An AI-powered WhatsApp Notification Router that intelligently classifies incoming messages as Notify, Digest, or Mute using a hybrid AI architecture powered by Google Gemini, rule-based reasoning, and personalized scoring.

SignalOS is an AI-powered notification routing system that reduces notification overload by intelligently deciding whether an incoming WhatsApp message should:

• 🔔 Notify immediately
• 📋 Be included in a digest
• 🔕 Be muted

The system combines deterministic rules, contextual AI reasoning, historical interactions, and personalized scoring to make reliable routing decisions.

• 🤖 Google Gemini AI Integration
• 🧠 Hybrid AI + Rule Engine
• 👤 Personalized Notification Routing
• 💬 Text Message Analysis
• 🖼️ Image & Screenshot Support
• 🎤 Voice Note Support
• 📊 Analytics Dashboard
• 📱 Fully Responsive UI
• 📈 Live Routing Simulation
• 📂 CSV Export
• 🔒 Secure API Key Management
• ⚡ Fast & Lightweight Architecture

.
├── ai/                     # Gemini AI prompts and reasoning
├── assets/                 # Static assets
├── config/                 # Configuration files
├── dataset/                # Competition datasets
├── evaluation/             # Benchmark & evaluation scripts
├── history/                # Historical interaction retrieval
├── loaders/                # Dataset loaders
├── models/                 # Data models
├── outputs/                # Generated predictions
├── public/                 # Public frontend assets
├── routing/                # Notification routing engine
├── src/
│   ├── assets/
│   ├── components/
│   ├── App.tsx
│   ├── index.css
│   ├── main.tsx
│   └── types.ts
├── tests/                  # Test cases
├── utils/                  # Helper utilities
├── .env.example
├── .gitignore
├── bun.lock
├── index.html
├── main.py                 # Backend entry point
├── metadata.json
├── output.csv
├── package.json
├── README.md
├── requirements.txt
├── server.ts               # Frontend server
├── tsconfig.json
└── vite.config.ts

Incoming Message
        │
        ▼
Dataset Loader
        │
        ▼
Rule Engine
        │
        ▼
History Retrieval
        │
        ▼
Gemini AI
        │
        ▼
Score-Based Router
        │
        ▼
Validation
        │
        ▼
Notify / Digest / Mute

Handles deterministic cases:

• Emergency alerts
• Scam detection
• Phishing
• OTP
• Promotions
• Spam

Uses

• Sender information
• Group metadata
• Business accounts
• User history
• Previous interactions

Provides contextual reasoning by understanding

• Intent
• Urgency
• Importance
• Context
• User relevance

Calculates the final routing decision using:

• Importance
• Urgency
• Trust
• Risk
• Personalization

Ensures:

• Emergencies are never muted
• Scams are never notified
• Final routing is consistent

The web dashboard includes:

• 📊 Analytics
• 🔔 Live Feed
• 🧪 Message Simulator
• 📚 Knowledge Base
• 📤 CSV Export
• 📈 AI Decision Visualization

• React
• TypeScript
• Vite
• CSS

• Python
• Gemini API
• Pandas

• Google Gemini
• Rule Engine
• Context Retrieval
• Hybrid Routing Pipeline

The application reads:

• messages.csv
• users.csv
• groups.csv
• images.csv
• business_accounts.csv
• message_history.csv
• message_events.csv

along with local media files.

The system generates

output.csv

containing

• Message ID
• Routing Action
• Confidence
• Explanation
• Supporting Evidence

git clone https://github.com/yourusername/SignalOS.git
cd SignalOS

pip install -r requirements.txt

npm install

or

bun install

Create .env

GEMINI_API_KEY=YOUR_API_KEY
APP_URL=http://localhost:3000

python main.py

npm run dev

or

bun run dev

The project includes a dedicated evaluation pipeline for benchmarking routing performance on labeled samples and generating the required competition output.

• Environment Variables
• No Hardcoded API Keys
• Secure Gemini Integration
• Deterministic Safety Rules

• Offline AI Support
• Local LLM Fallback
• Multi-language Messages
• Better Voice Understanding
• Adaptive User Learning
• WhatsApp API Integration

Vijay Kumar

B.Tech Computer Science Engineering

AI • Machine Learning • Full Stack Development