LLM Throws a Syntax Error Tantrum: Teaching AI to Craft Graph Style Scripts

This blog post explores the application of Large Language Models (LLMs) in generating Graph Style Script (GSS) for customizing graph visuals, demonstrating how LLMs can be trained to understand and create GSS code within Memgraph Lab. GSS is an in-house developed language that enables users to style graphs by altering the appearance of nodes and edges based on defined rules, similar to CSS for web design. The author discusses the process of training an LLM to generate GSS, highlighting its initial challenges, such as syntax errors and unsupported functions, and how these were addressed through iterative feedback. The post offers insights into using LLMs to create complex graph styling rules, emphasizing the need for comprehensive documentation and examples to facilitate learning both for LLMs and users. It concludes by encouraging users to experiment with GSS in Memgraph Lab and suggests potential enhancements for the GSS language based on the LLM interactions, showcasing the mutual learning process between AI and human developers.