Skip to content
BranchesTags

Repository files navigation

ShieldCoreAI - Complete Setup & Demo Guide

Everything needed to make sportshield-ai work from zero for the Google Solution Challenge 2026.

Live Deployment

Part 1 - Firebase Console Setup

Go to: https://console.firebase.google.com

Select your project: sheildcoreai

1. Enable Google Authentication

  • Authentication -> Sign-in method -> Google -> Enable -> Save
  • Authentication -> Settings -> Authorized domains -> Add localhost

2. Create Firestore Database

  • Firestore Database -> Create database
  • Choose Start in test mode
  • Region: asia-southeast1 (or your preferred region) This enables the Firestore API.

3. Create Storage Bucket

  • Storage -> Get Started
  • Choose Start in test mode
  • Check your bucket URL in the Firebase Console (e.g., sheildcoreai.firebasestorage.app)

4. Enable Realtime Database

  • Realtime Database -> Create database
  • Choose Test mode
  • Region: asia-southeast1
  • Expected URL format: https://sheildcoreai-default-rtdb.asia-southeast1.firebasedatabase.app

5. Download Service Account Key

  • Project Settings -> Service Accounts -> Generate new private key
  • Download the JSON and rename it to serviceAccountKey.json
  • Put it in sportshield-ai/backend/

6. Get Web App Config

  • Project Settings -> General -> Your apps
  • Copy the values from the Config view for the frontend .env.

Part 2 - Create Your Env Files

sportshield-ai/backend/.env

GEMINI_API_KEY=your_gemini_api_key_here
FIREBASE_PROJECT_ID=sheildcoreai
FIREBASE_CREDENTIALS_JSON=./serviceAccountKey.json
FIREBASE_DATABASE_URL=https://sheildcoreai-default-rtdb.asia-southeast1.firebasedatabase.app
FIREBASE_STORAGE_BUCKET=sheildcoreai.firebasestorage.app
GOOGLE_CSE_KEY=
GOOGLE_CSE_CX=
SENDGRID_API_KEY=
SENDGRID_FROM_EMAIL=
DEMO_MODE=true
PORT=8000

(Get Gemini API key from https://aistudio.google.com)

sportshield-ai/frontend/.env

VITE_API_URL=https://sheildcoreai-dp80.onrender.com/
VITE_FIREBASE_API_KEY=paste_from_firebase_console
VITE_FIREBASE_AUTH_DOMAIN=sheildcoreai.firebaseapp.com
VITE_FIREBASE_DATABASE_URL=https://sheildcoreai-default-rtdb.asia-southeast1.firebasedatabase.app
VITE_FIREBASE_PROJECT_ID=sheildcoreai
VITE_FIREBASE_STORAGE_BUCKET=sheildcoreai.firebasestorage.app
VITE_FIREBASE_MESSAGING_SENDER_ID=paste_from_firebase_console
VITE_FIREBASE_APP_ID=paste_from_firebase_console

Part 3 - Firebase Rules For Local Testing

Storage Rules

rules_version = '2';
service firebase.storage {
  match /b/{bucket}/o {
    match /{allPaths=**} {
      allow read, write: if true;
    }
  }
}

Firestore Rules

rules_version = '2';
service cloud.firestore {
  match /databases/{database}/documents {
    match /{document=**} {
      allow read, write: if true;
    }
  }
}

These rules are only appropriate for local testing and demos.

Part 4 - Critical Backend Fix (Historical)

The scan route should not upload to Storage before returning the initial response. This repo already includes that fix in: sportshield-ai/backend/routes/scan.py

Behavior now:

  • the API returns scan_id immediately
  • file upload happens inside the background scan task
  • scan processing continues even if Storage upload fails

Part 5 - Fix PyTorch On Windows

If Torch is broken on Windows, reinstall the CPU wheels:

cd sportshield-ai/backend
pip uninstall torch torchvision -y
pip install torch torchvision --index-url https://download.pytorch.org/whl/cpu

If Torch still fails, the backend will continue running and deepfake detection will stay disabled until the dependency issue is resolved.

Part 6 - Run The Project & Demo Preparation

Terminal 1 - Backend

cd sportshield-ai/backend
pip install -r requirements.txt
python main.py

(Note: Windows users might see Uvicorn hot-reload issues. reload=False has been set in main.py to prevent crashes. Stop and start the server manually after file changes.)

Expected backend log highlights:

  • Firebase Admin Initialized Successfully
  • either deepfake model loaded, or deepfake skipped with a warning

Terminal 2 - Seed Demo Data

To fully prepare the Judge Demo Gallery, you must seed the database and cache the demo files:

cd sportshield-ai/backend
# 1. Download offline images for the demo gallery
python download_demo_dataset.py

# 2. Seed database with authentic organizations, global violations, and demo scans
python seed_demo_data.py

Terminal 3 - Frontend

cd sportshield-ai/frontend
npm install
npm run dev

Open your browser to: http://localhost:5173

Part 7 - The 5-Minute Judge Demo Flow

We highly recommend using the newly created Demo Gallery (/demo) for the live presentation to ensure a seamless, error-free demonstration.

  1. Dashboard → Show active tracking, 247 global scans, and $5,400 protected rights value.
  2. Demo Gallery (/demo) → Click SCAN THIS on the Authentic Olympic Sprint to show a 7-step verified pipeline (watermark + registry match).
  3. Demo Gallery (/demo) → Click SCAN THIS on the Synthetic Deepfake to show 94% threat detection and EXIF stripping analysis.
  4. Demo Gallery (/demo) → Click SCAN THIS on the Dual Threat to show deepfake + IP theft detection simultaneously.
  5. Violations Feed → Show global threat map, click on any violation, and generate an automated legal DMCA takedown notice via Gemini.

Part 8 - Deepfake Model Integration

The project already supports a Hugging Face model in: sportshield-ai/backend/models/deepfake.py

Current default:

MODEL_ID = 'dima806/deepfake_vs_real_image_detection'

Option A - Use Your Hugging Face Model

Change:

MODEL_ID = 'your-username/your-model-name'

Option B - Use Your Local PyTorch .pt Or .pth Model

Replace models/deepfake.py with a local-model implementation and point MODEL_PATH to your file in backend/models/.

Option C - Use Your Keras .h5 Model

Load the model with TensorFlow in models/deepfake.py and adapt preprocessing/inference to your trained architecture.

Part 9 - Gemini Integration

Gemini is already wired in: sportshield-ai/backend/services/gemini.py and sportshield-ai/backend/services/dmca.py utilizing the gemini-1.5-flash model.

Once GEMINI_API_KEY is set, these features become available:

  • image manipulation analysis
  • legal report generation
  • multilingual report output

Part 10 - Final Checklist

  • sportshield-ai/backend/.env exists
  • sportshield-ai/frontend/.env exists
  • sportshield-ai/backend/serviceAccountKey.json exists
  • Firestore is enabled
  • Storage is enabled
  • Realtime Database is enabled
  • Google Auth is enabled
  • localhost is in Firebase authorized domains
  • Firestore rules allow local testing
  • Storage rules allow local testing
  • python main.py starts successfully
  • python seed_demo_data.py succeeds
  • python download_demo_dataset.py succeeds
  • npm run dev starts successfully
  • Google sign-in works, or local preview is used until Firebase Auth is configured

Quick Reference

Key Where to get it What breaks without it
GEMINI_API_KEY Google AI Studio Gemini analysis and legal reports
FIREBASE_PROJECT_ID Firebase Project Settings Backend Firebase integrations
FIREBASE_CREDENTIALS_JSON Service account JSON path Backend Firebase Admin init
VITE_FIREBASE_API_KEY Firebase web app config Frontend Firebase access
VITE_FIREBASE_AUTH_DOMAIN Firebase web app config Google sign-in
VITE_FIREBASE_DATABASE_URL Realtime Database Live feed and dashboard stats
VITE_FIREBASE_STORAGE_BUCKET Firebase web app config Frontend storage references

Generated for ShieldCoreAI / sportshield-ai. Best of luck at the Solution Challenge!

About

No description, website, or topics provided.

Resources

Readme
Activity

Stars

0 stars

Watchers

0 watching

Forks

0 forks
Report repository

Releases

Packages

Contributors

Languages