search

sagemathinc / cocalc-docker

DEPRECATED (was -- Docker setup for running CoCalc as downloadable software on your own computer)
https://cocalc.com
Other
398 stars 103 forks source link

Change/adjust/disable Token limit per user or globally not possible? #218

Open bombaglad opened 5 months ago

bombaglad commented 5 months ago

Hi! Relating to Issue #217 , after solving the payment issue and using ChatGPT for a while, there's a new error which I can't seem to be able to fix via the settings-GUI.

"Error: You may use at most 50000 tokens per hour. Please try again later."

It's definitely NOT a limitation by the OpenAI-API, as a different user on the same server is still able to use ChatGPT normally.

Is there a proper way to disable the check on a self-hosted docker instace, or to change the limits? I couldn't find it in the documentation.

image

Seems to me like it is related to some of these files, to prevent abuse and high costs: src/packages/server/llm However I am basically the only user on this server and want to use the API as much as I need, while keeping track of the costs through OpenAI's website myself.

Thanks in advance!!

williamstein commented 5 months ago

These parameters are hardcoded here in the source code: https://github.com/sagemathinc/cocalc/blob/master/src/packages/server/llm/abuse.ts

You'll either have to edit them directly and rebuild your cocalc-docker install, or wait until we make those parameters configurable in admin settings.

https://github.com/sagemathinc/cocalc/issues/7242

williamstein commented 5 months ago

Incidentally, fully rebuilding the cocalc-docker server after making edits to any source code just means doing this as root (which should take about 5 minutes):

umask 022
cd /cocalc/src
time pnpm build

Then restarting. Of course, there are much faster ways to build just the code for a particular package. However, the above would work for anything.