Implementing Fall-Back Models for Agent Reliability

When selecting models for an agent, it would be nice if we could select a set of fall-back models. That way an agent could still function even if the primary provider is experiencing some issues/downtime.