Skip to content

LLamaSharp.SemanticKernel

LLamaSharp.SemanticKernel are connections for SemanticKernel: an SDK for integrating various LLM interfaces into a single implementation. With this, you can add local LLaMa queries as another connection point with your existing connections.

For reference on how to implement it, view the following examples:

ITextCompletion

1
2
3
4
5
using var model = LLamaWeights.LoadFromFile(parameters);
// LLamaSharpTextCompletion can accept ILLamaExecutor. 
var ex = new StatelessExecutor(model, parameters);
var builder = new KernelBuilder();
builder.WithAIService<ITextCompletion>("local-llama", new LLamaSharpTextCompletion(ex), true);

IChatCompletion

1
2
3
4
5
using var model = LLamaWeights.LoadFromFile(parameters);
using var context = model.CreateContext(parameters);
// LLamaSharpChatCompletion requires InteractiveExecutor, as it's the best fit for the given command.
var ex = new InteractiveExecutor(context);
var chatGPT = new LLamaSharpChatCompletion(ex);

ITextEmbeddingGeneration

1
2
3
4
5
6
7
using var model = LLamaWeights.LoadFromFile(parameters);
var embedding = new LLamaEmbedder(model, parameters);
var kernelWithCustomDb = Kernel.Builder
    .WithLoggerFactory(ConsoleLogger.LoggerFactory)
    .WithAIService<ITextEmbeddingGeneration>("local-llama-embed", new LLamaSharpEmbeddingGeneration(embedding), true)
    .WithMemoryStorage(new VolatileMemoryStore())
    .Build();