chatgpt-api: Starting to get 404 when I send a message

Verify latest release

I verified that the issue exists in the latest chatgpt release

Verify webapp is working

I verify that the ChatGPT webapp is working properly for this account.

Environment details

Node.js v18.13.0 Windows 10 Chrome Version 109.0.5414.120 (Official Build) (64-bit)

Describe the Bug

I’ve been using the module with great success, but from today I started running into issues: When I send a message to ChatGPT I get an error code 404:

ChatGPTError: ChatGPTAPI error 404 at ChatGPTAPIBrowser.sendMessage (…node_modules/chatgpt/build/index.js:1604:23) at process.processTicksAndRejections (node:internal/process/task_queues:95:5) at async …chatgpt.js:49:21 { statusCode: 404, statusText: ‘’

Logging in still works for me. It looks like an update from ChatGPT that needs to be incorporated into the module.

About this issue

Original URL
State: closed
Created a year ago
Comments: 16 (3 by maintainers)

Most upvoted comments

Update: the text-chat-davinci-002-20230126 model has been disabled by OpenAI, which is what this library’s currently using under the hood.

OpenAI released this model in stealth to YC and select partners last week. It was fine-tuned for chat and used to some extent by ChatGPT itself. After @waylaidwanderer discovered it, we updated this lib to use it.

This is a temporary setback as we look for a workaround. It is very likely that OpenAI partners still have access to a similar model. If you have any info on workarounds, please share them with one of the admins or in #general.

As @qbosen mentioned for a temporary workaround, you can override the model to use text-davinci-003 to keep your applications running. This model isn’t fine-tuned for chat like ChatGPT, but it will still produce reasonable results.

  const api = new ChatGPTAPI({
    apiKey: process.env.OPENAI_API_KEY,
    completionParams: {
      model: 'text-davinci-003'
    }
  })

NOTE: this method is not free and your OpenAI account will be charged for usage. Double check their pricing to make sure this is okay for your use case.

transitive-bullshit on Feb 2, 2023

A temporary solution is to set the model

const api = new ChatGPTAPI({completionParams: {model: 'text-davinci-003'}});

qbosen on Feb 2, 2023