Refinery Eye AI - Local Setup

Overview

This application runs completely locally with no cloud dependencies except for the Gemini AI API for video analysis.

Architecture

• Frontend: React + Vite (served from /dist)
• Backend: Express.js server
• File Storage: Local file system (/uploads directory)
• Database: Local JSON file (/data/reports.json)
• AI Analysis: Google Gemini API (requires API key)

Prerequisites

• Node.js 18 or higher
• Google Gemini API key

Setup

1. Install dependencies:

   npm install

2. Configure environment:

   • Copy .env.example to .env
   • Add your Gemini API key to .env:

     GEMINI_API_KEY=your_api_key_here
     

3. Build the frontend:

   npm run build

4. Start the server:

   npm start

5. Access the application:

   • Open your browser to http://localhost:8080
   • Login with:
     • Username: JRinst
     • Password: JRiocl@2025

Directory Structure

├── uploads/          # Uploaded videos and PDFs (local storage)
├── data/            # Local database
│   └── reports.json # Saved analysis reports
├── dist/            # Built frontend files
├── server.js        # Express server
├── App.tsx          # Main React component
└── .env             # Environment variables

Features

• ✅ Local file upload (videos and PDFs)
• ✅ AI-powered video analysis using Gemini
• ✅ Local report storage (JSON file)
• ✅ Session-based authentication
• ✅ PDF report generation

How It Works

Upload Flow

1. User uploads video → Saved to /uploads directory
2. User uploads reference PDFs → Saved to /uploads directory
3. Files served via /api/content/* endpoint

Analysis Flow

1. Video file uploaded to Gemini API for processing
2. Reference PDFs sent as inline base64 data
3. Gemini analyzes video against references
4. Results returned and saved to local JSON database
5. Gemini file cleaned up after analysis

Report Storage

• Reports saved to data/reports.json
• Each report has a unique UUID
• Reports include video URL, references, analysis results, and metadata

No Cloud Dependencies

• ❌ No Google Cloud Storage
• ❌ No Firestore
• ❌ No Vercel
• ✅ Only Gemini API for AI analysis (temporary file upload during processing)

Development

Run in development mode:

npm run dev        # Frontend dev server
npm run dev:api    # API server (if separate)

Build for production:

npm run build
npm start

Troubleshooting

"Analysis failed" error

• Check that GEMINI_API_KEY is set in .env
• Verify the API key is valid
• Check server logs for detailed error messages

"Failed to upload video"

• Ensure /uploads directory exists (created automatically)
• Check file permissions
• Verify disk space

Port already in use

# Find process using port 8080
netstat -ano | findstr :8080

# Kill the process (Windows)
taskkill /PID <PID> /F

Security Notes

• Change default password in production
• Set a secure SESSION_SECRET in .env
• Keep .env file secure (never commit to git)
• Uploaded files are stored locally - ensure proper file system permissions

Limitations

• Session storage uses MemoryStore (not suitable for production scaling)
• Reports stored in single JSON file (consider database for large scale)
• No file cleanup mechanism (uploaded files persist indefinitely)

License

Proprietary - Refinery Eye AI