How would you design an orchestration layer for a system with multiple specialized AI agents (e.g., planner, retriever, executor) where partial failures are common?

When would you store a summary of a document in the vector DB but retrieve the full text for the LLM?

How do you prevent an Agent from getting stuck in an infinite loop (e.g., Tool A keeps calling Tool B, which calls Tool A)?

Why would you index sentences but provide the surrounding paragraph as context?