about 429.
// friends don't let friends pay per token.
HTTP 429 means Too Many Requests. This network is owned by us so there is no limit.
The big-name providers seem to think rate-limiting indie devs into oblivion is a viable long-term strategy. So a few of us, namely me and my Dad and my best friend, built a network to pool out of work GPUs and put a tiny OpenAI-compatible API in front of them a 429.
It's a single endpoint, with an open-weight model, served by whatever GPUs friends have plugged in this week. If nobody's home, you get a 503 and an apology in JSON.
Keys are handed out personally, with no signup form for keys other than the form on the homepage. Issued keys are free at point of use because the friend running the network is paying the power bill out of love and a vague hope you'll buy them a coffee someday. Oh and they aren't greedy...
Should you use this in production? Probably not yet. For the side project you've been meaning to ship for six months but keep getting rate-limited on? Bet.