Better Usage Limit Handling #1130
Description
A thread from Discord:
It would be great if when recources are used up, it would simply respond with exactly same json structure as the response from any model, but with custom message telling the user about end of recourses and suggest to upgrade the account or buy more resources if doable.
That's not a bad idea. It would be kind of annoying to implement though since usage limits are supposed to be a concept that's general to all driver interfaces rather than just AI chat, but maybe each interface can define a "when usage has run out" implementation.
So, that's the idea behind this issue; add an implementation of the puter-chat-completion interface that informs the user that their allotted amount for costly services has run out. We have a fake chat service (for test_mode and when moderation fails) that might be a good reference for this.