Adding memory to Semantic Kernel

Semantic Kernel uses text embeddings to simulate memory. When an AI agent wants to see what it knows, it searches a database of text embeddings and pulls back the results with the largest similarity score.

Memory is available via a separate NuGet package. Because this capability is still early in development, we’ll need to explicitly reference its version and we’ll need to disable a series of SKEXP flags, as shown here:

#r "nuget: Microsoft.SemanticKernel.Plugins.Memory, 1.16.0-alpha"
using Microsoft.SemanticKernel.Memory;
#pragma warning disable SKEXP0001, SKEXP0010, SKEXP0050

These #pragma statements effectively acknowledge that we understand semantic memory is considered a preview feature, is subject to change, and should not yet be used for real-world production scenarios.

Reminder about preview features

As I mentioned earlier in this chapter, Semantic Kernel is still rapidly evolving. Expect that portions of...