Make DVC projects robust to migration


I work for a company as a data scientist and am beginning to integrate DVC into ML projects. Over the summer, we plan to transition from our on-premises setup to Azure Cloud for remote storage and moving from Azure DevOps platform to GitHub for Git projects.

I have several DVC cross-projects which import data from each other (for example, a dataset DVC project that serves multiple ML projects). The main issue is that a DVC import freezes the URL of its origin. If I migrate, I’ll need to modify the history of my Git projects by replacing URLs, which is both painful and messy.

I came across this discussion where the idea of an HTTP proxy, like DAGsHub, is suggested, but it’s not directly accessible through my company.

Do you have any ideas for migration or best practices to make a DVC project robust to migration?

PS: The same applies to the DVC remote.

Thanks in advance.