Comment by bandana
I'm making an OpenAI API proxy to stay within a spending limit (like 1$ per hour then a 429), AFAIK they only support "budget alerts" and I'm not comfortable releasing anything without a hard limit on the spend. https://github.com/goverture/goxy - Still a work in progress, I plan to support streaming as well and might support other providers if there's a demand for it.
OpenAI gives an allotment of free daily tokens if you agree to hand over the inputs as training. I’d love a proxy that places the limit just before you exhaust those free tokens, to avoid incurring any expenses for small hobbyist projects.