Openai api Error that i m unable to overcome

jpeg said:

Show the request.
From what I've seen you have to change the model you're using in your request.

Click to expand...

GuruGazette29 said:

Same issue i face

Click to expand...

ImBackBaby said:

Once I’m home and have time I can check how I do it, but as @jpeg said it looks like you’re using the wrong model for what you’re doing?

Would definitely help if you posted some code, specifically the request

Click to expand...

Bambarbiakirgudu said:

Offcourse i can share it. but, the code is a bit longer and to mutch modules. I will try making an other snippid.

Click to expand...

bdsto said:

You're using the wrong (legacy) endpoint. I guess you switched the model to a newer one (3.5 or 4) and then the error appeared?
You need to rewrite the whole request, if you want to use a newer model. The /v1/completions Endpoint is not compatible with modern openai models. Switch to /v1/chat/completions or use the legacy endpoint but with a compatible model (gpt-3.5-turbo-instruct or davinci).
https://platform.openai.com/docs/models/model-endpoint-compatibility

Click to expand...

Surfboykj said:

your model is wrong

Click to expand...

Bambarbiakirgudu said:

I use gpt 3.5 turbo already. Can You share an example please?

Click to expand...

completion = gptclient.completions.create(
    model="gpt-3.5-turbo",
    prompt="abcd."
)

completion = gptclient.chat.completions.create(
    model="gpt-3.5-turbo",
    messages=[
      {"role": "system", "content": "You are a helpful assistant."},
      {"role": "user", "content": "abcd"}
    ]
  )

ImBackBaby said:

Hmmm I've been really thinking about this for a while now and can't figure out what the issue might be either. Below I'll post my JS implementation, but I don't see it being different.
Only 2 small things I caught, but might be nothing

1. I don't pass 2 individual properties "messages" and "model", I pass 1 JSON containing both. But since I don't really code python anymore, maybe it's just language specific?
2. I use another model, could you try it with gpt-4o ? Just out of curiosity

JavaScript:

const chat = async (newMessageText, messageHistory = []) => {
    messageHistory.push({
        role: "user",
        content: newMessageText
    });

    const chatCompletion = await openai.chat.completions.create({
        messages: messageHistory,
        model: "gpt-4o",
    });

    messageHistory.push(chatCompletion.choices[0].message);

    return ({
        text: chatCompletion.choices[0].message.content,
        history: messageHistory
    });
}

Click to expand...

Bambarbiakirgudu said:

Well buddy. Tried gpt4 too.

A Temprary solution here.

Firstly, Uninstall newer version.
pip uninstall openai

Then downgrade to the older one.
pip install openai==0.27.0


but, it's stupid why can't we find a working solution for Openai v1. You should know that openai official docs looked outdated to me also. If i m not wrong.

Just check this out: https://platform.openai.com/docs/guides/text-generation/chat-completions-api

Click to expand...

ImBackBaby said:

Yeah been staring at the documentation aswell the last 20 minutes, you literally do what they ask you. Documentation might really be outdated here

But checking openAI's Github it also looks just like your implementation. Only they first pass messages and then the model, but I suppose the way it's written the order doesn't matter in Python?
You could always open an issue with them on GitHub and see if it's an actual bug in the API

Click to expand...