fix: adjust regex in ModelService to recognize o1- models returned from OpenAI-API query

[juwalter]

Summary

API query for OpenAI returns list of models. Their names are filtered using a regex. The regex did not yet account for model names starting with o1- - with this change, model names starting with "o1-" are also recognized and will be shown in UI.

closes danny-avila#4339

Change Type

Please delete any irrelevant options.

• Bug fix (non-breaking change which fixes an issue)

Testing

performed docker build with this change and tested manually

Checklist

Please delete any irrelevant options.

• My code adheres to this project's style guidelines
• I have performed a self-review of my own code
• A pull request for updating the documentation has been submitted.

[weipienlee]

Tnx @juwalter, that works. However, I still get the message using an o1 model. Do you know what's going on?

Something went wrong. Here's the specific error message we encountered: An error occurred while processing your request. Please contact the Admin.

I have access to the model, checked by api call (langchain) and OpenAI playground.

[juwalter]

Hello @weipienlee - just checked again. I am now using latest container ghcr.io/danny-avila/librechat-dev:latest (just did docker exec -ti <container id> ash and confirmed that the change is in api/server/services/ModelService.js)

I have also made sure to comment out #OPENAI_MODELS=gpt-4o,... and indeed it works as intended.

If the models show in the drop-down like so:

then it might be that you do not have API access to o1-type models yet (I had it first in my chatgpt plus subscription, but not API access, and just last week it started to work via API as well).

can you try using e.g. https://github.com/simonw/llm to see if API access really works for you?

[weipienlee]

I already did almost the same:

• ghcr.io/danny-avila/librechat-dev:latest is defined in the docker compose file (rebuild after checkout bdc2fd3, 1 after your merge)
• commented the model list
• did extra container check as describe
  • found const regex = /(text-davinci-003|gpt-|o1-)/;
  • the models were already listed in the app, though

I already checked API access and is working fine (only accepts t=1):

model = "o1-mini"; temperature=1
llm = ChatOpenAI(model=model, temperature=temperature)

So, not sure what's going. It would help if there was more feedback like a log or something