The advent of Large Language Models (LLMs) has shown considerable promise in various fields,including engineering system design, due to their capabilities as few-shot or even single-shot learners. Thispaper investigates the integration of LLMs in the architectural design of fluid power systems for constructionmachinery. The primary contribution is the development of a methodology that transforms textual inputs intoformal architectural definitions, utilizing micro-templates within prompts to enhance output repeatability andconsistency. In an additional step LLMs can generate Python code that generates the system architecture, fromuser input, in a more narrow put more reliable design space. The iterative refinement process facilitated by LLMsallows for the expansion and optimization of design spaces. This can have a great impact on the early stages ofengineering design by automating the generation of system architectures.