defaultLlm

The deployment default: either an LLM name, or a role from llms or roles. It's good practice to override this in configuration.

A model name is resolved once, against the services registered at startup. A role is resolved per call, through the same RoleResolver chain as any other role - which is what lets a deployment whose key arrives at runtime have a working default without a restart. A model name that nothing registers falls back to the role chain too, so default-llm: gpt-5.6-luna keeps working whichever way it is meant.