mirror of
https://github.com/open-webui/open-webui.git
synced 2026-05-08 04:16:03 -05:00
[GH-ISSUE #7239] enh: customisable model external icon in the selector #14672
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 @hg0428 on GitHub (Nov 22, 2024).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/7239
If you connect a locally-run OpenAI-compatible API, the models from it show up as external even if they are local. For example, models from LM Studio or Ollama's experimental OpenAI API.
This is one of those small but very annoying problems with Open WebUI. Let's get this fixed.
Also, in set task model, all my local models show up in the external models dropdown.
Perhaps there should be a local->external toggle for each individual connection? Or perhaps custom labels (e.g. Groq, Ollama, my custom API endpoint, LM Studio). I think this later option is the best. Just have the same little link symbol but the text can be customized. No need to change the icon, I think it works for letting you know the source.
@hg0428 commented on GitHub (Nov 22, 2024):
Custom labels would also help you distinguish between models if you have the same model from multiple providers. For example, I have Llama3.1 8B locally and Llama3.1 8B via Groq and there is nothing to distinguish them besides manually changing the name.
@greg-d128 commented on GitHub (Nov 26, 2024):
I was about to put in similar request. I have two connections to ollama, one local to the laptop and another that points at a box more suitable to running llm. I have multiple models that are very difficult to distinguish.
Ideally though, I would like to put a tag, or an icon on a connection and have all the models from that connection inherit these. That way it will not need to be constantly kept up to date. Even having an option of modifying the model name based on the host it is coming from would help.
Hmm.. I wonder if model files are accessible, maybe a small script that automatically changes model names based on the IP they are coming from?