsync(searchEmbedding).ToListAsync();

    // Print the first search result.
    Console.WriteLine("Score for first result: " + searchResult.FirstOrDefault()?.Score);
    Console.WriteLine("Hotel description for first result: " + searchResult.FirstOrDefault()?.Record.Description);
}

Embedding dimensions

Vector databases typically require you to specify the number of dimensions that each vector has when creating the collection. Different embedding models typically support generating vectors with different dimension sizes. E.g. Open AI text-embedding-ada-002 generates vectors with 1536 dimensions. Some models also allow a developer to choose the number of dimensions they want in the output vector, e.g. Google text-embedding-004 produces vectors with 768 dimension by default, but allows a developer to choose any number of dimensions between 1 and 768.

It is important to ensure that the vectors generated by the embedding model have the same number of dimensions as the matching vector in the database.

If creating a collection using the Semantic Kernel Vector Store abstractions, you need to specify the number of dimensions required for each vector property either via annotations or via the record definition. Here are examples of both setting the number of dimensions to 1536.

[VectorStoreRecordVector(Dimensions: 1536)]
public ReadOnlyMemory<float>? DescriptionEmbedding { get; set; }

new VectorStoreRecordVectorProperty("DescriptionEmbedding", typeof(float)) { Dimensions = 1536 }

Coming soon

More info coming soon.

Coming soon

More info coming soon.