Support for gpt-3.5-turbo and gpt-4

[NezihAkin]

Hi,

OpenAIAnswerGenerator class only allows to use GPT-3 models from OpenAI now. in predict function, endpoint url is defined only for gpt-3 models as /v1/completions. A small update here would enable to use chat completion models (gpt-3.5-turbo and gpt-4), since these models use /v1/chat/completions.

    if top_k is None:
        top_k = self.top_k

    # convert input to OpenAI format
    prompt, input_docs = self._build_prompt(query=query, documents=documents)

    # get answers from OpenAI API
    url = "https://api.openai.com/v1/completions"

    payload = {
        "model": self.model,
        "prompt": prompt,
        "max_tokens": self.max_tokens,
        "stop": self.stop_words,
        "n": top_k,
        "temperature": self.temperature,
        "presence_penalty": self.presence_penalty,
        "frequency_penalty": self.frequency_penalty,
    }

    headers = {"Authorization": f"Bearer {self.api_key}", "Content-Type": "application/json"}
    response = requests.request("POST", url, headers=headers, data=json.dumps(payload), timeout=timeout)
    res = json.loads(response.text)

I would create a PR here, but I am not sure about your wow.

Best regards

[benharrell]

The response format changed as well, best to check the API reference if you plan to implement this

https://platform.openai.com/docs/guides/chat/introduction

[NezihAkin]

That's totally true. Thanks for the heads up!

[AalexisS1]

Oh, thanks too)

[vblagoje]

@NezihAkin our longer-term plan is to retire OpenAIAnswerGenerator and transition users to PromptNode. We support both models in PromptNode; did you find some difficulties in using PromptNode?

[NezihAkin]

oh, sorry. Let me go in the direction of PromptNode. I focused on OpenAIAnswerGenerator directly, before checking PromptNode. Thanks a lot for the clarification!

[vblagoje]

No worries; our docs need an update as well. @dfokina let's not forget to discourage the use of https://docs.haystack.deepset.ai/docs/answer_generator#openaianswergenerator, eventually replacing that note with an official deprecation notice as we currently have for Seq2SeqGenerator and RAGenerator