ChatGPT custom model

Hi,

I am using the "Use ChatGPT (Optionally on Selected Text). Copy Result to Clipboard / Variable" trigger with my own API. The custom model works only with gpt-4o-mini, which is the model provided by BTT. I tried using other models such as gpt-4o or the exact model (gpt-4o-2024-08-06), and it never works. Is there any fix for this? Is this happening to anyone else?

I tried debugging further. It is definitely an issue with the "Use ChatGPT (Optionally on Selected Text)" trigger because custom models work well with the "Transform & Replace Selection With ChatGPT" trigger. Can you please fix it for the "Use ChatGPT (Optionally on Selected Text)" trigger? Moreover, the API key vanishes after clicking save in this trigger option, whereas it shows fine for the "Transform & Replace Selection With ChatGPT" trigger.

Hi @thanos, welcome to the community!

I tried the "Use ChatGPT (Optionally on Selected Text)" action and it's also not working for me.

Using the Proxyman app, I inspected the HTTP request that BTT is making when using the "Use ChatGPT (Optionally on Selected Text)" action. I see that the response code is 200 and the response content looks correct.

This makes me think there is indeed a bug with the "Use ChatGPT (Optionally on Selected Text)" action.

My device information:

  • Device: MacBook Pro (14-inch, 2021)
  • macOS: 15.2 Beta (24C5073e)
  • BTT Version: 4.955 (2024120623)
  • Alpha: 1 uninstallable update behind latest Alpha version 4.955 (2024120624)
  • Regular: Installed version is ahead of latest Regular version 4.910.

Thank you for your reply @fortred2. Yes, there is definitely a bug because when I use the gpt-4o-mini with my API key, I can see the corresponding usage on the OpenAI activity page. There are a few other old posts, and it looks like it has not been addressed for a long time. I hope Andreas can shed some light on this!

thanks for reporting, I‘ll look into it later today

2 Likes

can you check whether 4.957 alpha (uploading) solves this for you?

1 Like

Thank you so much, @Andreas_Hegenberg. It worked!!