Minor improvements

[oskarst]

Small note, needs some documentation, currently it's not very clear which OpenAI model you should use for this extension, and if chosen incorrectly it will cause errors. for example o1-mini atm does not work and throws error during API request to OpenAI.

The API Model select input could also use some sorting as atm there are many choices but they are all over the place, simple alphabetical sort should clear it up nicely, could also make the case that preview models should not be used.

[ryansunxl]

@oskarst thanks for using this module and the feedback.
I agree an alphabetic order is handy.
The module was tested w/ gpt modules, I was not aware and surprised that o1 would cause issues.
I don't have access to openai anymore, can you share the error message w/ o1?

[cmuench]

In the tool n8n they also filter the models.
Maybe this code reference is helpful: https://github.com/n8n-io/n8n/blob/46dcce341fbfa1c2a44a08f3dc93f1f8f16808c8/packages/%40n8n/nodes-langchain/nodes/llms/LMChatOpenAi/methods/loadModels.ts#L18

[ryansunxl]

I tried to filter models list copying n8n code per @cmuench suggestion, but I think that code snippet is outdated (e.g. 4o is missing).
I also think it is hard to filter the list properly, especially considering people may have their own fine-tuned models with special id.

The current model list is quite long with many models (e.g. the audio and image ones) not appropriate for this use case, I'm thinking about turning it back into free form text field.

Anyone has any thought on this?

[ryansunxl]

@oskarst I'll update code to stay compatible with the openAI api update. o1-mini will not work, you have to choose a model that supports developer role message https://community.openai.com/t/developer-role-not-accepted-for-o1-o1-mini-o3-mini/1110750/7

I advise you using a gpt model.