Model Management

OllamaApiClient exposes the full Ollama model-management API. All operations are available on the IOllamaApiClient interface and have convenient extension-method overloads for common cases.

Listing local models

var ollama = new OllamaApiClient("http://localhost:11434");

IEnumerable<Model> models = await ollama.ListLocalModelsAsync();

foreach (var model in models.OrderBy(m => m.Name))
    Console.WriteLine(model.Name);

Listing running models

Query which models are currently loaded in memory:

IEnumerable<RunningModel> running = await ollama.ListRunningModelsAsync();

foreach (var model in running)
    Console.WriteLine($"{model.Name} — expires {model.ExpiresAt}");

Pulling a model

Pull a model from the Ollama model hub and report download progress:

await foreach (var status in ollama.PullModelAsync("qwen3.5:35b-a3b"))
    Console.WriteLine($"{status.Percent:0}%  {status.Status}");

PullModelAsync streams PullModelResponse objects, each containing a Status string and a Percent value (0–100).

Pushing a model

Push a locally created model to a registry (requires a valid Ollama account):

await foreach (var status in ollama.PushModelAsync("myuser/my-custom-model:latest"))
    Console.WriteLine(status?.Status);

Copying a model

Create a local copy of a model under a new name:

await ollama.CopyModelAsync("qwen3.5:35b-a3b", "qwen3.5:35b-a3b-backup");

Deleting a model

await ollama.DeleteModelAsync("qwen3.5:35b-a3b-backup");

Showing model information

Retrieve detailed metadata for a locally available model, including its Modelfile, parameters and template:

ShowModelResponse info = await ollama.ShowModelAsync("qwen3.5:35b-a3b");

Console.WriteLine(info.ModelInfo);
Console.WriteLine(info.Parameters);

Creating a custom model

Build a new model from an existing one with a custom system prompt or other Modelfile instructions:

await foreach (var status in ollama.CreateModelAsync(new CreateModelRequest
{
    Model = "my-assistant",
    From = "qwen3.5:35b-a3b",
    System = "You are a helpful assistant that only speaks like a pirate.",
}))
{
    Console.WriteLine(status?.Status);
}

Generating embeddings

Although embeddings are primarily used in the context of RAG (retrieval-augmented generation) or semantic search, they are managed through the same client. See also the Chat and Generate page for usage alongside chat models.

var ollama = new OllamaApiClient("http://localhost:11434", "nomic-embed-text");

EmbedResponse response = await ollama.EmbedAsync("The quick brown fox jumps over the lazy dog");
float[] vector = response.Embeddings[0];

Console.WriteLine($"Embedding dimension: {vector.Length}");

Checking server availability

bool running = await ollama.IsRunningAsync();
Console.WriteLine(running ? "Ollama is running" : "Ollama is not available");

Getting the Ollama version

string version = await ollama.GetVersionAsync();
Console.WriteLine($"Ollama version: {version}");

Unloading a model from memory

To immediately free GPU/CPU memory occupied by a loaded model, use RequestModelUnloadAsync:

await ollama.RequestModelUnloadAsync("qwen3.5:35b-a3b");

This sends a request with KeepAlive = "0s", telling Ollama to unload the model right away.

Tip

You can control how long a model stays loaded with the KeepAlive property on any request. Set it to "5m" for five minutes, "0s" for immediate unload, or omit it to use the server default.

Working with blobs

Blobs (Binary Large Objects) are used when creating models from local files. First check whether a blob already exists on the server before uploading:

var digest = "sha256:29fdb92e57cf...";

if (!await ollama.IsBlobExistsAsync(digest))
{
    var bytes = await File.ReadAllBytesAsync("my-model-weights.bin");
    await ollama.PushBlobAsync(digest, bytes);
}

Then reference the digest in a CreateModelRequest:

await foreach (var status in ollama.CreateModelAsync(new CreateModelRequest
{
    Model = "my-local-model",
    Files = new Dictionary<string, string> { ["my-model-weights.bin"] = digest },
}))
{
    Console.WriteLine(status?.Status);
}