✨ VibeGuard: An AI-Powered Content Moderation Engine

---

VibeGuard is an automated Trust & Safety NLP pipeline, designed to instantly analyze text and categorize harmful behavior across multiple risk vectors in real time.

---

🚦 The Problem

Modern digital platforms scale rapidly, but traditional manual content moderation is:

• Slow
• Expensive
• Emotionally taxing for human moderators

As user-generated content grows, automated, real-time solutions are critical to maintain trust and safety without sacrificing user experience.

---

🛠️ The Technical Solution

VibeGuard leverages state-of-the-art NLP to automatically classify content for risk, with:

• Base Model: Fine-tuned distilbert-base-uncased (fast, lightweight inference)
• Dataset: Jigsaw Toxic Comment Classification dataset
• Performance: 86.67% Accuracy on the validation set

---

🌟 Key Features

• Multi-Label Detection: Evaluates text for multiple risk categories at once (not just binary toxic/not).
• Targeted Categories: Detects the following risk vectors:
  • Toxic
  • Severe Toxic
  • Obscene
  • Threat
  • Insult
  • Identity Hate
• User-Friendly Dashboard: Minimalist, dark-mode Streamlit web UI for rapid inference and testing.

---

💻 Tech Stack

• Deep Learning: PyTorch, Hugging Face Transformers, Datasets
• Frontend: Streamlit
• Deployment: Hugging Face Spaces

---

📁 Repository Structure

notebooks/
  └── VibeGuard_Training.ipynb   # Data prep, fine-tuning, and evaluation notebook
src/
  └── app.py                     # Streamlit dashboard frontend
requirements.txt                 # Dependency list
README.md                        # Project documentation (this file)

---

🚀 Try the Live Demo

Test VibeGuard instantly on Hugging Face Spaces.

---

🖥️ How to Run Locally

1. Clone the repository

   git clone https://github.com/Ilyan321/VibeGuard-Toxicity-Detector.git
   cd VibeGuard-Toxicity-Detector

2. (Recommended) Create a virtual environment

   python -m venv venv
   source venv/bin/activate  # on Unix or Mac
   .\venv\Scripts\activate   # on Windows

3. Install dependencies

   pip install -r requirements.txt

4. Run the app

   streamlit run src/app.py

   The app will open in your browser at http://localhost:8501.

---

📖 License

MIT © Ilyan321
See the LICENSE file for more details.

---

🤝 Contributing

Contributions, bug reports, and suggestions are welcome! Please open issues or submit pull requests.

---

📬 Contact

Questions? Reach out via GitHub Issues.