How to Route LLM Requests: Building a Cost-Optimized Fallback Chain

I got paged twice in the last two months because a single model dependency went down. One was a...