[Question]: Clarification on redistributing onnxruntime.dll and DirectML.dll from Microsoft.Windows.AI.MachineLearning (license §3 Distributable Code) #27

What's your question?

Hello Windows ML team,

I develop a small, free, MIT-licensed Windows desktop app (unpackaged, distributed as a portable ZIP) that runs ONNX models with the ONNX Runtime and DirectML binaries from the Microsoft.Windows.AI.MachineLearning NuGet package (version 2.5.96, self-contained deployment).

While reviewing the package's license.txt, I found that §3(a)(i) describes the distributable code as files binplaced by the "WindowsAppSDK NuGet package". Since I reference Microsoft.Windows.AI.MachineLearning directly, rather than the main Microsoft.WindowsAppSDK package, I would like to confirm my understanding before releasing. Note that I build with CMake, using the package's own build/cmake integration, which places these DLLs next to my executable.

1. Are onnxruntime.dll and DirectML.dll from Microsoft.Windows.AI.MachineLearning 2.5.96 considered Distributable Code under this license, so that I may include them in my app's ZIP package?
2. My app only uses the ONNX Runtime C/C++ API with the DirectML execution provider. It does not use the Windows ML APIs (ExecutionProviderCatalog, etc.), so I do not ship Microsoft.Windows.AI.MachineLearning.dll. Is shipping only onnxruntime.dll and DirectML.dll a supported and permitted configuration?
3. For an unpackaged app distributed as a ZIP (no installer), is presenting an end-user license agreement on first launch, which the user must accept before the app loads these DLLs, an acceptable way to meet the §3(b)(ii) requirement?

Thank you for your help.

Related sample (if any)

No response

Context

No response