Hermae

Hermae was an AI assistant built to help enterprise teams generate frontend code by leveraging their design systems. The core idea was to use artificial intelligence to understand a team's specific design components and patterns, then produce production-ready code that adhered to those standards. It aimed to bridge the gap between design documentation and implementation, specifically targeting the challenge of providing large language models with the necessary context from a proprietary design system.

The tool worked by converting Storybook documentation into Markdown format. This converted documentation was then used to create vector embeddings, which powered a Retrieval-Augmented Generation (RAG) system. This approach allowed the AI to access and utilize the specific rules and components of a team's design system when generating code. While the concept of coupling design systems with AI was validated by the subsequent emergence of tools like Vercel v0 and Cursor, the specific implementation of Hermae was an experiment that has concluded.

Hermae was most relevant for frontend developers and product teams working with established design systems who wanted to accelerate their UI development process. Its primary use case was taking the defined components in a Storybook and enabling an AI to generate code that consistently matched those specifications, reducing manual coding time and potential inconsistencies. The project provided valuable insights into the evolving space of AI-assisted development, though the service itself is no longer active.