The Vision: Democratizing 3D Design
Complex CAD (technical drawing) software often presents a steep learning curve, creating a barrier between a designer’s idea and a digital model. This project explores a new frontier in language-to-geometry interfaces, building an automated pipeline that transforms natural language instructions into editable 3D models with fully defined parameters.
 |
 |
|---|---|
| Low Pressure Turbine Blade | Design Lamp |
How It Works: A Prompt-Driven Modeling Pipeline
- Instruction Input — the user provides verbal instructions for a 3D part.
- LLM Interpretation — an LLM (via OpenAI API) translates these into executable Python code.
- Code Execution — the script runs in FreeCAD to automatically construct a parametric CAD model.
Key Features & Innovations
- Automated Pipeline — end-to-end workflow from text input to CAD model, orchestrated with LangChain.
- AI-Powered Error Correction — a Retrieval-Augmented Generation (RAG) component detects and fixes common CAD code errors (e.g., misapplied rotations, bad feature constraints).
- Parametric by Design — outputs are true parametric CAD models, not static meshes, making them editable and reusable for design iteration.
Tech Stack
- Core AI: OpenAI API, LangChain
- AI Technique: Retrieval-Augmented Generation (RAG)
- CAD Software: FreeCAD
- Language & Environment: Python