STEP File Analysis API (OCCT)

Co-planning checklist

This implementation is scoped to an API-only service for STEP ingestion and geometric extraction. Confirm the following before expanding feature fidelity:

1. Units: STEP files are assumed to be millimeters and converted to inches. If your data uses inches, set the conversion to 1.0.
2. Feature detection tolerance: Heuristic hole/pocket detection is used. If you need certified feature recognition, we should add a dedicated feature recognition pipeline.
3. Tool accessibility: Curvature sampling is used as a proxy. If you require voxelization or ray-casting, we can add that pass and GPU support.

Feasibility notes (unfiltered)

• Thin wall thickness: Reliable for parallel planar faces; curved/organic thin walls require a signed distance field or medial axis computation.
• Hole detection: Cylindrical faces are robustly identified; complex hole blends and cross-holes need deeper topology analysis.
• Pocket detection: Current implementation flags planar faces that are inward-facing, which is a proxy. True cavity extraction requires boolean subtractions or feature graph analysis.

Running locally

pip install -r requirements.txt
uvicorn app.main:app --reload

API usage

POST /analyze with a .step or .stp file.

Example response structure:

{
  "file_name": "part.step",
  "metrics": [
    {
      "key": "bounding_box",
      "display_name": "Bounding Box Dimensions",
      "description": "Length × Width × Height and min/max coordinates",
      "unit": "in",
      "category": "Geometry",
      "occt_extraction": "Bnd_Box from BRepBndLib::Add()",
      "feeds": ["stock sizing", "fixturing", "stock volume estimation"],
      "value": {
        "min": [0, 0, 0],
        "max": [1, 2, 3],
        "dimensions": [1, 2, 3]
      }
    }
  ],
  "units": {
    "length": "in",
    "area": "in^2",
    "volume": "in^3"
  },
  "notes": [
    "Units assume STEP geometry is authored in millimeters; converted to inches.",
    "Pocket and hole detection are heuristic and should be validated per part."
  ]
}

Render.com deployment

This repo includes render.yaml and requirements.txt. Connect the GitHub repo to Render and deploy the web service.

startCommand: uvicorn app.main:app --host 0.0.0.0 --port $PORT