Kernel Memory and .NET Aspire

[NikiforovAll]

https://nikiforovall.github.io/dotnet/ai/2024/09/04/typical-rag-dotnet.html

https://x.com/nikiforovall/status/1831242148533973289

var builder = WebApplication.CreateBuilder(args);

builder.AddServiceDefaults();
builder.AddNpgsqlDataSource("rag-db"); // Adds OpenTelemetry and health checks for Postgres

builder.ConfigureAiOptions();
var aiOptions = builder.GetAiOptions();

builder.Services.AddKernelMemory<MemoryServerless>(memoryBuilder =>
{
    memoryBuilder
        .WithPostgresMemoryDb(
            new PostgresConfig()
            {
                ConnectionString = builder.Configuration.GetConnectionString("rag-db")!
            }
        )
        .WithSemanticKernelTextGenerationService(
            new AzureOpenAIChatCompletionService(
                deploymentName: aiOptions.Deployment,
                endpoint: aiOptions.Endpoint,
                apiKey: aiOptions.ApiKey
            ),
            new SemanticKernelConfig()
        )
        .WithSemanticKernelTextEmbeddingGenerationService(
            new AzureOpenAITextEmbeddingGenerationService(
                deploymentName: "text-embedding-ada-002",
                endpoint: aiOptions.Endpoint,
                apiKey: aiOptions.ApiKey
            ),
            new SemanticKernelConfig()
        );
});

var app = builder.Build();

app.MapDefaultEndpoints();
app.AddKernelMemoryEndpoints(apiPrefix: "/rag");
app.Run();

[dluc]

nice :-). One can also use WithAzureOpenAITextGeneration and WithAzureOpenAITextGeneration and have support for Managed Identities for more security

[NikiforovAll]

Great, it would be great to managed identities support for WithSemanticKernel* methods, I primarily used it because of additional Otel traces produced by SK