Interesting deconstruction of agents Most agents are prompt + LLM + output parser in a while loop This makes that explicit by forcing users to implement the while loop – more customizability

→ View original post on X — @hwchase17,