Should agent architectures be fixed… or generated at runtime? [D]

About this article

I’m sharing a research prototype exploring a different approach to LLM-based multi-agent systems. Most current agent frameworks rely on fixed architectures: a predefined set of agents, roles, and interaction patterns. In this work, I explore an alternative: → dynamically generating and orchestrating agents at runtime depending on the task. Use case: root cause analysis (RCA) in microservice systems using logs, metrics, and traces. Approach: - A Parser converts a natural language incident into...