LangGraph Deployment

[qichaolan]

We want to leverage LangGraph to build an AI assistant with a main or supervisor agent, multiple sub-agents are deployed across different networks or GKE clusters? Some of the sub-agents must access isolated data sources that cannot be public-facing, while we want to maintain a single exposed interface for customer interaction. What architectural approach or mechanisms would allow seamless communication between the main agent and sub-agents across these isolated environments while maintaining security and data integrity?