Implement a way to communicate current scene status to the model:
Visual Language Models could benefit from overview (top-down) with some scale.
Scene API to query placement overall or intersection with a given shape (box). The intersection would be so that the model decides whether it is suitable to add a new model in the place.
Implement a way to communicate current scene status to the model: