MLflow tracking server behind authentication layer

[Daniel Quang]

I currently have an MLflow tracking server. The UI can be accessed on a browser, (eg https://www.mymlflowserver.com). However, the UI is behind an authentication layer, which may or may not have 2FA (eg Okta). Users must sign in before they can view the UI. If I wanted to do model tracking, I can do post requests to the server in Python. All it requires that I do is make sure the headers argument in the requests.post function has the proper info. I'd like to just do the model tracking purely from the Python API, such as with the MlflowClient class. Do you have any advice on how to do this? My best guess on an approach right now is to modify the MLflowClient and/or TrackingServiceClient classes and pass in the proper authentication credentials.