Whether you’re building AI solutions or enhancing existing projects with advanced search capabilities, you now have the option of using Chroma as a database provider in your .NET applications.

What is Chroma?

Chroma is an open-source database for your AI applications.

With support for storing embeddings, metadata filtering, vector search, full-text search, document storage, and multi-modal retrieval, you can use Chroma to power semantic search and Retrieval Augmented Generation (RAG) features in your app.

For more details, check out the Chroma website.

Get started with Chroma in your C# application

In this scenario, we’ll be using the ChromaDB.Client package to connect to a Chroma database and search for movies using vector search.

The easiest way to start is locally using the Chroma Docker image. You can also deploy an instance in Azure.

Note: the ChromaDB.Client is an open-source community-supported library.

Connect to the database

1. Create a C# console application.
2. Install the ChromaDB.Client NuGet package.
3. Create a ChromaClient with configuration options.

   using ChromaDB.Client;
   
   var configOptions = new ChromaConfigurationOptions(uri: "http://localhost:8000/api/v1/");
   using var httpClient = new HttpClient();
   var client = new ChromaClient(configOptions, httpClient);

When using a hosted version of Chroma, replace the uri with your hosted endpoint.

Create a collection

Now that you have a client, create a collection to store movie data.

var collection = await client.GetOrCreateCollection("movies");

To perform operations on that collection, you’ll then need to create a collection client.

var collectionClient = new ChromaCollectionClient(collection, configOptions, httpClient);

Add data to your collection

Once your collection is created, it’s time to add data to it. The data we’re storing will consist of:

• Movie IDs
• Embeddings to represent the movie description.
• Metadata containing the movie title

List<string> movieIds = ["1", "2", "3", "4", "5" ];

List<ReadOnlyMemory<float>> descriptionEmbeddings = [

    new [] { 0.10022575f, -0.23998135f },
    new [] { 0.10327095f, 0.2563685f },
    new [] { 0.095857024f, -0.201278f },
    new [] { 0.106827796f, 0.21676421f },
    new [] { 0.09568083f, -0.21177962f },
];

List<Dictionary<string,object>> metadata = 
[
    new Dictionary<string, object> { ["Title"] = "The Lion King" },
    new Dictionary<string, object> { ["Title"] = "Inception"  },
    new Dictionary<string, object> { ["Title"] = "Toy Story"  },
    new Dictionary<string, object> { ["Title"] = "Pulp Fiction" },
    new Dictionary<string, object> { ["Title"] = "Shrek"  },
];

await collectionClient.Add(movieIds, descriptionEmbeddings, metadata);

Search for movies (using vector search)

Now that your data is in the database, you can query it. In this case, we’re using vector search.

List<ReadOnlyMemory<float>> queryEmbedding = [new([0.12217915f, -0.034832448f])];

var queryResult = await collectionClient.Query(
    queryEmbeddings: queryEmbedding,
    nResults: 2,
    include: ChromaQueryInclude.Metadatas | ChromaQueryInclude.Distances);

foreach (var result in queryResult)
{
    foreach (var item in result)
    {
        Console.WriteLine($"Title: {(string)item.Metadata["Title"] ?? string.Empty} {(item.Distance)}");
    }
}

The result should look similar to the following output.

Title: Toy Story 0.028396977
Title: Shrek 0.032012463

Watch it live

Join Jiří Činčura on the .NET Data Community Standup on February 26 to learn more about how to use Chroma and the new C# SDK.

Conclusion

This latest addition enhances the growing AI ecosystem in .NET. It paves the way for a simpler implementation of the existing Semantic Kernel connector and seamless integration into your .NET apps using foundational components like Microsoft.Extensions.VectorData and Microsoft.Extensions.AI.

We’d like to thank @ssone95 for his work and contributions to the project. We’re excited to continue building partnerships and working with the community to enable .NET developers to build AI applications.

To learn how you can start building AI apps using databases like Chroma, check out the .NET AI documentation.

Try out the Chroma C# SDK today and provide feedback.