🛡️ PromptGuard

PromptGuard is a full-stack AI safety system designed to analyze user prompts and prevent unsafe or unethical requests before they reach generative AI models.

It combines rule-based filtering, machine learning risk scoring, and explainable decisions in a clean, production-style architecture.

---

🚀 Why PromptGuard?

Recent incidents involving generative AI misuse highlight the need for pre-generation safety layers.

PromptGuard focuses on:

• Preventing harmful prompts (explicit, abusive, exploitative)
• Providing transparent explanations for decisions
• Suggesting safe rewrites instead of hard blocking when possible

---

🧠 How It Works

1. User enters a prompt
2. Rule-based checks detect explicit violations
3. ML model computes a risk score
4. Decision engine classifies the prompt as:
   • ✅ SAFE
   • ✏️ REWRITE
   • ❌ BLOCK
5. User receives:
   • Decision
   • Risk score
   • Explanation
   • Suggested rewrite (if applicable)

---

🏗️ Tech Stack

Frontend

• React (Vite)
• Tailwind CSS v4

Backend

• FastAPI
• Hugging Face Transformers
• Rule-based + ML hybrid safety logic

---

📸 Screenshots

🧠 Prompt Analysis Interface

---

❌ Blocked Prompt Example

---

✏️ Rewrite Suggestion Example

---

✅ Safe Prompt Example

---

⚙️ Run Locally

Backend

cd backend
python -m venv .venv
.venv\Scripts\activate
pip install -r requirements.txt
uvicorn app:app --reload

### Frontend

```bash
cd frontend/promptguard-ui
npm install
npm run dev