Upload a part, ask anything about it — dimensions, features, manufacturability, tolerances — and get answers from the exact geometry the engineer authored, not a guess from a blurry render.
A STEP or IGES file is built for engineering software, not for language models. Upload one to an AI assistant and it chokes — the file is a dense graph of surface equations and topology, not a document. Show it a screenshot instead and the model can see the shape, but it can't tell you the part is exactly 8.000 mm thick or that two holes sit 30.000 mm apart. In engineering, the number is the whole point.
CADVERT bridges that gap. It reads the CAD file directly, extracts the exact geometry — dimensions, holes, fillets, wall thicknesses, tolerances — and writes it out as clean, structured text any LLM can reason over.
The key word is exact. CADVERT doesn't approximate the part into a mesh of triangles the way most tools do. It reads the true analytical geometry straight from the source, so every dimension it reports is the value that was actually in the file — not a measurement, not an estimate.
Two ways to use CADVERT, depending on who you are.
Upload your file — STEP, IGES, BREP, STL, or OBJ — and ask. "How thick is this wall?" "What size tap do these holes need?" "Is this part suitable for CNC machining?" CADVERT reads the exact geometry and the AI answers from real numbers, not guesses.
pip install cadvert
Three lines take a CAD file to AI-ready context you can drop into any model. Use it to build:
Whatever you build, the geometry underneath is exact. CADVERT does the translation; your model does the reasoning. Full API reference and code examples are on the developer docs.
Upload a CAD file and start asking questions — or install the library and build on it.