I'm currently developing a system with a steering agent and four sub-agents (A, B, C, and D) and facing a challenge with seamless transitions between these sub-agents. Ideally, I want to enable users to switch from one sub-agent to another during interactions. For instance, a user in a session with sub-agent A might want to switch to sub-agent B, and so forth. However, direct transitions coded like ${AGENT: B} in A and ${AGENT: A} in B are causing looping issues in the playbook.
This direct approach is also not scalable. For example, with four sub-agents, I need to create 12 transition examples to handle all possible switches. This could become unmanageable with more agents, as a system with ten agents would require creating 90 examples just to handle these transitions.
I'm looking for feedback or solutions that could help optimize this process, possibly through dynamic decision-making or another efficient method that avoids hardcoded transitions and extensive manual setup. Any suggestions or insights would be greatly appreciated!
LinkedIn
Twitter
