Content Shield

Content Shield is a web app that helps creators check a video for social platform policy risk before posting it.

The Problem It Solves

If you post a video without reviewing it carefully, you can run into:

• platform takedowns
• limited reach or demonetization
• missed policy issues in the audio, visuals, or on-screen text
• uncertainty about whether a video is risky enough to rewrite or re-edit

Content Shield is built to make that first review faster. You upload a video, choose the platform, and get a simple risk assessment with flagged issues and a short explanation.

What The App Does

• lets you choose a target platform: YouTube, TikTok, Instagram, or X
• uploads a video to the backend for analysis
• uses Gemini to review the video
• pulls platform policy context from local policy documents using retrieval
• returns a Low, Medium, or High risk level
• returns a short summary of why that risk level was assigned
• returns flagged issues with timestamps, snippets, and rationale

How It Works

1. The React frontend lets the user pick a platform and upload a video.
2. The FastAPI backend accepts the file at POST /analyze.
3. The backend retrieves policy context from documents in backend/policy_docs.
4. Gemini analyzes the uploaded video with that policy context.
5. The app returns a structured result for the frontend to display.

Stack

• frontend: React, Vite, TypeScript, Tailwind
• backend: FastAPI
• AI: Google Gemini
• retrieval: LlamaIndex + ChromaDB
• caching and rate limiting: Redis

Run With Docker

Docker is the easiest way to run the full project because it starts the frontend, backend, and Redis together.

1. Create the backend env file

Copy backend/env.example to backend/.env and add your Gemini key:

GEMINI_API_KEY=your_api_key_here

2. Start the app

docker compose up --build

3. Open the app

• frontend: http://localhost:8080
• backend docs: http://localhost:8000/docs

Run Without Docker

Use this if you want to run the frontend and backend yourself.

Requirements

• Python 3.10+
• Node.js 20+
• Redis
• FFmpeg

1. Backend setup

Make sure Redis is running locally, then create backend/.env and add your Gemini key. Set REDIS_URL to your local Redis instance.

Example:

GEMINI_API_KEY=your_api_key_here
REDIS_URL=redis://localhost:6379/0

Install backend dependencies:

cd backend
python -m venv venv

Activate the virtual environment, then install packages:

pip install -r requirements.txt

Start the backend from the backend folder:

uvicorn app.main:app --reload --port 8000

2. Frontend setup

From the project root:

npm ci
npm run dev -- --port 8080

Open http://localhost:8080.

Project Notes

• policy documents are stored in backend/policy_docs
• the local vector database is stored in backend/chroma_db
• Redis is used for caching analysis results and rate limiting
• the frontend defaults to http://localhost:8000 for the API if VITE_API_URL is not set