satyakottu080704/ACP-Reconstruction — reverse-engineered prompt

Build me a Python app that turns a hand drawn asbestos survey sketch into a clean digital floor plan in the Acorn style. I want to upload an image, have the model detect rooms, walls, doors, stairs, floors and ACM areas, then reconstruct tidy geometry with straight snapped walls, closed rooms, shared walls, no gaps, and doors attached to the correct walls. It should also read handwritten room names, room numbers, sample IDs and floor labels using the OpenAI vision model from my env file, then merge that into the final plan.

Please make it easy to run from the command line, expose it through an API, and include a simple Streamlit interface for testing. Export results to JSON, SVG, PNG, PDF, DXF and Visio when available. Use the trained weights from the expected models folder, keep the image size consistent with training, and preserve the current review only publishing flow. I also want tests that can run without a GPU or Visio installed. Use the existing repo layout and look up current docs online if you need to.