Good to meet you @pixelcruncher
If you have a look at the video I linked to in my previous comment, it's very easy. You just describe it with words, and it doesn't affect the rest of the render.
ComfyUI is a tool that helps you work with AI models, so you'd need to train a model (e.g. Qwen or Seedream) to do these kinds of operations for you precisely.
Yes, we do have an API (you can test it here https://rapidapi.com/myarchitectai-team-myarchitectai-team-default/api/architecture-rendering-api) but we didn't include the editing tools in it yet as they're still in beta.