A Hybrid Role-Based Reference Architecture for LLM-Enhanced Multi-Agent Systems
Abstract
Large language models (LLMs) are transforming how we build multi-agent systems (MAS); yet, many LLM-centric frameworks still lack the engineering rigour that agent-oriented software engineering (AOSE) provides, resulting in systems that are powerful but difficult to maintain and scale. In our previous work, we critically examined the “role” concept across definition, specification, and implementation, and proposed a preliminary hybrid role-based architecture where roles are treated as first-class run-time entities that support four different action implementation types. However, that earlier work remained at a conceptual level: it identified the need for typed actions and runtime roles but did not provide a formal meta-model specifying how these constructs relate to one another, nor did it offer a concrete realization or validation. Building on that foundation, this paper closes this gap by elaborating on and operationalising the proposed architecture as a reference architecture, making the role construct general, reusable, prescriptive and actionable. In particular, we define a role meta-model for LLM-enhanced agents that specifies the core role constructs, their interfaces, constraints, and interaction relationships—providing clear variation points for implementing roles consistently from design-time and run-time. We implemented this meta-model as a set of Java annotations, enabling roles and their relationships to be specified declaratively in code and validated at run-time. Our solution is framework-agnostic: any Java-based agent framework can adopt the annotations to expose roles, actions, and interaction points in a uniform way and thereby support LLM-enabled behaviors. We demonstrate the applicability of our approach by implementing a hotel reservation scenario in the SCOP framework, where each agent type is realized through dedicated role specifications and role implementations combining hybrid action types. Finally, we discuss practical design considerations—including deliberation-execution separation, action-type boundary decisions, and observability and debugging strategies—offering actionable guidance toward production-grade LLM-enhanced MAS.
Önder Gürcan
Senior Researcher
My research interests include multi-agent systems, collective intelligence, self-organization and self-adaptation, simulation of biological systems, distributed clock synchronization and behavioural economics.