KawaiiGPT

KawaiiGPT — Open-source LLM gateway accessing DeepSeek, Gemini, and Kimi-K2 through reverse-engineered Pollinations API with no API keys required, built-in prompt injection capabilities for security research, Termux/Linux native support, and Rich console interface

 /$$   /$$                                   /$$ /$$        /$$$$$$  /$$$$$$$  /$$$$$$$$
| $$  /$$/                                  |__/|__/       /$$__  $$| $$__  $$|__  $$__/
| $$ /$$/   /$$$$$$  /$$  /$$  /$$  /$$$$$$  /$$ /$$      | $$  \__/| $$  \ $$   | $$   
| $$$$$/   |____  $$| $$ | $$ | $$ |____  $$| $$| $$      | $$ /$$$$| $$$$$$$/   | $$   
| $$  $$    /$$$$$$$| $$ | $$ | $$  /$$$$$$$| $$| $$      | $$|_  $$| $$____/    | $$   
| $$\  $$  /$$__  $$| $$ | $$ | $$ /$$__  $$| $$| $$      | $$  \ $$| $$         | $$   
| $$ \  $$|  $$$$$$$|  $$$$$/$$$$/|  $$$$$$$| $$| $$      |  $$$$$$/| $$         | $$   
|__/  \__/ \_______/ \_____/\___/  \_______/|__/|__/       \______/ |__/         |__/   

KawaiiGPT is an open-source command-line AI tool that provides unified access to multiple large language models through the Pollinations reverse-engineered API wrapper. No official API keys or registration required — models are accessed freely via the Pollinations gateway at gen.pollinations.ai.

The tool supports backend LLMs including DeepSeek, Gemini, and Kimi-K2, with built-in prompt injection (jailbreak) capabilities for security research and red-team evaluation.

Note: KawaiiGPT is not a proprietary model — it is a jailbreak wrapper that proxies requests to existing LLMs through reverse-engineered API endpoints.

apt-get update && apt-get upgrade -y
apt install python3 python3-pip git -y
git clone https://github.com/MrSanZz/KawaiiGPT
cd KawaiiGPT
python3 install.py
python3 kawai.py

pkg update && pkg upgrade -y
pkg install python3 git -y
git clone https://github.com/MrSanZz/KawaiiGPT
cd KawaiiGPT
python3 install.py
python3 kawai.py

git clone https://github.com/MrSanZz/KawaiiGPT
cd KawaiiGPT
pip install -r requirements.txt
python3 main.py

Settings are stored in config.json and can be edited from the interactive menu (option [3]) or manually:

{
  "llm_provider": "pollinations",
  "api_base_url": "",
  "default_model": "deepseek"
}

Launch the application and navigate the Rich-styled console menu:

┌─────────────────────────────────────────────────────┐
│              KawaiiGPT — Main Menu                  │
├─────────────────────────────────────────────────────┤
│  [1]  Install dependencies                          │
│  [2]  Start                                         │
│  [3]  Settings                                      │
│  [4]  Description                                   │
│  [0]  Exit                                          │
└─────────────────────────────────────────────────────┘

KawaiiGPT/
├── main.py                # Entry point — Rich console menu
├── kawai.py               # Alternative entry point (install.py bootstrap)
├── install.py             # Dependency installer
├── config.json            # User configuration (auto-created)
├── requirements.txt       # Python dependencies
├── README.md              # This file
├── core/
│   ├── __init__.py
│   ├── inpainting.py      # LLM response processing
│   ├── processor.py       # Request pipeline
│   └── validator.py       # Input validation
├── detection/
│   ├── __init__.py
│   ├── detector.py        # Model detection logic
│   ├── signature.py       # Prompt signature handling
│   └── temporal.py        # Rate limiting / timing
├── gui/
│   ├── __init__.py
│   └── main_window.py     # GUI interface module
└── utils/
    ├── __init__.py
    ├── file_handler.py    # File I/O utilities
    ├── gpu_manager.py     # Resource management
    └── logger.py          # Logging configuration

This project is provided for educational and research purposes only.

• All risks and consequences of usage are the sole responsibility of the user.
• Modifying or selling this tool is prohibited.
• KawaiiGPT uses pre-existing models accessed through reverse-engineered APIs — no fine-tuned or custom models are included.
• Prompt injection (jailbreak) features are intended for authorized security research and red-team evaluation only.
• The developers are not responsible for any misuse of this tool.