[Feature Request] Load custom models as ENV variables in docker or some other way to load custom models.

[shizuo-x]

🥰 Feature Description

I am using own locally run model api under open ai name, but the models for open ai are hard coded into the ui, and without going into settings and changing the custom model name each time will output a wrong model error.

I is not a problem if my PC has only one model loaded, but multiple loaded model need to specify what model to use in the api request. I am using this chat UI as front end to expose on a website to the public. And a common man wont know what model are loaded on my pc and that he has to go to settings to manually load a model.

Please provide a feature to load models on the server side and show it as a list to the public to select from.

🧐 Proposed Solution

Please provide a feature to load models on the server side and show it as a list to the public to select from.

📝 Additional Information

No response

[QAbot-zh]

#5001

[shizuo-x]

use: CUSTOM_MODELS as env variable in docker and put all models together separated by comma was solution for now.
I dont understand chinese, it is a bit hard to understand and look the translations.

[Kosette]

TL;DR the custom model format is <model_name>@<provider>.

If you are using an OpenAI-compatible API, your custom model can be written as <model_name>@OpenAI, similarly, if it is a Claude-compatible API, it is written as <model_name>@Anthropic. Note that the provider name in @ is capitalized. If you are using a built-in model name, you can use lowercase.