OpenAI Chat Completions API error: "InvalidRequestError: Unrecognized request argument supplied: messages"

Question

I am currently trying to use OpenAI's most recent model: gpt-3.5-turbo. I am following a very basic tutorial.

I am working from a Google Collab notebook. I have to make a request for each prompt in a list of prompts, which for sake of simplicity looks like this:

prompts = ['What are your functionalities?', 'what is the best name for an ice-cream shop?', 'who won the premier league last year?']

I defined a function to do so:

import openai

# Load your API key from an environment variable or secret management service
openai.api_key = 'my_API'

def get_response(prompts: list, model = "gpt-3.5-turbo"):
  responses = []

  
  restart_sequence = "\n"

  for item in prompts:

      response = openai.Completion.create(
      model=model,
      messages=[{"role": "user", "content": prompt}],
      temperature=0,
      max_tokens=20,
      top_p=1,
      frequency_penalty=0,
      presence_penalty=0
    )

      responses.append(response['choices'][0]['message']['content'])

  return responses

However, when I call responses = get_response(prompts=prompts[0:3]) I get the following error:

InvalidRequestError: Unrecognized request argument supplied: messages

Any suggestions?

Replacing the messages argument with prompt leads to the following error:

InvalidRequestError: [{'role': 'user', 'content': 'What are your functionalities?'}] is valid under each of {'type': 'array', 'minItems': 1, 'items': {'oneOf': [{'type': 'integer'}, {'type': 'object', 'properties': {'buffer': {'type': 'string', 'description': 'A serialized numpy buffer'}, 'shape': {'type': 'array', 'items': {'type': 'integer'}, 'description': 'Array shape'}, 'dtype': {'type': 'string', 'description': 'Stringified dtype'}, 'token': {'type': 'string'}}}]}, 'example': '[1, 1313, 451, {"buffer": "abcdefgh", "shape": [1024], "dtype": "float16"}]'}, {'type': 'array', 'minItems': 1, 'maxItems': 2048, 'items': {'oneOf': [{'type': 'string'}, {'type': 'object', 'properties': {'buffer': {'type': 'string', 'description': 'A serialized numpy buffer'}, 'shape': {'type': 'array', 'items': {'type': 'integer'}, 'description': 'Array shape'}, 'dtype': {'type': 'string', 'description': 'Stringified dtype'}, 'token': {'type': 'string'}}}], 'default': '', 'example': 'This is a test.', 'nullable': False}} - 'prompt'

Woolfell

messages isn't the correct argument. Guess you need prompt: [] – Bib

@Bib the messages argument is the one provided in the documentation. However, implementing your solution leads to another error message (check the most recent edit) – Woolfell

But the prompt just need to be your question: prompt: item – Bib

@Bib This leads to a different error which I believe has to do with the model (your solution would work, e.g., with a davinci model. InvalidRequestError: This is a chat model and not supported in the v1/completions endpoint. Did you mean to use v1/chat/completions? – Woolfell

Oke I made some code myself and can't reproduce your problem. Works fine over here. – Bib

Sure you are using the latest version of the openai package? – Bib

error I get is openai.NotFoundError: Error code: 404 - {'error': {'message': 'This is a chat model and not supported in the v1/completions endpoint. Did you mean to use v1/chat/completions?', 'type': 'invalid_request_error', 'param': 'model', 'code': None}} – Kittle

Answer 1

Problem

You used the wrong method to get a completion. When using the OpenAI SDK, whether you use Python or Node.js, you need to use the right method.

Which method is the right one? It depends on the OpenAI model you want to use.

Solution

The tables below will help you figure out which method is the right one for a given OpenAI model.

STEP 1: Find in the table below which API endpoint is compatible with the OpenAI model you want to use.

API endpoint	Model group	Model name
/v1/chat/completions	• GPT-4 • GPT-3.5	• gpt-4 and dated model releases • gpt-4-32k and dated model releases • gpt-4-1106-preview • gpt-4-vision-preview • gpt-3.5-turbo and dated model releases • gpt-3.5-turbo-16k and dated model releases • fine-tuned versions of gpt-3.5-turbo
/v1/completions (Legacy)	• GPT-3.5 • GPT base	• gpt-3.5-turbo-instruct • babbage-002 • davinci-002
/v1/assistants		All models except gpt-3.5-turbo-0301 supported. Retrieval tool requires gpt-4-1106-preview or gpt-3.5-turbo-1106.
/v1/audio/transcriptions	Whisper	• whisper-1
/v1/audio/translations	Whisper	• whisper-1
/v1/audio/speech	TTS	• tts-1 • tts-1-hd
/v1/fine_tuning/jobs	• GPT-3.5 • GPT base	• gpt-3.5-turbo • babbage-002 • davinci-002
/v1/embeddings	Embeddings	• text-embedding-ada-002
/v1/moderations	Moderations	• text-moderation-stable • text-moderation-latest

STEP 2: Find in the table below which method you need to use for the API endpoint you selected in the table above.

Note: Pay attention, because you have to use the method that is compatible with your OpenAI SDK version.

API endpoint	Method for the Python SDK v0.28.1	Method for the Python SDK >=v1.0.0	Method for the Node.js SDK v3.3.0	Method for the Node.js SDK >=v4.0.0
/v1/chat/completions	openai.ChatCompletion.create	openai.chat.completions.create	openai.createChatCompletion	openai.chat.completions.create
/v1/completions (Legacy)	openai.Completion.create	openai.completions.create	openai.createCompletion	openai.completions.create
/v1/assistants	/	openai.beta.assistants.create	/	openai.beta.assistants.create
/v1/audio/transcriptions	openai.Audio.transcribe	openai.audio.transcriptions.create	openai.createTranscription	openai.audio.transcriptions.create
/v1/audio/translations	openai.Audio.translate	openai.audio.translations.create	openai.createTranslation	openai.audio.translations.create
/v1/audio/speech	/	openai.audio.speech.create	/	openai.audio.speech.create
/v1/fine_tuning/jobs	/	openai.fine_tuning.jobs.create	/	openai.fineTuning.jobs.create
/v1/embeddings	openai.Embedding.create	openai.embeddings.create	openai.createEmbedding	openai.embeddings.create
/v1/moderations	openai.Moderation.create	openai.moderations.create	openai.createModeration	openai.moderations.create

Python SDK v1.0.0 working example for the gpt-3.5-turbo model

If you run test.py, the OpenAI API will return the following completion:

Hello! How can I assist you today?

test.py

import os
from openai import OpenAI

client = OpenAI(
    api_key = os.getenv("OPENAI_API_KEY"),
)

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

print(completion.choices[0].message.content.strip())

Node.js SDK v4.0.0 working example for the gpt-3.5-turbo model

If you run test.js, the OpenAI API will return the following completion:

Hello! How can I assist you today?

test.js

const OpenAI = require("openai");

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY,
});

async function main() {
  const completion = await openai.chat.completions.create({
    model: "gpt-3.5-turbo",
    messages: [
      { role: "system", content: "You are a helpful assistant." },
      { role: "user", content: "Hello!" },
    ],
  });

  console.log(completion.choices[0].message.content.trim());
}

main();

Cheri

This: I think the naming practices at OpenAI made it a bit confusing, Why would you have this in the introduction example: import openai openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Who won the world series in 2020?"}, {"role": "assistant", "content": "The Los Angeles Dodgers won the World Series in 2020."}, {"role": "user", "content": "Where was it played?"} ] ) – Woolfell

I agree, it's a bit confusing. I think they should copy-paste the example from the documentation. – Cheri

This is Pytthon, right? What's the nodejs equivalent? I had const completion = await openai.createCompletion({....}) but I get the error "This is a chat model and not supported in the v1/completions endpoint. Did you mean to use v1/chat/completions?" – Igenia

Super helpful~! In my case, I used a wrong class, so instead of openai.ChatCompletion.create(), I used the wrong one openai.Completion.create(). And OpenAI happens to have both ChatCompletion and Completion which makes this issue harder to be found. – Westcott

Answer 2

response = openai.ChatCompletion.create(
model='gpt-3.5-turbo',
  messages=[
    {"role": "user", "content": "What is openAI?"}],
max_tokens=193,
temperature=0,
)

print(response)
print(response["choices"][0]["message"]["content"])

Alcine

This answer could benefit from literally any explanation at all. Code-only answers are rarely useful, especially as time goes on. – Ultrasonics

Answer 3

For gpt4 this is what worked for me

        for prompt in prompts:
            # from openai import OpenAI
            # client = OpenAI(api_key=openai.api_key)
            print(f'{self.model_name=}')
            response = openai.chat.completions.create(
            # response = client.chat.completions.create(
                model = self.model_name,
                messages = 
                [
                    # {"role": "system", "content": "You are a helpful assistant."},
                    {"role": "user", "content": prompt},
                ],
                temperature = temperature 
            )
            generated_text: str = response.choices[0].message.content.strip()
            generated_texts.append(generated_text)

when error

openai.NotFoundError: Error code: 404 - {'error': {'message': 'This is a chat model and not supported in the v1/completions endpoint. Did you mean to use v1/chat/completions?', 'type': 'invalid_request_error', 'param': 'model', 'code': None}}

appeared.

Kittle

Answer 4

You should define messages=[{"role": "user", "content": prompt}] outside of your response variable and call that in response variable like:

messages=[{"role": "user", "content": prompt}]
for item in prompts:
      response = openai.Completion.create(
      model=model,
      messages=messages,
      temperature=0,
      max_tokens=20,
      top_p=1,
      frequency_penalty=0,
      presence_penalty=0
    )

Amalee