Encamina.Enmarcha.SemanticKernel.Plugins.Memory 8.1.6

Semantic Kernel - Memory Plugin

Memory Plugin is a project that offers functionality as a plugin for searching the memory of the Semantic Kernel. When text is stored in memory within the Semantic Kernel, it is saved in the form of embeddings that can be queried through the Memory Plugin.

Setup

Nuget package

First, install NuGet. Then, install Encamina.Enmarcha.SemanticKernel.Plugins.Memory from the package manager console:

PM> Install-Package Encamina.Enmarcha.SemanticKernel.Plugins.Memory

.NET CLI:

First, install .NET CLI. Then, install Encamina.Enmarcha.SemanticKernel.Plugins.Memory from the .NET CLI:

dotnet add package Encamina.Enmarcha.SemanticKernel.Plugins.Memory

How to use

To use MemoryQueryPlugin, the usual approach is to import it as a plugin within Semantic Kernel. The simplest way to do this is by using the extension method ImportMemoryPlugin, which handles the import of the Plugin into Semantic Kernel.

// Entry point
var builder = WebApplication.CreateBuilder(new WebApplicationOptions
{
   // ...
});

// ... 

builder.Services.AddScoped(sp =>
{
      var kernel = new KernelBuilder()
        .WithAzureTextEmbeddingGenerationService("<YOUR DEPLOYMENT NAME>", "<YOUR AZURE ENDPOINT>", "<YOUR API KEY>")
        //.WithOpenAITextEmbeddingGenerationService("<YOUR MODEL ID>", "<YOUR API KEY>")
        /// ...
        .Build();

    // ...

    kernel.ImportMemoryPlugin(ILengthFunctions.LengthByTokenCount);

    return kernel;
});

Now you can inject the kernel via constructor, and the memory capabilities are already available.

public class MyClass
{
    private readonly Kernel kernel;

    public MyClass(Kernel kernel)
    {
        this.kernel = kernel;
    }

    public async Task TestMemoryQueryAsync()
    {
        var contextVariables = new ContextVariables();
        contextVariables.Set(PluginsInfo.MemoryQueryPlugin.Functions.QueryMemory.Parameters.Query, "What is the weather like in Madrid?");
        contextVariables.Set(PluginsInfo.MemoryQueryPlugin.Functions.QueryMemory.Parameters.CollectionsStr, "collection-1:collection-2");
        contextVariables.Set(PluginsInfo.MemoryQueryPlugin.Functions.QueryMemory.Parameters.CollectionSeparator, ":");
        contextVariables.Set(PluginsInfo.MemoryQueryPlugin.Functions.QueryMemory.Parameters.MinRelevance, "0.8");
        contextVariables.Set(PluginsInfo.MemoryQueryPlugin.Functions.QueryMemory.Parameters.ResponseTokenLimit, "1500");
        contextVariables.Set(PluginsInfo.MemoryQueryPlugin.Functions.QueryMemory.Parameters.ResultsLimit, "2");

        var functionMemoryQuery = kernel.Func(PluginsInfo.MemoryQueryPlugin.Name, PluginsInfo.MemoryQueryPlugin.Functions.QueryMemory.Name);

        var resultContext = await kernel.RunAsync(contextVariables, functionMemoryQuery);
    }
}

Within resultContext, you will find texts that meet the conditions.