YakGPT

A simple, locally running ChatGPT UI that makes your text generation faster and chatting even more engaging!

Features

Run locally on browser – no need to install any applications
Faster than the official UI – connect directly to the API
Easy mic integration – no more typing!
Use your own API key – ensure your data privacy and security
All state stored locally in localStorage – no analytics or external service calls
Access on https://yakgpt.vercel.app or run locally!

Screenshots

Mobile	Voice Mode	Light Theme	Dark Theme

🚀 Getting Started

Visit YakGPT to try it out without installing, or follow these steps to run it locally:

Prerequisites

You'll need the following tools installed on your computer to run YakGPT locally.

Git
Yarn (or npm or pnpm)
Any modern web browser like Google Chrome, Mozilla Firefox, or Microsoft Edge

Installation

Clone the repository:

$ git clone https://github.com/yakGPT/YakGPT.git

Install dependencies, build the bundle and run the server

$ yarn
$ yarn build
$ yarn start

Then navigate to http://localhost:3000

Congratulations! 🎉 You are now running YakGPT locally on your machine.

🔑 API Key Configuration

To utilize YakGPT, you'll need to acquire an API key for OpenAI. The app should prompt you to insert you key.

🐳 Docker

To build a Docker image, run:

docker build -t yakgpt:latest .

To run in a container:

docker run -it -p 3000:3000 yakgpt:latest

🎤 Microphone Integration

YakGPT makes chatting a breeze with its microphone integration! Activate your microphone using your browser's permissions, and YakGPT will automatically convert your speech into text.

You can also toggle the mic integration as needed by clicking on the microphone icon in the app.

Remember to use a supported web browser and ensure your microphone is functioning properly.

🛡️ Data Privacy and Security

YakGPT ensures your data privacy and security by letting you use your own API key. Your conversation with YakGPT takes place directly between your browser and OpenAI's GPT-3 API, with no intermediary servers.

📃 License

This project is licensed under the MIT License - see the LICENSE file for details

🙌 Acknowledgments

OpenAI for building such amazing models and making them cheap as chips.
opus-media-recorder A real requirement for me was to be able to walk-and-talk. OpenAI's Whisper API is unable to accept the audio generated by Safari, and so I went back to wav recording which due to lack of compression makes things incredibly slow on mobile networks. opus-media-recorder saved my butt by allowing cross-platform compressed audio recording via web worker magic. 🤗

Got feedback, questions or ideas? Feel free to submit an issue!