Why should you care?

This is the key wiring lesson: SK + Ollama = private agents with a familiar .NET API.

See it live — copy this example

Run CLI steps in a terminal. Run C# samples in a .NET 8 console or Web API project with Ollama running on localhost:11434.

// Example pattern (package versions vary — check current SK Ollama connector)
using Microsoft.SemanticKernel;

var builder = Kernel.CreateBuilder();
builder.AddOllamaChatCompletion(
    modelId: "phi3:mini",
    endpoint: new Uri("http://localhost:11434"));
var kernel = builder.Build();

var result = await kernel.InvokePromptAsync("Reply with one word: pong");
Console.WriteLine(result);

What happened?

• AddOllamaChatCompletion points SK at localhost Ollama.
• InvokePromptAsync sends a simple prompt and returns the model text.

Practice next

1. Add the correct SK + Ollama connector package for your SK version.
2. Start Ollama and ensure the model exists.
3. Run InvokePromptAsync.
4. Use mistral instead of phi3.
5. Wrap invoke in try/catch and log errors.

Remember

SK can target Ollama locally. Model id must match pulled tags. Keep endpoint in config.

Private prompt path

HR FAQ agent never leaves the building network.

Outcome: Kernel → Ollama on localhost.