Mediating Modes of Thought: LLM's for design scripting

TL;DR

This work investigates using Large Language Models as mediators between designer intent and parametric scripting to broaden access to computational design. A multi-agent system converts natural language prompts into geometric logic, which is then mapped to Grasshopper scripts via a JSON schema. Case studies on a truss, umbrella, and suspension bridge demonstrate the approach's potential to capture design intent and generate runnable scripts, while highlighting reliability and mapping challenges at higher complexity. The results suggest a promising direction for integrating conversational, AI-assisted mediators into design environments, with future work focusing on conversational interfaces, multimodal inputs, and improved robustness.

Abstract

Architects adopt visual scripting and parametric design tools to explore more expansive design spaces (Coates, 2010), refine their thinking about the geometric logic of their design (Woodbury, 2010), and overcome conventional software limitations (Burry, 2011). Despite two decades of effort to make design scripting more accessible, a disconnect between a designer's free ways of thinking and the rigidity of algorithms remains (Burry, 2011). Recent developments in Large Language Models (LLMs) suggest this might soon change, as LLMs encode a general understanding of human context and exhibit the capacity to produce geometric logic. This project speculates that if LLMs can effectively mediate between user intent and algorithms, they become a powerful tool to make scripting in design more widespread and fun. We explore if such systems can interpret natural language prompts to assemble geometric operations relevant to computational design scripting. In the system, multiple layers of LLM agents are configured with specific context to infer the user intent and construct a sequential logic. Given a user's high-level text prompt, a geometric description is created, distilled into a sequence of logic operations, and mapped to software-specific commands. The completed script is constructed in the user's visual programming interface. The system succeeds in generating complete visual scripts up to a certain complexity but fails beyond this complexity threshold. It shows how LLMs can make design scripting much more aligned with human creativity and thought. Future research should explore conversational interactions, expand to multimodal inputs and outputs, and assess the performance of these tools.

Figures (7)

• Figure 1: A diagram that outlines the flow of the prototyped system, leading from the user input to the final script.
• Figure 2: A table of shortened responses from the first two agents to the prompt “a truss”. The result of the first agent serves as the input for the second.
• Figure 3: A visual script and CAD preview that was parsed from the third agent's JSON onto the Grasshopper canvas for the prompt “a truss”.
• Figure 4: A table of shortened responses from the first two agents to the prompt “an umbrella”. The result of the first agent serves as the input for the second.
• Figure 5: A visual script and CAD preview that was parsed from the third agent's JSON onto the Grasshopper canvas for the prompt “an umbrella”.