From Treatise’s Geometry to an AI-Driven Procedural Framework for Shape Grammars

This paper presents a procedural, AI-assisted workflow for translating the proportional and com-positional logic of classical architecture into a parametric modeling system based on shape grammar theory. Grounded in Scamozzi’s L’Idea dell’architettura universale, the research formalizes a corpus of architectural elements by encoding their geometric rules into modular, reproducible data structures. Leveraging Blender’s open-source Python API and large language models for natural-language-to-code generation, the method has shifted from primitive-based approximations to analytically defined Bezier and NURBS profiles, enabling accurate reconstruction of moldings, entasis, and proportional dependencies. The resulting Blender add-on, Ars Modulandi 2.0, provides an accessible tool for generating classical components while enhancing transparency and scientific traceability. Overall, the study demonstrates how AI-driven scripting can act as a mediator between textual description, mathematical reasoning, and architectural form, opening new avenues for procedural heritage reconstruction and computational design.