As of the last few weeks, all chatbot Spaces claiming to utilize the GPT4 API have invariably either returned instant errors (with a message that the Space is “too busy”) or get stuck in an infinite loading screen, despite them having worked perfectly for the previous four months, even if a functional OpenAI API key is prompted when required. This issue is not localized to just me, as several others have also expressed complaints here: Chat-with-GPT4 - a Hugging Face Space by yuntian-deng
I have contacted HuggingFace about this issue, they have yet to respond or even acknowledge my message. This is a problem that should have been seen quickly and fixed immediately, yet nobody else seems to be aware of it. Is it just localized to a small number of people, me included?
I can confirm that this is happening with other APIs too. Any of my spaces that make a call to an external API are stuck on an infinite load screen. Huggingface doesn’t really seem to have any SLAs in place. I think their one SLA is - we are open source so live with it (pro account memberships at $9 a month don’t seem to matter).