In the field of Human-Computer Interaction (HCI), the development of interactive devices represents a significant area of focus. The advent of novel hardware and advanced fabrication techniques has underscored the demand for specialized design tools that democratize the prototyping process for such cutting-edge devices. While these tools simplify the process through parametric design and simulation, they typically require a certain learning curve and often fall short in facilitating creative ideation. In this study, we employ fluidic computation interface as a case study to investigate the potential of augmenting design tools of physical devices with Large Language Model (LLM) agents. Enhanced by LLM agents, the Generative Design Tool (GDT) can comprehend the capabilities and limitations of newly developed devices; it can propose varied, insightful, and practical application scenarios, and recommend device designs that are technically and contextually appropriate. Furthermore, it generates the necessary design parameters for the traditional part of the design tool to visualize results and produce support files for fabrication. This paper outlines the GDT's framework, implementation, and performance, while also contemplating its prospects and the obstacles encountered.

翻译：在人机交互领域，交互式设备的开发是一个重要的研究方向。新型硬件与先进制造技术的出现，凸显了对专用设计工具的需求，这些工具旨在降低此类前沿设备的原型开发门槛。虽然现有工具通过参数化设计和仿真简化了流程，但它们通常存在一定的学习曲线，且在激发创造性构思方面往往不足。本研究以流体计算界面为案例，探讨了利用大型语言模型代理增强物理设备设计工具的潜力。经LLM代理增强的生成式设计工具能够理解新型设备的功能与局限；提出多样、深刻且实用的应用场景；并推荐在技术及情境层面均适宜的设备设计方案。此外，它还能为传统设计工具生成必要的设计参数，以可视化结果并生成制造所需的支持文件。本文阐述了该生成式设计工具的框架、实现与性能，同时对其发展前景与面临的挑战进行了展望。