OpenAIChatGPTConnecter

Hi ​@couch! That does really sound like a network-related issue! 

Which endpoint were you trying to ping? You can right-click> ‘Embed’>‘Edit Transformer’ to see the exact endpoints the connector calls, which include: https://api.openai.com/v1.

Once you confirm those endpoints, if it's still unreachable from your command prompt, your IT team would be the best next step to check for any firewall or proxy restrictions on that https://api.openai.com domain.

Another suggestion would be to check your API Key to ensure it’s tied to an active payment plan and that the model you use isn’t deprecated!

Hi, Thanks I am using the - OpenAIChatGPTConnecter and one of the parameters to enter is the endpoint. I agree it looks firewall related. I see the openapi caller needs a json file to enter the credentuals required. I can maybe look at this method but probs get same issue. I think the model used on our AI platform is GPT-5 this options diesnt seem to be available within the OpenAIChatGPTConnecter transformer - maybe this is the issue.

Ah, I see! Thank you ​@couch, for correcting me on that. In this case, I’d recommend the OpenAIConnector, if possible!

They essentially do the same thing! The OpenAIChatGPT one is the community version, whereas the OpenAIConnector is our Safe Software Verified version, which means it's maintained more frequently.

I was able to query GPT-5 on that OpenAIConnector, so it’s worth a try! 🙂

Let me know how that goes, and/or what version of the OpenAIChatGPT connector you’re on! You should be able to see that in the Navigator
 

Thanks - yes version 7 for chatgpt connector. The open API caller seems to request a json file, what the syntax of that json file required? what needs to be included? thanks

Would you be willing to share a screenshot of where it's requesting the JSON file? That will help narrow down exactly what endpoint is expecting a JSON file. I don’t seem to get a similar error when testing the ChatGPTConnector so it’s difficult to say for sure.

That said, it may be referring to the request body for the Chat Completions API, the basic structure looks like this:

{
  "model": "gpt-5-2025-08-07",
  "messages": [
    {"role": "user", "content": "Your prompt here"}
  ],
  "temperature": 1
}

You can find the full list of supported fields in the Chat Completions documentation and available models and their ID here.

As a side note, if you haven't already, it's worth trying a reinstall of the transformer:
Navigate to %APPDATA%/Safe Software\FME\FME Store\Transformers, delete the ChatGPTConnector.fmx file, then close and reopen Workbench. That can sometimes resolve unexpected Hub Transformer behavior!