ollama trying to pull custom model

[shirofaii]

Describe the bug

When you add custom model for ollama (with id "custom" for example) provider trying to pull this model and failed

Steps to reproduce the bug

Steps to reproduce:

1. Create custom model with id="custom" for ollama
2. Run clean ollama server without any model loaded
3. Compile and run this code

using LangChain.Providers.Ollama;
using Ollama;

var provider = new OllamaProvider("http://127.0.0.1:11434/api");
var llm = new OllamaChatModel(provider, "custom");

var result = await llm.GenerateAsync("Hi!");

Actual result: provider trying to pull model "custom" and fail

Expected behavior

ollama load existed model "custom" and generate valid response

Screenshots

No response

NuGet package version

No response

Additional context

No response

[HavenDV]

Now it can be disabled in provider using CanPullModelsAutomatically property

[fyziktom]

I think it will help if you will use ListModelsAsync instad of ListRunningModelsAsync in OllamaProvider.cs. Like this:

public async Task PullModelIfRequiredAndAllowedAsync(
    string id,
    CancellationToken cancellationToken = default)
{
    if (!CanPullModelsAutomatically)
    {
        return;
    }

    try
    {
        // Pull the model if it is not running
        var runningModels = await Api.Models.ListModelsAsync(cancellationToken).ConfigureAwait(false);
        if (runningModels.Models != null &&
            runningModels.Models.All(x => x.Model1?.Contains(id) != true))
        {
            await Api.Models.PullModelAsync(id, cancellationToken: cancellationToken)
                .EnsureSuccessAsync().ConfigureAwait(false);
        }
    }
    catch (HttpRequestException)
    {
        // Ignore
    }
}

I discovered that previous version worked only when the model was really running. So it was possible to use custom models, but only when you will "wake them up" to running state. With change to ListModelsAsync I can see it list all the available models.