the maximum we are ready to add to our product is an API to plug your own processing module. Let’s say it will pass to the module all JS includes and tag content.

But I see (at least) two pitfalls:
1. JS code requires an access to DOM structure of the main document. PD4ML internally does not implement DOM.
2. JS addresses CSS styles of the main document. PD4ML does not implement any standard ways to request CSS properties.

May be there are some other issues I do not know yet.

So as you may see a support implementation is going to be limited and isolated from the main document content. On practice all dynamic image generation parameters would need to be inserted as JS constants to tag, for example.