Most CAD workflows still start with a manual sketching step; CADAM experiments with flipping that pipeline by letting users describe a part in plain English or upload reference images and receive a parametric OpenSCAD model instantly in the browser. The core insight is combining LLM/AI-guided shape synthesis with an OpenSCAD WASM engine and parameter extraction so models are both editable and exportable without heavyweight tooling.
What Sets It Apart
- Text + image → parametric OpenSCAD output: Instead of producing only meshes, CADAM aims to generate editable OpenSCAD code with recognized parameters (sliders) so dimensions can be tuned without re-querying the AI. This keeps models reproducible and modifiable.
- Browser-first architecture: Uses OpenSCAD compiled to WebAssembly and Three.js for live preview, allowing rapid iteration in-browser without installing native CAD software. That lowers the barrier for quick prototyping and classroom demos.
- Integration with AI and server functions: Uses Anthropic for the generative step and Supabase edge functions (and ngrok for local dev) to handle image URLs and API calls—so while the renderer is client-side, generation relies on an AI backend.
Who It's For & Tradeoffs
Great fit if you want a quick way to turn concept sketches or text prompts into parameterized models for 3D printing, education, or prototyping, and prefer editable OpenSCAD output rather than a static mesh. It’s also useful for people learning parametric CAD because generated models expose sliders and dimensions.
Look elsewhere if you need industrial-grade CAD (complex assemblies, precise mechanical tolerances, constraint-driven sketches) or want a fully offline pipeline: CADAM relies on OpenSCAD semantics and an AI service (Anthropic) for generation, and the project is distributed under GPLv3 which affects commercial redistribution.
Where It Fits
Think of CADAM as a bridge between conversational design prompts and editable parametric code: faster than hand-modeling simple parts, more flexible than black-box mesh generators, and well suited for makers, educators, and rapid iteration during concept exploration.
