LLM-Assisted Modeling of Semantic Web-Enabled Multi-Agents Systems with AJAN
Hacane Hechehouche, Andre Antakli, Matthias Klusch
TL;DR
AJAN addresses the challenge of modeling Semantic Web-enabled MAS with RDF/RDFS/OWL and SPARQL Behavior Trees by introducing the AJAN-Editor, an LLM-assisted development environment that generates SPARQL-BTs and SPARQL queries from natural language. The system combines a modular SBT node factory, action and entity linking, and semantic search over AJAN documentation to support end-to-end agent engineering in offline and online modes. It demonstrates improved accessibility, modularity, and knowledge-graph independence, enabling non-experts to design and deploy semantic web agents in dynamic environments. The shelf-assembly MOSIM/Blocks World demonstration showcases practical integration with 3D simulation and interactive NL interfaces, underscoring the approach's potential to broaden adoption of Semantic Web MAS tooling.
Abstract
There are many established semantic Web standards for implementing multi-agent driven applications. The AJAN framework allows to engineer multi-agent systems based on these standards. In particular, agent knowledge is represented in RDF/RDFS and OWL, while agent behavior models are defined with Behavior Trees and SPARQL to access and manipulate this knowledge. However, the appropriate definition of RDF/RDFS and SPARQL-based agent behaviors still remains a major hurdle not only for agent modelers in practice. For example, dealing with URIs is very error-prone regarding typos and dealing with complex SPARQL queries in large-scale environments requires a high learning curve. In this paper, we present an integrated development environment to overcome such hurdles of modeling AJAN agents and at the same time to extend the user community for AJAN by the possibility to leverage Large Language Models for agent engineering.