Ollama Connector

Supports

• text generation
• chat completion
• embedding generation

:warning: Embedding generation: Is experimental in the semantic kernel.

Quick Start

• Install from nuget

  dotnet add package Codeblaze.SemanticKernel.Connectors.Ollama

• Text Generation

  Configure the kernel

  var builder = Kernel.CreateBuilder(); // provide the HTTP client used to interact with Ollama API builder.Services.AddTransient<HttpClient>(); builder.AddOllamaTextGeneration( config["Ollama:Model"], // Ollama model Id config["Ollama:BaseUrlGeneration"] // Ollama endpoint ); var kernel = builder.Build();

  Usage

  const string prompt = """ Bot: How can I help you? User: {{$input}} --------------------------------------------- The intent of the user in 5 words or less: """; var result = await kernel.InvokePromptAsync(prompt, new KernelArguments { {"input", input} }); System.Console.WriteLine(result.GetValue<string>());

• Chat Completion

  Configure the kernel

  var builder = new KernelBuilder(); // provide the HTTP client used to interact with Ollama API builder.Services.AddTransient<HttpClient>(); builder.AddOllamaChatCompletion( config["Ollama:Model"], // Ollama model Id config["Ollama:BaseUrlGeneration"] // Ollama endpoint ); var kernel = builder.Build();

  Usage

  var chat = _Kernel.GetRequiredService<IChatCompletionService>(); var history = new ChatHistory(); // add messages to current chat history as required history.AddSystemMessage("..."); history.AddAssistantMessage("..."); history.AddUserMessage(input); // result is a list of all chat messages // including the output of current prompt var result = await chat.GetChatMessageContentsAsync(history); // Print the last message System.Console.WriteLine(result[^1].Content);