🚀 Sentinel AI – Backend

Backend service for a Deepfake Detection Platform that allows users to analyze images/videos and determine whether they are AI-generated or real.

🧠 Features

🔐 User Authentication (JWT + API Key)
🔑 Per-user API Key system
🔄 API Key regeneration
📡 Deepfake detection (URL & file upload)
📜 Scan history tracking
🗄️ PostgreSQL (Supabase)
⚡ High-performance APIs with FastAPI

🏗️ Tech Stack

Framework: FastAPI
Database: PostgreSQL (Supabase)
ORM: SQLAlchemy
Auth: JWT (python-jose)
Hashing: Passlib (bcrypt)
Containerization: Docker

📁 Project Structure

.
├── app/
│   │
│   ├── detectors/
│   │   ├── image_model.py
│   │   └── video_detector.py
│   │
│   ├── models/
│   │   ├── image_detect/
│   │   │   └── xception_deepfake_base.keras
│   │   └── video-detect/
│   │       ├── model.ipynb
│   │       └── video_model.keras
│   │
│   ├── routes/
│   │   ├── __pycache__/
│   │   ├── api.py
│   │   └── auth.py
│   │
│   ├── services/
│   │   ├── __pycache__/
│   │   ├── image_scraper.py
│   │   └── ytdlp_service.py
│   │
│   ├── auth.py
│   ├── db.py
│   ├── image_model.py
│   ├── main.py
│   ├── models.py
│   ├── schemas.py
│   └── utils.py
│
├── venv/
├── .env
├── .gitignore
├── cookies.txt
├── LICENSE
└── README.md

🐳 Docker Setup (Recommended)

Run the backend without installing Python or dependencies.

✅ Prerequisites

Install Docker Desktop

🚀 Getting Started

1️⃣ Clone the repository

git clone <your-repo-url>
cd deepfake-backend

2️⃣ Create `.env` file

DATABASE_URL=postgresql://username:password@host:port/database
use ipv4 instead of ipv6 in case of supabase
GEMINI_API_KEY=your_gemini_api_key
SECRET_KEY=your_secret_key_for_jwt

⚠️ Use Supabase Session Pooler if required.

3️⃣ Run the backend

docker compose up --build

4️⃣ Access the API

Backend: 👉 http://localhost:8080
Swagger Docs: 👉 http://localhost:8080/docs

🛑 Stop the server

docker compose down

🔄 Rebuild after changes

docker compose up --build

⚠️ Notes

First build may take time (Docker image build)
Ensure port 8080 is free
.env file is required
No need to install Python locally

🔐 Authentication

The API supports two authentication methods:

1️⃣ JWT Token (Recommended)

Use the token received after login:

Authorization: Bearer YOUR_JWT_TOKEN

2️⃣ API Key

Use the API key received during registration/login:

x-api-key: YOUR_API_KEY

✅ You can use either JWT OR API Key to access protected endpoints.

🔑 Authentication Flow

Register

Creates a new user
Returns API key

Login

Returns:
- JWT token
- API key

Regenerate API Key

Generates a new API key
Invalidates the old one

📡 API Endpoints

Auth

POST /auth/register → Register user
POST /auth/login → Login & get JWT + API key
POST /auth/regenerate-key → Generate new API key

Core API

POST /analyse/url → Analyze media from URL
POST /analyse/analyse_upload → Analyze uploaded file
GET /history → Get user scan history
POST /factcheck → Fact-check a claim using Gemini API'

🧪 Example Requests

Using JWT

POST /analyse/url

Headers:
  Authorization: Bearer YOUR_JWT_TOKEN

Body:
{
  "url": "https://example.com/video.mp4"
}

Using API Key

POST /analyse/url

Headers:
  x-api-key: YOUR_API_KEY

Body:
{
  "url": "https://example.com/video.mp4"
}

⚡ Notes on Usage

Swagger UI supports authentication via the 🔒 button
JWT is recommended for frontend apps
API keys are useful for scripts and external integrations

🧠 Future Improvements

🤖 Model improvements
⏱️ Rate limiting (Redis)
🌐 Chrome extension
📊 Dashboard
🔐 Multiple API keys

💀 Notes

bcrypt limit: 72 characters
Use URL-safe DB credentials
Prefer Supabase Session Pooler for production

👨‍💻 Author

Built as part of a hackathon project — Sentinel AI

⭐ Contribute

Pull requests are welcome. Open an issue for major changes.

If you want next-level polish, I can:

add badges (Docker, FastAPI, stars, etc.)
or make it look like a top-tier GitHub repo (with screenshots + demo GIF)

FilesExpand file tree

README.md

Latest commit

History