OpenAI Chat Completions API error: "Request failed with status code 400"

Question

I want to upgrade my GPT-3 to GPT-3.5 Turbo in Node.js. But I have a problem with that.

My code:

const askAi = async (message) => {
  try {
    const openAIInstance = await _createOpenAIInstance()

    const response = await openAIInstance.createCompletion({
      model: 'gpt-3.5-turbo',
      messages: [{ role: 'user', content: message }]
    })

    const repliedMessage = response.data.choices[0].message.content

    return repliedMessage
  } catch (err) {
    logger.error('', '', 'Ask AI Error: ' + err.message)
    return sendInternalError(err)
  }
}

But when I try to input the message:

askAi('Suggest me a job position for Auto CAD user')

It's returning an error:

Request failed with status code 400 

Answer 1

Problem

You used the wrong function to get a completion for the given model.

The function createCompletion works with the Completions API. In other words, the function createCompletion works with GPT-3 models (e.g., text-davinci-003) or GPT base models (e.g., davinci-002).

Solution

Note: OpenAI NodeJS SDK v4 was released on August 16, 2023, and is a complete rewrite of the SDK. Among other things, there are changes in method names. See the v3 to v4 migration guide.

Model	NodeJS function (SDK v3)	NodeJS function (SDK v4)
GPT-3.5 and GPT-4	openai.createChatCompletion	openai.chat.completions.create
GPT base and GPT-3	openai.createCompletion	openai.completions.create

• If you have the OpenAI NodeJS SDK v3: Change createCompletion to createChatCompletion

const askAi = async (message) => {
  try {
    const openAIInstance = await _createOpenAIInstance()

    const response = await openAIInstance.createChatCompletion({
      model: 'gpt-3.5-turbo',
      messages: [{ role: 'user', content: message }]
    })

    const repliedMessage = response.data.choices[0].message.content

    return repliedMessage
  } catch (err) {
    logger.error('', '', 'Ask AI Error: ' + err.message)
    return sendInternalError(err)
  }
}

• If you have the OpenAI NodeJS SDK v4: Change createCompletion to chat.completions.create

Note: There are also changes in extracting the message content in v4.

const askAi = async (message) => {
  try {
    const openAIInstance = await _createOpenAIInstance()

    const response = await openAIInstance.chat.completions.create({
      model: 'gpt-3.5-turbo',
      messages: [{ role: 'user', content: message }]
    })

    const repliedMessage = response.choices[0].message.content

    return repliedMessage
  } catch (err) {
    logger.error('', '', 'Ask AI Error: ' + err.message)
    return sendInternalError(err)
  }
}