I built an LLM Request Cascade proxy that auto-switches models before you ever timeout

You're mid-task in Claude Code. You hit enter. Then... nothing. 12 seconds later, either the response...