Chat GPT Instruct Model - Possible Bug!

[GunnerJnr]

Hi,

I saw that in the models for the Chat GPT piece, the new gpt-3.5-turbo-instruct was in there, so I was messing around to see how it faired. However, on running the sample data test, I get the following error each time:

{
  "token": 404,
  "error": {
    "message": "This is not a chat model and thus not supported in the v1/chat/completions endpoint. Did you mean to use v1/completions?",
    "type": "invalid_request_error",
    "param": "model",
    "code": null
  }
}

I assume this is a bug or not set up correctly in ActivePieces. Looks like it is calling the wrong API endpoint.

Kind regards

[ahwork20230915]

Was going to start a topic about this. Hope this can be solved asap.

[GunnerJnr]

Thank you @ahwork20230915

[Abdul]

Hello GunnerJnr !

I opened an issue on Github for this, we will check it from here and notify once it is solved

[BUG]: ChatGPT 3.5 turbo-instruct

[GunnerJnr]

Thank you. That is greatly appreciated.

[ashrafsam]

Hey all, this is not an Activepieces bug. According to OpenAI’s docs models page, this model is only compatible with the legacy text completion endpoint and not the chat endpoint.

I moved this topic to Discussions to discuss whether we’ll add a new action for completions or wait for OpenAI to make it compatible with Chat.

[ahwork20230915]

I do hope we take initiative instead of waiting, granted if the task is not time consuming

[ashrafsam]

Hi @GunnerJnr and @ahwork20230915 It’s possible to use gpt-3.5-turbo-instruct within the OpenRouter piece instead of OpenAI. Here is a screenshot:

[GunnerJnr]

@ashrafsam,

Awesome, thanks man.