CityFlow: An Advanced Case-based Design System Integrated with Expert Workflow and LLM Retriever

Urban environments present increasingly complex challenges, which demand innovative, interdisciplinary solutions. This paper introduces a novel AI-enhanced methods-as-case framework, designed to support decision-making in urban design. Addressing the limitations of traditional rule-based and case-based design paradigms, our framework combines these approaches to enhance case retrieval and interpretable revision processes. The resulting CityFlow platform provides a user-friendly, low-code environment for data analysis, urban evaluation, and simulation, fostering collaboration and knowledge sharing among diverse users. By integrating LLMs and a knowledge graph, CityFlow facilitates rapid prototyping and iterative refinement of urban design workflows. We discuss the platform’s current capabilities, aligned with early stages of AI advancement (L1-L2), and outline future directions towards more autonomous AI assistance (L3), ultimately redefining human-computer interaction in urban science and paving the way for more sustainable and collaborative urban design and research.