Introducing GenXDefinition (GXD): A Grasshopper Definition Generator.
GenXDefinition (GXD) operationalizes large language models (LLMs) within Rhino/Grasshopper by synthesizing C# that assembles complete Grasshopper definitions. Given a textual description of a parametric task or an image, GXD generates code that instantiates components, assigns parameters, and wires connections on the Grasshopper canvas via the Grasshopper/RhinoCommon SDKs. The produced script is pasted into a Grasshopper C# Script component, where it executes to assemble the definition.
Prompt: The user writes a natural-language prompt describing the target object and any element. (e.g. “Generate an organic lamp with texture”).
LLM Code Generation: The model returns C# source code tailored to Grasshopper’s C# scripting environment.
Paste into C# Script: The user creates a C# Script component (found under Maths > Script) on the Grasshopper canvas and pastes the generated code into it. Remove output [a]. Add a toggle in input [0] and a button in input [1].
Run and Assemble: When the C# component runs, it creates the necessary components and parameters, and connect their inputs and outputs. The GH canvas will be populated with the components and connections as defined by the code.
Copy-Paste definition so to become functional.
Getting Started with GXD
To try GXD or to see examples, visit genxdefinition.com or search for it in the GPT store OpenAI.
As with any generative AI tool, it’s wise to review the output code. In practice, GPT-5’s thinking mode tend to yield better results. Sometimes prompts need tweaking, structured and short prompts help the model to produce valid graph layouts.
Troubleshooting & known limitations - GXD is an LLM-assisted generator, so it may occasionally produce compile or runtime errors (e.g., missing using directives, mismatched Rhino/Grasshopper API calls, version differences). This McNeel Forum thread will be the central place to report issues, discuss errors, and collect troubleshooting tips and fixes. When posting, please include Rhino/Grasshopper version, OS, your prompt, and a minimal snippet of the generated code that reproduces the problem.
Currently
This project is in active development. Current research will document ongoing experiments, benchmarks, and case studies from GXD deployments in architecture and interaction design. The research is led by R. Konstantinidis (Researcher) and N. Kourniatis (Academic Coordinator), will prioritise methodological rigor and reproducibility, with links to datasets, scripts, and preprints as they become available.
Proof of concept
GXD originated as a proof-of-concept in the M.Sc. in Interaction Generative Design at the Hellenic Open University, under the thesis “Synergies of Generative Design Systems.” It investigates large language models as mediators for prompt-to-parametric workflows in Grasshopper. The work was authored by R. Konstantinidis, supervised by A. Laskari, with academic coordination by N. Kourniatis.
