Consider adding a boolean to the new model dialog with "Load all Models" with the help text of "If this field is toggled, we will call the API to retrieve all models from the endpoint; you can edit the model names, inactivate unwanted models, in the edit dialog." When retrieving all the models, you should check for ones already loaded by that endpoint, only adding new ones. To prevent the retrieval of models from the API that are not wanted, you should introduce a new field in the model's table/editor to disable or deactivate them rather than remove them.
This feature would allow the retrieval of all models available at the endpoints, leveraging the APIs of each provider. If you have added a few models to Ollama, you could grab them all quickly and then edit/inactivate/rename them after retrieving them. It there a new model from Openai? But you forgot the technical name? Just get all, then manage them in your interface.
tkg61
commented
1 month ago
The Feature
Consider adding a boolean to the new model dialog with "Load all Models" with the help text of "If this field is toggled, we will call the API to retrieve all models from the endpoint; you can edit the model names, inactivate unwanted models, in the edit dialog." When retrieving all the models, you should check for ones already loaded by that endpoint, only adding new ones. To prevent the retrieval of models from the API that are not wanted, you should introduce a new field in the model's table/editor to disable or deactivate them rather than remove them.
This would need to be coupled with the feature outlined here: https://github.com/BerriAI/litellm/issues/4466
Motivation, pitch
This feature would allow the retrieval of all models available at the endpoints, leveraging the APIs of each provider. If you have added a few models to Ollama, you could grab them all quickly and then edit/inactivate/rename them after retrieving them. It there a new model from Openai? But you forgot the technical name? Just get all, then manage them in your interface.
Twitter / LinkedIn details
No response