Spent quite some time normalizing APIs, handling tool-calls, and managing prompt caching, but the end result sounds very cool: You always get the absolute best value for your \$ at the exact moment of inference.
Currently runs perfectly on a Roo and Cline fork, and on any OpenAI compatible BYOK app (so kind of everywhere)
Feedback very much welcomed! Please tear it apart: [https://makehub.ai](https://makehub.ai/)