Natural Language Based CAD Design Automation Using Large Language Models

This repository contains the full implementation for the thesis “Query2CAD: Automated CAD Model Generation and Validation Using Large Language Models.” The project focuses on generating FreeCAD models directly from natural-language descriptions. It integrates a structured prompt builder, an example library and strict modeling rules. A two-level validation pipeline checks both code structure and final geometry. Invalid outputs are automatically repaired using a controlled regeneration loop. The system evaluates multiple LLMs across moderate and challenging query sets. All scripts, datasets, example models and validation logs are included for reproducibility. Performance metrics cover success rates, timing behavior and common failure patterns. It provides a complete, extensible framework for AI-driven parametric CAD generation.