📡 Opportunity Radar: Autonomous Market Intelligence Swarm

Empowering Investors with Multi-Agent Intelligence.
Developed for the ET Gen AI Hackathon 2026

Opportunity Radar is a production-grade market intelligence portal that autonomously monitors BSE/NSE corporate disclosures, institutional bulk deals, and market sentiment. By leveraging a self-orchestrating swarm of 5 AI agents (powered by CrewAI, Sarvam AI, and Gemini), the platform filters through thousands of routine filings to extract high-conviction "Material Alpha" in real-time.

---

🚀 Core Platform Features

1. The Intelligence Portal (Frontend)

A premium, responsive React dashboard featuring four primary views:

• Dashboard: Real-time NIFTY/SENSEX market stats and the Top 5 most recent high-conviction signals.
• Live Signals (Radar): A comprehensive feed of all agent-generated alerts with dynamic filtering by sector, category, and intelligent search.
• Watchlist: Track specific stocks with AI sentiment dynamically mapped from the latest intelligence signals (Bullish, Bearish, Neutral).
• Backtest Lab: Validate signal precision. Replay our proprietary SMA 20/50 "Golden Cross" detection algorithm against historical price action to simulate win rates and equity curves.

2. Multi-Channel Alert System

High-conviction alerts are autonomously routed directly to investors:

• Email (Resend): Triggered for signals with a Conviction Score ≥ 7.0.
• WhatsApp (Twilio): Immediate priority delivery for extreme conviction signals (Score ≥ 9.0).
• Both channels format alerts identically to the specified PRD templates.

3. Absolute Auditability (Database)

The system is designed for 100% traceability via Supabase PostgreSQL:

• raw_events: The original, unmodified JSON scraped from the exchanges.
• agent_outputs: The raw "thoughts", reasoning, and sentiment of individual agents.
• signals: The final, synthesized intelligence result shown to the user.

---

🧠 Hybrid Swarm Architecture

We utilize a Hierarchical Multi-Agent System that splits cognitive load for maximum efficiency and minimum latency.

graph TD
    Data[Market Data Streams: BSE, NSE, Screener] --> Orch{Orchestrator}
    Orch --> A[Filing Watcher]
    Orch --> B[Deal Tracker]
    Orch --> C[Results Analyzer]
    Orch --> D[Sentiment Analyzer]
    A & B & C & D --> E[Signal Scorer Agent]
    E --> DB[(Supabase Audit DB)]
    DB --> UI[React Frontend]
    DB --> Notify[Resend/Twilio Alerts]

Loading

The 5 Agents

1. BSE Filing Analyst (Powered by Gemini 1.5 Flash): Chews through high-volume corporate announcements to distinguish "Routine Administrative" from "Material Impactful" disclosures.
2. Institutional Deal Tracker (Powered by Gemini 1.5 Flash): Monitors NSE 'Bulk & Block' deals to identify "Smart Money" accumulation and distribution patterns.
3. Results Analyzer: Parses quarterly earnings (Screener.in) to identify PAT beats and margin expansions.
4. Management Sentiment Analyzer: Extracts forward-looking confidence metrics from earnings concall transcripts.
5. Signal Conviction Scorer (Powered by Sarvam-M 24B): The final intelligence layer. Synthesizes the diverging viewpoints of all other agents into a single, definitive conviction score (0-10) and determines the final action_suggestion.

---

🛠 Technical Stack

• Intelligence: Sarvam-M 24B (Reasoning) & Gemini 2.5 Flash (Parsing) orchestrated by CrewAI.
• Backend / Pipeline: FastAPI + Uvicorn Python backend running an APScheduler high-frequency polling loop.
• Data Integrations: yfinance (Live Price/Indices), Custom BSE/NSE parsers, screener_client.py.
• Database: Supabase (PostgreSQL) leveraging JSONB columns for raw exchange data storage.
• Frontend: React + Vite + CSS Variables (Glassmorphism UI) + Recharts + Framer Motion.
• Notifications: httpx direct API integration with Twilio (WhatsApp) and Resend (Email).

---

🏁 Quick Start & Setup

1. Install Dependencies

# Frontend
cd frontend
npm install

# Backend
cd backend
pip install -r requirements.txt

2. Environment Variables

Create a .env file in the backend/ directory using .env.example as a template:

# AI Models
SARVAM_API_KEY=your_sarvam_key
GEMINI_API_KEY=your_gemini_key

# Database
SUPABASE_URL=your_project_url
SUPABASE_KEY=your_service_role_key

# Notifications
RESEND_API_KEY=your_resend_api_key
TWILIO_SID=your_twilio_sid
TWILIO_TOKEN=your_twilio_token
TWILIO_WHATSAPP_FROM=whatsapp:+14155238886
ALERT_EMAIL=investor@example.com
ALERT_WHATSAPP=whatsapp:+919876543210

3. Launch the Platform

Start the backend orchestrator (port 8000) and frontend portal (port 5173) in separate terminals:

# Terminal 1: Backend Swarm
cd backend
python main.py

# Terminal 2: Intelligence Portal
cd frontend
npm run dev

---

Created for the ET Gen AI Hackathon 2026.