Koan.Ai.Provider.Ollama 0.5.2

Sylin.Koan.Ai.Provider.Ollama

Ollama AI provider for Koan: local LLM chat, stream, and embeddings via Ollama endpoint.

• Target framework: net9.0
• License: Apache-2.0

Install

dotnet add package Sylin.Koan.Ai.Provider.Ollama

Minimal setup

Register Koan + the Ollama provider (typical ASP.NET Program.cs):

// using Koan.AI; using Koan.Ai.Provider.Ollama; using Koan.AI.Web;
var builder = WebApplication.CreateBuilder(args);
builder.Services
		.AddKoan()
		.AddAi()
		.AddOllama(o =>
		{
				o.BaseAddress = new Uri("http://localhost:11434"); // default
				o.DefaultModel = "qwen3:4b"; // or any local model/tag
		})
		.AddAiWeb(); // optional HTTP endpoints under /ai
var app = builder.Build();
app.MapControllers();
app.Run();

Then query the default engine:

using Koan.AI;
var res = await Engine.Prompt("Explain quantum entanglement briefly.");
Console.WriteLine(res.Text);

Or via HTTP (if AddAiWeb is enabled):

POST /ai/chat { "model": "qwen3:4b", "messages": [{ "role": "user", "content": "Explain quantum entanglement briefly." }] }

Using reasoning (think)

Set the optional think flag in prompt options. The Ollama adapter emits a top-level think in its JSON payload for models that support it (e.g., Qwen3/R1/DeepSeek‑V3.1 quants).

using Koan.AI;
using Koan.AI.Contracts.Models;

var res = await Engine.Prompt(
		"Briefly explain quantum entanglement.",
		model: "qwen3:4b",
		opts: new AiPromptOptions { Think = true }
);
Console.WriteLine(res.Text);

HTTP example:

POST /ai/chat { "model": "qwen3:4b", "messages": [{ "role": "user", "content": "Explain quantum entanglement briefly." }], "options": { "think": true } }

Vendor-specific options (passthrough)

Unknown fields posted under options are forwarded to Ollama’s options bag in the request. Use this for provider-specific parameters like mirostat, repeat_penalty, etc. Known Koan fields still map to Ollama (temperature, top_p, num_predict, stop). If keys overlap, your posted value wins.

HTTP example:

POST /ai/chat { "model": "qwen3:4b", "messages": [{ "role": "user", "content": "Summarize Bell’s theorem in one sentence." }], "options": { "temperature": 0.6, "topP": 0.95, "maxOutputTokens": 128, "mirostat": 2, "mirostat_tau": 5.0, "repeat_penalty": 1.1 } }

Programmatic example:

using Koan.AI;
using Koan.AI.Contracts.Models;

var opts = new AiPromptOptions
{
		Temperature = 0.6,
		TopP = 0.95,
		MaxOutputTokens = 128,
		// Add arbitrary vendor params via object initializer syntax in JSON calls
};
var res = await Engine.Prompt("Summarize Bell’s theorem in one sentence.", "qwen3:4b", opts);

Notes

• The Web API uses messages[] instead of a top-level prompt.
• Embeddings are supported via POST /ai/embeddings with model and input (array of strings).
• Streaming is supported via POST /ai/chat/stream with the same body as /ai/chat.

Links

• Ollama: https://ollama.com