This is a demonstration of my proposal that keyboards could be significantly improved through the integration of AI. Let's refer to it as Keyboard 2.0, which offers AI assistant features.
First and foremost, this project consists of a simple iOS keyboard extension (featuring two layouts by default: English and Turkish) located in the Relax directory. Additionally, there's a straightforward API that utilizes ChatGPT to correct grammatical errors in real-time and explore other experimental functionalities, found in the keyboard_api directory.
Run the API:
cd keyboard_api/
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
uvicorn main:app --host 0.0.0.0 --port 8000
NOTE: You should to rename .env_example to .env in the keyboard_api directory and adjust the API_KEY to match your own OpenAI API Key. When deploying the API on your local network, ensure that you accurately specify the correct address (192.168.1.xxx) as the endpoint in APIService.swift.
IMPORTANT NOTE: After installing the app on your device or simulator, please navigate to Settings -> General -> Keyboard -> Keyboards to add the keyboard. Make sure to enable Allow Full Access for it.
Once the API is running, open the Relax project in Xcode and deploy it on your simulator or device to begin testing.
I believe it would be feasible to incorporate features such as text-to-image, where the response could be inserted into the Pasteboard for sending as a message on iOS. Additionally, there's potential to develop a new model specifically designed to rectify grammatical errors based on the source language of the given text.
Please feel free to share your thoughts and suggest new ideas by opening issues. Your feedback is highly appreciated.