Transform & Replace Selection Wtih ChatGPT does't work with "gpt-5" model.

Bugs or Unexpected Behavior
1

Device information:

  • Type of Mac: 2021 Macbook Pro M1Max
  • macOS version: Sequoia 15.6
  • BetterTouchTool version: 5.570

I changed the model name from gpt-4.1 to gpt-5, but then I got an error and it stopped working. When I checked the logs, I found the following error message.

    error =     {
        code = "missing_required_parameter";
        message = "Missing required parameter: 'messages[0].role'.";
        param = "messages[0].role";
        type = "invalid_request_error";
    };

I’m guessing the API specs might have changed, but I’d be really happy if BTT could use the new model!

Thank you!

2

I remember (or am I hallucinating?) that Andreas mentioned that exact model names should be used. So maybe test with any of these - https://platform.openai.com/docs/guides/latest-model#model-name-reference ?

3

It seems to work ok here (apart from that gpt-5 with default settings is very slow for me)
Are you using this API url? https://api.openai.com/v1/chat/completions

1 Like
4

Thank you for the information!

I checked out the links you shared and tried using some of the newer model names, like gpt-5 and gpt-5-mini, but unfortunately, I still got the same missing_required_parameter error.

Just to double-check, I even tried putting in a model name that doesn’t exist, like gpt-4.2. In that case, I got a different error saying “The model does not exist.”

So it looks like the problem isn’t with the model name itself.

5

Up until now, I’d been leaving the "URL To Call:" field blank, and it was working just fine. When I set the "URL To Call:" to the chat/completions URL, it worked as expected. And just as you mentioned, the response from GPT-5 was indeed slow.

Thanks a lot.

6

Ah interesting, I'll check why it doesn't work with blank url anymore.

Also next version will allow to set additional API parameters. The slowness is caused by the default gpt-5 settings that always put the model in "thinking" mode, which takes a long time.

However this week I'll also finally release my h@llo.ai labeled AI stuff into BTT, with that release these legacy actions will mostly become obsolete (but they will continue to be supported).

2 Likes
7

In 5.575 you can now specify extra api parameters. When entering these parameters gpt-5 should become faster:

{"reasoning_effort": "low", "verbosity": "low"}
1 Like
8

Awesome! I tried it right away. Responses are faster so it feels much smoother. Being able to adjust the reasoning_effort for each trigger is really handy. Thanks a lot.