I’m not sure if it’s just me but I’m not unable to do anymore then 25 chat commands using the plugin before ChatGPT says i need to wait a few hours. This makes plugin testing really difficult. Previously it was unlimited and based on GPT 3.5. Is it possible to get the previous plugin with the 3.5 settings? cc @logankilpatrick
Also, is there a process for changing the domain after the plugin is live? I’m currently I’m using the dev version for the live plugin which means there is an occasion error as i do development. I can’t seem to figure out how to change the live version and dev version urls.
Another alternative might be to allow chat usage to be billed to our api accounts by usage. Presumably part of the problem is current chat usage is either free or fixed subscription, so openAI is looking to limit it’s cost exposure. Some developers need, and might be willing to pay for, more access.
My plugin ChatShareX works in a similar fashion, and this has not brought about the limit. I am 99% sure it means userInput and AI response. If you prompt the chatGPT and the AI assistant triggers 10 request to the plugin, it should only count as one message from the User and one message from chatGPT.
I’m having a fantastic time developing the plugins and working with ChatGPT
“A Whole New World” plays in the background hahahah
I was super excited to see (confirm?) that the plugins are now using GPT4. However, the limit of 25 messages per 3 hours has severely decreased my velocity in developing and testing plugins.
Until now, the message output speed is slower than 3.5+, so development already involved a fair bit of chillin while the outputs were generated. But this a limit of 25 messages every 3 hours is not nearly enough to test the flexibility of the plugins I’m making.
Even if there was an option to use my GPT4 API Key (still waiting, help a dev out ), I’d jump at the ability to be charged per message for testing my plugin.
TLDR: It’s hard to release a plugin that facilitates high quality interactions between the API <> ChatGPT and ChatGPT <> User (supporting/anticipating prompts) with a 3 hour throttle.
I’m concerned this may no longer be the case. Just hit the limit and I couldn’t count more than ~10 messages in the past few hours. These “request chains” counting as more than one message would explain it.
Sorry my post wasn’t worded well. Plugin makes multiple requests to the plugin’s API, and I’m not sure how that is being treated against the GPT-4 3-hour limit. Last time I hit the limit it appeared that I had sent fewer than 25 messages. I started a tally and time log on a note card to double-check my usage next time I hit the limit.
Just hit the usage cap after 7 messages, so it seems like multiple requests to the plugin’s API do count towards the 25 message usage cap, albeit not at a 1:1 ratio.
That’s not what I found unfortunately :(, if it calls the plugin 10 times it won’t count, but if it executes the plugin 10 times and after each time, it has to output something back to you in form or response or a followup, this will count, so 10 plugin call + 1 response per call, will count as 20. that’s my observation.
Would be nice to see a plugin developer dashboard.
), I’d jump at the ability to be charged per message for testing my plugin.