@mefengl I'm going to temporarily remove your status check to build an initial 0.0.1 release so I can test CORS passes when using ES6 import from CDN so I can update ES6 instructions, then you can try to implement it cleaner in a branch or I will add final clean solution in main, does that sound good?

I'm working on a new method I want to call getChatID. what is it? basically a rewritten fetch function. this is what I got so far:

// Needs to be called before document loading
const originalFetch = window.fetch;

window.fetch = function (url, options) {
  return originalFetch.apply(this, arguments)
    .then((response) => {
      if (response.url.includes('://chat.openai.com/backend-api/conversations')) {
        response.json().then((data) => { console.log(data); });
      }
      return response;
    });
};

the chat id is the url identifier

why rewriting fetch instead of splitting the url? because you can't get the current chat id unless you refresh the page

why am I opening the issue? because even tho it works (the chats ids are in the items object key)

it breaks the loading of conversations and I don't know how to fix it

Cause ChatGPT web is using Tailwind, if we pre-tag elements with id, it will make other scripts' development easier.

Such as, I used to get sendButton by the last button in the form, but after installing the wide-screen mode script, that function actually returned the wrong button.

this should not be much work, cause can use the function defined in #1

After adding the starter pack to my website, I encountered an issue. When I restart the extension and refresh the browser page, the request is not being made until I navigate to the chat.open.ai page. Is there a solution to call ChatGPT in the background from other pages?

I finally got around to testing it, did you notice what users have to type:

It's just too unfriendly because they will mess up all the mixed ending symbols, could you change it to be like this instead:

chatgpt.sidebar.append('dropdown', [['label1', 'value1'], ['label2', 'value2']])

current is by function run every second

I can provide a Nepali translation for the README file since there is no option for the README to be in the Nepali language.
Should I proceed?

https://github.com/mefengl/chatgpt-utils

If you like, I will add them to this repo

I tried running the following script, and every object prefixed with: chatgpt. returns null. Greasemonkey did download the minified chatgpt js file, but it cannot work at all with the object chatgpt . chatgpt.getLastReply, chatgpt.getLastResponse all return null.

// ==UserScript==
// @name ChatGPT Userscript
// @description A Greasemonkey template to start using chatgpt.js like a boss
// @author chatgpt.js
// @namespace https://chatgpt.js.org
// @Version 2024.1.28
// @match https://chat.openai.com/*
// @ICON https://raw.githubusercontent.com/kudoai/chatgpt.js-greasemonkey-starter/main/media/images/icons/robot/icon48.png
// @require https://cdn.jsdelivr.net/npm/@kudoai/[email protected]/dist/chatgpt-2.6.4.min.js
// @grant GM_getValue
// @grant GM_setValue
// @Noframes
// @homepageURL https://github.com/kudoai/chatgpt.js-greasemonkey-starter
// ==/UserScript==

(async () => {
// Init config
console.log("Inside async ");

// Print chatgpt.js methods
await chatgpt.isLoaded(); 
console.log("Past GPT loading ");


function logHello() {
       (async () => {
  					console.log("Hello again");
       try {
          let lastResponse = await chatgpt.getLastReply();
         
          if (typeof lastResponse === 'string') {
              console.log("Hello " + lastResponse);
          } else {
              console.error("Hello Error: lastResponse is not a string");
          }   

               console.log("Hello " + lastResponse);
          } catch (error) {
              console.error("Async Hello Error:", error);
          }
      })();
 }
// Run the logger when the DOM is ready
window.addEventListener('DOMContentLoaded', logHello());

})();

No matter what I try I get a CORS policy error.

Tried using the CDN to include the script still errors. Uploaded it to perplexity and asked it to give me code that sends a message and gets a response and perplexity says there's an error in the send on line 1575 ;(

I'm not sure if adding a file upload feature is already on your product roadmap, but I believe it is a very important capability to have.

Uncaught TypeError: Cannot set properties of null (setting 'value')

As the chat GPT page has a ton of new features pertaining to GPT versions and custom GPTs, it would be very handy to have the ability to get or set chat context (amongst multiple historical chats or to a particular custom GPT) from the library.

{ Thank you so much for the time and effort you've put into making this library possible :) }

You can move the newest code to https://github.com/chatgptjs/chatgptjs-tampermonkey-starter to find out if it works.

That repo can be a substitute for a test.

@mefengl

IDK, then it will be two styles?

The style still matches:

Or do you like original one:

I would like to use this extension to save the current chat as json file.

I've noticed that when chat history is set to off, methods like getLastResponse will return messages from my last saved conversation, not the current temporary conversation.

I'm interested in getting the markdown content of messages. I know it exists locally as the copy to clipboard message at the bottom of each response will put it in your keyboard without hitting any API endpoint.

Is this something that could be supported?

Visual reference:

var chatgpt = {
    // ...
    eventEmitter: {
        events: {},

        on: function(eventName, callback) {
            if (!this.events[eventName]) {
                this.events[eventName] = [];
            }
            this.events[eventName].push(callback);
        },

        once: function(eventName, callback) {
            var self = this;
            function oneTimeCallback() {
                callback.apply(null, arguments);
                self.removeListener(eventName, oneTimeCallback);
            }
            this.on(eventName, oneTimeCallback);
        },

        removeListener: function(eventName, callback) {
            if (this.events[eventName]) {
                this.events[eventName] = this.events[eventName].filter(function(cb) {
                    return cb !== callback;
                });
            }
        },

        emit: function(eventName) {
            if (this.events[eventName]) {
                var args = Array.prototype.slice.call(arguments, 1);
                this.events[eventName].forEach(function(callback) {
                    callback.apply(null, args);
                });
            }
        }
    },
    // ...
};

emit is used when status change or other situation

user usage:

// Adding event listeners
chatgpt.eventEmitter.on('beforeSend', function() {
    console.log('Before sending message...');
});

chatgpt.eventEmitter.on('afterSend', function() {
    console.log('After sending message...');
});

chatgpt.eventEmitter.on('onGenerating', function() {
    console.log('Generating a response...');
});

chatgpt.eventEmitter.on('onDidGenerate', function() {
    console.log('Response generated.');
});

// Removing an event listener
// chatgpt.eventEmitter.removeListener('beforeSend', callback);

// Listening to an event only once
// chatgpt.eventEmitter.once('beforeSend', function() {
//     console.log('Before sending message (only once)');
// });

-- from ChatGPT and mefengl

Users want access to as much AI as they can get, they dont want to manage 50 accounts, they want the fastest AI they want the cheapest AI, and you can provide all of that for them with this update.

in addition to or in place of integrating with any aggregators - Please integrate APIpie so devs can access them all from one place/subscription and plus it also provides:

-The most affordable, reliable and fastest AI available
-One API to access ~500 Models and growing
-Language, embedding, voice, image, vision and more
-Global AI load balancing, route queries based on price or latency
-Redundancy for major models providing the greatest up time possible
-Global reporting of AI availability, pricing and performance

Its the same API format as openai, just change the domain name and your API key and enjoy a plethora of models without changing any of your code other than how you handle the models list.

This is a win win for everyone, any new AI's from any providers will be automatically integrated into your stack with this one integration. Not to mention all the other advantages.

I'm using a userscript I made to rewrite the fetch function calls and log the details, and I found an API-way to get a response. I'm working on it, and it might even be better than the DOM solution

const endpoints = {
    session: 'https://chat.openai.com/api/auth/session',
    chat: 'https://chat.openai.com/backend-api/conversations',
    singleChat: 'https://chat.openai.com/backend-api/conversation'
};

    test: function(chatIdx, responseIdx, regeneratedIdx) {
        chatIdx = chatIdx ? chatIdx : 0; // leave it as is because getChatDetails already takes care of the -1
        responseIdx = responseIdx ? responseIdx - 1 : 0;
        regeneratedIdx = regeneratedIdx ? regeneratedIdx - 1 : 0;
        return new Promise((resolve) => { chatgpt.getAccessToken().then(token => {
            getChatData(token).then(data => { resolve(data); });});});

        function getChatData(token) {
            return new Promise((resolve, reject) => {
                const xhr = new XMLHttpRequest();
                chatgpt.getChatDetails(chatIdx).then(chat => {
                    xhr.open('GET', endpoints.singleChat + '/' + chat.id, true); // GET request to the conversation
                    xhr.setRequestHeader('Content-Type', 'application/json');
                    xhr.setRequestHeader('Authorization', 'Bearer ' + token);
                    xhr.onload = () => {
                        if (xhr.status !== 200) return reject('🤖 chatgpt.js >> Request failed. Cannot retrieve chat messages.');
                        const data = JSON.parse(xhr.responseText).mapping; // all messages
                        const userMessages = [];
                        const responses = [];
                        for (const key in data) if (data[key].message && data[key].message.author.role === 'user') userMessages.push(data[key].id); // filter out user messages
                        for (const key in data)
                            // match responses with their indexed parent (user message)
                            if (data[key].message && data[key].message.author.role === 'assistant' && data[key].parent === userMessages[responseIdx]) // use responseIdx because at each user message corresponds a response so it doesn't matter if the variable says response
                                responses.push(data[key].message); // push only chatgpt responses
                        responses.sort((a, b) => a.create_time - b.create_time); // sort by creation time (in case of the response being regenerated)
                        resolve(responses[regeneratedIdx].content.parts[0]); // resolve promise giving the regenerate value requested
                    };
                    xhr.send();
                });
        });}
    },

https://github.com/kudoai/chatgpt.js/blob/main/chatgpt.js#L480
https://github.com/kudoai/chatgpt.js/blob/main/chatgpt.js#L488
https://github.com/kudoai/chatgpt.js/blob/main/chatgpt.js#L510

the above lines treat asynchronous functions as normal ones

@mefengl in your chatgpt-infinite you need to attribute this repo or chatgptjs.org and/or mention copyright me & you under MIT license (and any other scripts you use the code in)

hello i could not figure out how to call send message and get the response right away. theres a way for it ?

I think we should store an access token somewhere in the chatgpt object as a variable, and update it once it expires (the token has an expiration date, reference here) to avoid getting rate limited by the api and reducing the speed of the library

problem is idk about the security of doing this

edit: it shouldn't even matter since there's literally a function that retrieves the access token, so storing it in the object is not even a big deal, and removing the object from the window will literally make it disappear

vLLM supports an OpenAI-Compatible API. is it possible to configure this code to connect to another service other than OpenAI?
I've tried modifying the code but i am getting errors.

such add button to input

add button into the sidebar (like the auto clear history scripts)

          also (this is just an idea of mine), I think we should store an access token somewhere in the chatgpt object as a variable, and update it once it expires (the token has an expiration date, reference [here](https://github.com/kudoai/chatgpt.js/blob/main/chatgpt.js#L379)) to avoid getting rate limited by the api and reducing the speed of the library

problem is idk about the security of doing this

edit: it shouldn't even matter since there's literally a function that retrieves the access token, so storing it in the object is not even a big deal, and removing the object from the window will literally make it disappear

Originally posted by @madkarmaa in #79 (comment)

Dear chatgpt.js developer,

Greetings! I am vansinhu, a community developer and volunteer at InternLM. Your work has been immensely beneficial to me, and I believe it can be effectively utilized in InternLM as well. Welcome to add Discord https://discord.gg/gF9ezcmtM3 . I hope to get in touch with you.

Best regards,
vansinhu

chatgpt-conversation-1

chatgpt-conversation-1-prompt

chatgpt-conversation-1-response-1

chatgpt-conversation-1-response-2

'message' in data[key] 会导致后面的data[key].message.author.role报错
'message' in data[key] 替换成 data[key].message != null 后正常运行

Hi,
the extension currently fails to run with a 404 error from chatgpt.js line 30
Failed to load resource: the server responded with a status of 404 () https://raw.githubusercontent.com/KudoAI/chatgpt.js/main/data/_locales/en_GB/messages.json does not exist

By the way - thank your work - I do not understand why OpenAI insists on a narrow chat window...

It doesn't work for me

I couldn't find a delete chat func, would be great if added.

Here is playwright code to delete current chat:

        page.get_by_label("Chat history").get_by_role("button").nth(1).click()
        page.get_by_text("Delete chat").click()
        page.get_by_role("button", name="Delete").click()

The parameter btns in chatgpt.alert() doesn't take count of functions that have multiple words (eg. function hello_world()) and renders them in the buttons, taken the previous example, as "Hello_world". I think it could be a nice addition.

We may use a bundle tool like esbuild, then can split this single file into different files and then bundle them into a single file.

The main benefit is that conflict will be much less since working on a different file.

Another benefit is that it can auto-do compatibility work.

chatgpt.getLastResponse() is getting always this response: '1 / 1'

Hey yall,
Curious to get in touch with whoever is managing this project
https://juicebox.money/v2/p/588 (0xad029Eb29Dc085ed2BC83c59eaC72B5BF0493EDE)

Excited to help make it successful, but first want to confirm who owns it.

Dear devloper:

Thanks for the great app. I have to feature request below:

More width option

Here are three width now:

1. default;
2. wide screen;
3. full screen ( close sidebar )

It will be great that you can add the fourth option: between "default" and "wide screen", that will be good for some conversation.

TOC support

In some conversations, I have asked too many questions, it will be very easy to navigate to any question if there be an TOC on the right side.

Thank you.

I tried using both chatgpt.send() and creating my own logic, but none of them worked. it was working before, but the ChatGPT website got a UI update in which somehow it won't send the message automatically anymore (you have to manually modify the text first)

my own logic:

const sendButton = document.querySelector('form button[class*="bottom"]');
const chatInput = document.querySelector("form textarea");

if (chatInput) chatInput.value = "some text";
if (sendButton) {
  sendButton.removeAttribute("disabled");
  sendButton.click();
}

Looks like the script is not working with the new ChatGPT Teams subscriptions. Getting Uncaught (in promise) 🤖 chatgpt.js >> No chat with id = xxxxx-xxxx-xxx found.

@adamlui

Do you have any format config files to share?

I think this issue can be easily solved by asking ChatGPT.