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