GuardianAI 🛡️
Key Features
🧠 Multimodal by Design: Analyzes text (code, CVEs, reports), and images (website screenshots, logos) to identify threats.
🔐 Security-Focused: Trained on a specialized corpus of vulnerable code, security advisories, threat intelligence reports, and phishing data.
📖 Open & Transparent: All code, training data processing scripts, and model architectures are open-source to encourage trust, collaboration, and verification.
🧩 Modular Architecture: Built with a clean, component-based structure in PyTorch, making it easy for contributors to improve specific parts of the model.
🤝 Community-Driven: GuardianAI's strength comes from its community. We welcome contributions of all kinds, from data curation to model development.
Getting Started: Installation Follow these steps to set up your local development environment.
Prerequisites
Git: You must have Git installed to clone the repository. Download Git.
Python: You need Python 3.9, 3.10, or 3.11. Download Python.
Important: During installation, make sure to check the box that says "Add Python to PATH" or similar.
~~~~~~~~~~~~~~~~~🧩~~~~~~~~~~~~~~~~
Core Setup Steps
First, clone the repository to your local machine:
Generated sh
git clone https://github.com/YourUsername/GuardianAI.git
cd GuardianAI
Use code with caution.
Sh
Replace YourUsername with the actual GitHub organization or username.
Now, follow the instructions for your specific operating system. 💻 For Windows (Microsoft) We use PowerShell to set up the environment on Windows. Open PowerShell: Open a PowerShell terminal (do not use Command Prompt/CMD). Adjust Execution Policy (If Needed): By default, PowerShell may block script execution. Run the following command to allow scripts for your current session. This is a safe, temporary change.
Generated powershell
Set-ExecutionPolicy -ExecutionPolicy RemoteSigned -Scope Process
Use code with caution.
Powershell
Run the Setup Script: Generated powershell .\scripts\setup_environment.ps1 Use code with caution. Powershell This script will create a venv, activate it, and install all necessary dependencies.
🍎 For macOS (Apple) & 🐧 For Linux (including ChromeOS)
For macOS, Linux, and the Linux environment on ChromeOS, we use the
Bash setup script. Open your Terminal. Make the Script Executable: This is a one-time command to grant permission to run the script. Generated sh chmod +x scripts/setup_environment.sh Use code with caution. Sh
Run the Setup Script:
Generated sh
./scripts/setup_environment.sh
Use code with caution.
Sh
This will create and configure the Python virtual environment and install all dependencies.
📱 A Note on Mobile Platforms (Android & iOS) The GuardianAI development environment is designed for desktop operating systems (Windows, macOS, Linux). However, I am building this on a Phone. The tools required for training and developing a large AI model (PyTorch, GPU drivers, etc.) are not available on Android or iOS. However, a long-term goal for the GuardianAI project is to make the trained model accessible to mobile applications, likely through an API. This would allow a future Android or iOS app to send data (like a suspicious URL or a screenshot) to the GuardianAI model for analysis and receive a result. TL;DR: You develop GuardianAI on a desktop, but one day it could power security tools on your phone.
Usage
After running the setup script, your virtual environment (venv) will be active.
Running Tests
To ensure everything is working correctly, run our full test suite using pytest:
Generated sh
pytest
Use code with caution.
Sh
Code Style & Linting
We use black for code formatting and ruff for linting.
Generated sh
# To automatically format all code to match the project style
black .
# To check for potential errors and style issues
ruff .
Use code with caution.
Sh
Our Continuous Integration (CI) will fail if the code is not formatted or if ruff finds errors.
Project Structure
Here is a brief overview of the project's layout:
Generated code
/GuardianAI/
├── .github/ # GitHub-specific workflows (CI)
├── configs/ # Configuration files for models and training
├── data/ # Scripts and instructions for handling datasets (data is gitignored)
├── docs/ # Project documentation
├── evaluation/ # Scripts and benchmarks for evaluating model performance
├── notebooks/ # Jupyter notebooks for experimentation and analysis
├── scripts/ # Helper scripts (like environment setup)
├── src/guardian_ai/ # The main Python source code for the project
└── tests/ # The test suite for our source code
Use code with caution.
🪄 Contributing ✨ We welcome contributions of all kinds! Whether you're a data scientist, a software engineer, a security researcher, or just an enthusiastic user, you can help. Please read our CONTRIBUTING.md (coming soon!) for details on our code of conduct and the process for submitting pull requests. For major changes, please open an issue first to discuss what you would like to change. Join our community on Discord to chat with developers and get involved! License This project is licensed under the Apache 2.0 License. See the LICENSE file for details.
Here is the Basic Framework and Idea. Well make this Different then all the Others so we Don't Step on Each Others 💡 with Perplexity, Gemini, GitHub, DeepSeek. Everyone else irrelevant. We will make this within Laws, but we can deal with that and remake modifications if nesscary. Whoever Pitches in, when we start making Money in Advertising, etc, I am Righteous and well brake evenly. We will accept Donations for our Work. And Since I'm not seeing any other Major Product's here on GitHub that Public, Let's get it in Yal. What's a couple of functions? Were all on here to make money? So Let's Start 🔥🔥🤖✨Which we can do all that ourselves, We are Hackers.
/GuardianAI/
├── .github/
│ └── workflows/
│ └── python-ci.yml
│
├── .gitignore
├── LICENSE
├── README.md
├── requirements.txt # [+] Added Pillow, scikit-learn, timm
├── requirements-dev.txt
│
├── configs/
│ ├── model/
│ │ └── guardian_multimodal_7b_config.json
│ └── training/
│ └── default_trainer_args.json
│
├── data/
│ ├── README.md
│ ├── raw/
│ │ ├── cve_details/
│ │ ├── vulnerable_code/
│ │ ├── threat_reports/
│ │ └── phishing_images/ # [+] Directory for image datasets
│ │ ├── real_logos/
│ │ └── fake_logins/
│ └── processed/
│ ├── security_corpus.arrow
│ └── image_metadata.csv # [+] Metadata linking images to labels
│
├── docs/
│ ├── architecture.md # Will now describe the multimodal architecture
│ ├── data_pipeline.md
│ └── fine_tuning.md
│
├── evaluation/
│ ├── benchmarks/
│ │ ├── code_vuln_detection/
│ │ ├── secure_code_generation/
│ │ └── image_phishing_detection/ # [+] Benchmark for image tasks
│ ├── evaluate_model.py
│ └── leaderboards/
│
├── notebooks/ # [+] NEW FOLDER for experiments & baselines
│ ├── 1_data_exploration.ipynb
│ └── 2_sklearn_baseline_model.ipynb # [+] Scikit-learn baseline
│
├── scripts/
│ ├── setup_environment.sh
│ ├── setup_environment.ps1
│ ├── run_tests.sh
│ └── run_linter.sh
│
├── src/
│ └── guardian_ai/
│ ├── __init__.py
│ ├── data_processing/
│ │ ├── __init__.py
│ │ ├── code_parser.py
│ │ ├── cve_parser.py
│ │ ├── image_processor.py # [+] New module for handling images
│ │ └── text_cleaner.py
│ │
│ ├── model/
│ │ ├── __init__.py
│ │ ├── multimodal_architecture.py # Renamed from architecture.py
│ │ └── components/ # [+] NEW FOLDER for modular parts
│ │ ├── __init__.py
│ │ ├── text_encoder.py # Our original transformer
│ │ └── vision_encoder.py # The new Vision Transformer part
│ │
│ ├── inference/
│ │ ├── __init__.py
│ │ └── predictor.py # Will now handle text and/or image inputs
│ │
│ └── training/
│ ├── __init__.py
│ └── trainer.py
│
└── tests/
├── __init__.py
├── test_data_processing.py
├── test_image_processor.py # [+] New test file
└── test_model_forward_pass.py