mirror of
https://github.com/open-webui/open-webui.git
synced 2026-05-06 19:08:59 -05:00
[GH-ISSUE #5119] enh: Enable / Disable indivual OpenAI endpoints #13868
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: github-starred/open-webui#13868
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @aaronbolton on GitHub (Sep 3, 2024).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/5119
Originally assigned to: @tjbck on GitHub.
Is your feature request related to a problem? Please describe.
Feature request, disabling OpenAI API's disables all, I currently have several (OpenAI, Bedrock Proxy & Pipelines)
Describe the solution you'd like
I would like to be able to enable or disable individual OpenAI endpoints under Admin Panel>Setting>Connections>OpenAI API
Describe alternatives you've considered
disabling all
@adamoutler commented on GitHub (Oct 6, 2024):
I concur. This should be a feature. The only other option is to whitelist each model you wish to use and then logout and login as a user. This isn't practical.
There should be a global blacklist. There are models which are not suitable for any type of use and they should be hidden, even from admins. For example:
There are many good reasons to have a way to disable models, not just OpenAI endpoints, but the OpenAI endpoints are most certainly the largest portion of spam in everyone's list.
@aaronbolton commented on GitHub (Oct 7, 2024):
This is exactly why I'm asking I know have such a large list with combination of OpenAI, Bedrock Proxy, Pipelines and now groq, maybe also a tag in the list to show which endpoint the models is from.
@tjbck commented on GitHub (Nov 12, 2024):
Added to dev! Testing wanted here!
@aaronbolton commented on GitHub (Nov 12, 2024):
Fantastic thank you, the only issues I have found are, but aren't directly related to the functionality of the ability to enable/disable endpoints
@tjbck commented on GitHub (Nov 12, 2024):
@aaronbolton try the latest dev, prefix is definitely being applied even with the filter for both openai and ollama, I believe you're confusing the model id, with the model name, which does not contain the prefix id.
@aaronbolton commented on GitHub (Nov 12, 2024):
@tjbck Ah I see, I'm using the latest docker dev container is there any difference? but if I include the prefix yes it works for OpenAI connections
So this work for OpenAI connections
For Ollama I've added a prefix but that's not showing
@aaronbolton commented on GitHub (Nov 14, 2024):
@tjbck is there any specific testing or debugging I can do to assist with this?
@adamoutler commented on GitHub (Nov 14, 2024):
I tested and it worked for me. However, this is again, another whitelist. The blacklist feature is still missing. In the event there is a new model, I'd like to use it, but because whitelist is the only thing available, I won't know there is a new model available.
Eg. If I whitelist gpt-4o and they release gpt-5 I will never see it. However i could blacklist all gpt3.5 and others then when a new model drops it will just appear in the list for testing.
@adamoutler commented on GitHub (Nov 14, 2024):
I believe we need the whitelist and blacklist in the Workspace rather than on the sources. The reason I say this is certain functions install pipes on all models, eg. MCTS or Visual Tree of Thoughts function which duplicates all Ollama models. These can't be whitelisted on ollama. While I don't mind if users do a MCTS I'd rather they use smaller models.
I hate to say it, but I don't think this is a good implementation as-is. I mean, it's better than it was, but I think this will always be incomplete without a workspace whitelist and blacklist feature.
@aaronbolton commented on GitHub (Nov 14, 2024):
Did the Ollama functions work for you? Such as the prefix and whitelist I've not been able to get these working it works fine for OpenAI connection just not ollama
@adamoutler commented on GitHub (Nov 14, 2024):
Yes. Refresh the models list on connections tab and update button next to connection. It renamed all models.
However it only allows a whitelist which is ok, but only 1/2 of a comprehensive solution.
@aaronbolton commented on GitHub (Nov 14, 2024):
I must be going crazy i dont seem to have an update button next to connection.
@adamoutler commented on GitHub (Nov 14, 2024):
It's on the models screen.
@aaronbolton commented on GitHub (Nov 15, 2024):
ive tried this and i see the prefix when its updating, but from the drop-down list on the main chat page its not showing
@adamoutler commented on GitHub (Nov 15, 2024):
The drop down box is the name, not the ID. This changes the ID.