Text-to-Diagram for AI Systems: Designing a DSL an LLM Gets Right (zerowithdot.com)

A new tool called Gnomon has been developed to bridge the communication gap often found in multidisciplinary teams. In environments where stakeholders view projects from vastly different perspectives, Gnomon allows teams to create diagrams from natural language descriptions, ensuring all parties understand the project's architecture. The Gnomon system utilizes a domain-specific language (DSL) called gnomocode and features a compiler that validates statements against a carefully constructed ontology, establishing clear communication and reducing misunderstandings. Significantly, Gnomon addresses common pitfalls of traditional diagramming methods by distinguishing between different types of information and ensuring that overlapping interpretations are minimized. The system generates a standardized graph from the gnomocode, which can be depicted visually and queried as needed. By incorporating confidence levels into the language, it allows users to express uncertainty, thus fostering more nuanced conversations about project requirements. Built on Python using tools like Lark for parsing and rdflib for RDF graph creation, Gnomon positions itself as a vital resource for AI and ML professionals, aiming to enhance clarity and precision in project design and implementation.