I’m not familiar with ‘atomes’ within the context of Visual Paradigm. All Visual Paradigm does is follow existing standards for its supported modeling languages. So that is basically the thing to turn to if you wonder about requirements and specifications while working on a diagram type.
But about VP itself: nothing is really necessary, it all depends on context. If you need 2 diagrams (for example) you can just start a new project, create the diagrams and leave it at that. Comparable to what I explained in this post.
If things become more complex then you can chose to use a whole model structure, but it’s not necessary for your project. It will only make working on the project a whole lot easier.
Correct, that’s a good way to think about a VP project. And how you divide your work inside that tree structure is fully up to you.
Hope this can give you some ideas.