loeffeldude/my-chat-gpt

MyChatGPT is a ChatGPT UI clone with some extra features.

MyChatGPT

This is a OSS standalone ChatGPT client. It is based on ChatGPT.

The client works almost just like the original ChatGPT websites but it includes some additional features.

Hosted Version

You can use the hosted version of this client here.

Installation

Go to the realease page and download and install the latest release for your platform.

Setup

1. Head over to the settings and enter your OpenAI API key. You can get one here.

2. Choose a preamble. A basic default preamble is already set. You can change it to something more specific to your use case.

3. Start Chatting!

---

Why?

I wanted to use ChatGPT but I didn't want to pay a fixed price if I have days where I barely use it. So I created this client that almost works like the original.

The 20 dollar price tag on ChatGPT is a bit steep for me. I don't want to pay for a service I don't use. I also don't want to pay for a service that I use only a few times a month. Even with relatively high usage this client is much cheaper.

Pricing Comparison:

A ChatGPT conversation can hold 4096 tokens (about 1000 words). The ChatGPT API charges 0.002$ per 1k tokens.

Every message needs the entire conversation context. So if you have a long conversation with ChatGPT you pay about 0.008$ per message. ChatGPT needs to send 2500 (messages with full conversation context) a month to pay the same as the ChatGPT subscription.

You can delete previous messages with this client if they are no longer needed for the context. So you can have a lot more messages for the same price.

Features

Pay as you go.

You use your own API key and pay for the usage this turns out to be much cheaper than the original ChatGPT website with moderate usage.

More customization.

You can change the models behaior by changing the preamble. You can use this to create more customized chatbots.

Edit the chat history.

You can edit the chat history and the model will continue the conversation from the edited point.

Mark messages as important.

You can mark messages as important. Important messages will never be dropped from the conversation. (The model still has a token limit all not important messages will be dropped if the conversation gets too long.)

No annoying login flow.

You can use the client without having to login to OpenAI.

---

Development

This is just a React app with an Electron wrapper.

Building is done with Vite.

Fork this repo and start hacking. Feel free to open a PR if you want to contribute.

Setup

npm install

Run

npm run dev

Build

Build the electron app:

npm run build:electron

Just build React app:

npm run build:client

my-chat-gpt's Issues

Integrate APIpie.ai

Founder here plugging my own product.

I tried to test our API with your app but its locked down to OpenAI, APIpie.ai would be a great integration we have a huge range of models from different providers services and have also marked all Chat enhanced models (proper streaming and context memory) with subtype chatx so the models list can be filtered to only provide a list of models optimised for Chat experience, we are openai API compliant and we also update the models list frequently so would require no manual updating for new models if filtering on ChatX, you could further clean the list by removing duplicates if you don't mind what provider is serving the model.

https://apipie.ai/
https://discord.gg/fnVjanuCzB

Add a transcribe Audio feature

Add a transcribe audio feature to use the WhisperAPI to allow for input via voice.

Few suggestions

Hello, Author
I came across your app and I must say, it looks great! I have a few suggestions that might help improve it even more:

• Consider adding more topics or tags to your repository to make it easier for users to discover your app.
• I noticed that the ChatInput UX could use some improvement. Specifically, you could remove the textarea focus outline and set the default behavior to send messages on "Enter" and create a line-break on "Ctrl+Enter". You might also want to consider following the logic used on the official ChatGPT website.
• It would be helpful to include a Transcribe Audio button, which would allow users to select audio or mp3 files and transcribe them to text using the Whisper API.

FYI I also created a ChatGPT clone (https://github.com/ioanmo226/chatgpt-web-application). However, after exploring your app, I believe that it has many benefits compared to my clone.

Appimage incompatible with fuse3

$ dpkg -l | grep fuse3
ii  fuse3                                      3.10.5-1build1                          amd64        Filesystem in Userspace (3.x version)
ii  libfuse3-3:amd64                           3.10.5-1build1                          amd64        Filesystem in Userspace (library) (3.x version)

$ ./MyChatGPT-0.2.1.AppImage
dlopen(): error loading libfuse.so.2

AppImages require FUSE to run. 
You might still be able to extract the contents of this AppImage 
if you run it with the --appimage-extract option. 
See https://github.com/AppImage/AppImageKit/wiki/FUSE 
for more information

Move the toasts to be less annoying

Right now toasts get in the way of pressing buttons. I want to move them to the top right of the screen.

Automatic Chat Title

Implement an automatic chat title like the original ChatGPT

Invalid Characters

Hi, I found a little error when "delete all chats", the window that opens with invalid characters.

I want to congratulate you on the project. I'm a Linux user (Ubuntu Mate 23.04) and it's been great using ChatGPT through the program.

Regenerate Response button

Add a button to regenerate the response.

Fix the broken Chatinput

It's impossible to add new lines without shift+send setting. Also it should automatically add lines if the height overflows.

Make the chat messages window wider

Right now the messages look pretty small on a large screen.

Default start screen for easier API Key configuration

Right now we have to go to the settings to set the Initial setup if the Key is not setup show a ChatGPT like start screen.

Model selection?

Hi there, first of all let me say, this just works perfect. I was wondering what model is this using by default. GPT 3.5 Turbo, GPT-4?
Would be great to have a option to select the model we want to use. Thank you!