Structured rendering inputs: preserved vs. changed constraints, material/fixture specs, and presentation-protocol grounding

For automated rendering to be usable in a client deliverable, the inputs need to be precise. We would like to pass: a prompt grounded against our organization's visual presentation protocol, an anchor photo reference (file_id), preserved vs. changed architectural constraints, and exact material and fixture specs.

That lets the render keep what must stay the same (structure, layout, dimensions) and change only what we specify, with the specific finishes and fixtures we choose.