How does communication between them work?

[marhono05]

Hi! I’m interested in how the agents coordinate with each other. How do you handle conflicts between agents when they generate different solutions for the same task? Is there a "final decision" agent or some kind of ranking system?

[Feroan101]

There isn't a built-in ranking system where agents compete and the highest-scoring solution automatically wins. The "Agents Orchestrator" delegates tasks, collects the agents' outputs and confidence levels, and then synthesizes the results.

When agents disagree, the architecture uses contradiction detection followed by an arbitration step to determine which result to use, with human input when the conflict can't be resolved automatically. So there is a final decision layer, but it's handled through orchestration and arbitration rather than a simple numerical ranking.