Proper way to use OpenBLAS and AI model with this template

[SuperKogito]

First of all, thank you for sharing this great project ❤️ It is imo a life saver for anyone getting started with Audio plugins.
I already tested it on Windows 10 (mingw) and Ubuntu 22.04 and it works on both.

I am currently trying to develop an AI based plugin effect.
My Plugin uses OpenBLAS and although the resulting VST3 works on my machine, but when I try it on other machines it is not scanned/ recognised by the DAW.

I suspect this to be related to multi-threading issues when using OpenBLAS, or my architecture. I also noticed that my inference/ run is invoked even though my plugin is paused/ stopped. Hence I would like to ask if there is a way to handle this; something
something like void deactivate() or some way to bridge the run logic in this case. Does this make any sense?

My second question; what would be a suitable architecture for loading an AI model, invoking it and then cleaning it in the Audio plugin? atm I use this approach:

#include "DistrhoPlugin.hpp"

START_NAMESPACE_DISTRHO

// --------------------------------------------------------------------------------------------------------------------

class ImGuiPluginDSP : public Plugin
{

    // inits ...
    Model* model;

public:
   /**
      Plugin class constructor.@n
      You must set all parameter values to their defaults, matching ParameterRanges::def.
    */
    ImGuiPluginDSP()
        : Plugin(kParamCount, 0, 0), // parameters, programs, states
          model(nullptr)
    {
        // load model weights ...
       load_weights();
    }

    ~ImGuiPluginDSP() override
    {
        // free model
         free_model();
        // free weights
        free_weights();
    }

protected:
    // ----------------------------------------------------------------------------------------------------------------
    // Information

...

    // ----------------------------------------------------------------------------------------------------------------
    // Internal data



    // ----------------------------------------------------------------------------------------------------------------
    // Audio/MIDI Processing

   /**
      Activate this plugin.
    */
    void activate() override
    {
        // allocate and init model
	model_init();
    }

   /**
      Run/process function for plugins without MIDI input.
      @note Some parameters might be null if there are no audio inputs or outputs.
    */
    void run(const float** inputs, float** outputs, uint32_t frames) override
    {
        // run model inference/ routine
        model_process(model, inputs[0], inputs[1], frames, outputs);
        
    }

    // ----------------------------------------------------------------------------------------------------------------
    // Callbacks (optional)



    // ----------------------------------------------------------------------------------------------------------------

    DISTRHO_DECLARE_NON_COPYABLE_WITH_LEAK_DETECTOR(ImGuiPluginDSP)
};

// --------------------------------------------------------------------------------------------------------------------

Plugin* createPlugin()
{
    return new ImGuiPluginDSP();
}

// --------------------------------------------------------------------------------------------------------------------

END_NAMESPACE_DISTRHO

I should also mention that my model code is in C (not sure if this is relevent).
I am using windows 10 with mingw 13.2.0
I would appreciate any response. Thank you :)