Loop Engineering for RAG Generation: An LLM Cascade from a Cheap Local Model Up to a Hosted Flagship⭐8

Optimize RAG generation costs by cascading from a cheap local model to a hosted flagship, escalating only when validation fails.