Quantis - AI Portfolio Intelligence Platform

Last updated: May 16, 2026

An India-focused AI portfolio intelligence platform using deterministic portfolio construction, quantitative risk analytics, authenticated portfolio management, Monte Carlo simulation, and adversarial multi-agent debate orchestration.

Features

• Portfolio Optimization: Mean-variance optimization using PyPortfolioOpt
• Risk Analytics: VaR, Sharpe ratio, max drawdown, beta, alpha calculations
• Monte Carlo Simulation: 1000+ path simulations for scenario analysis
• Debate-Based Multi-Agent AI: Adversarial reasoning with Groq LLM
  • Aggressive Agent - maximizes growth, high-beta exposure
  • Historical Agent - evaluates drawdowns, regime performance
  • Risk Agent - enforces volatility, VaR, diversification constraints
  • Consensus Engine - resolves disagreements, produces final allocation
• Live Market Data: Real-time NSE market quotes via Polygon.io / Alpha Vantage / Yahoo Finance
• Hybrid Caching: In-memory + persistent CSV cache with incremental updates
• Mock Data Fallback: Realistic generated data when all APIs fail
• Authenticated Portfolio Management: Save portfolios, saved scenarios, watchlists, goals, alerts, notifications, and risk assessment
• Modern UI: Bloomberg-style dark interface with Recharts and shared authenticated layout

Tech Stack

Backend

• Python 3.11+
• FastAPI
• pandas, numpy, scipy
• PyPortfolioOpt
• yfinance (fallback)
• Polygon.io (primary data source)
• Alpha Vantage (secondary)
• Groq (LLM inference)

Frontend

• React 19
• TypeScript
• Vite
• Tailwind CSS
• Recharts
• JWT bearer token stored in localStorage by the auth context

Quick Start

Prerequisites

• Node.js 18+
• Python 3.11+

Backend Setup

cd backend
python -m venv venv
.\venv\Scripts\activate  # Linux/Mac: source venv/bin/activate
pip install -r requirements.txt

# Copy and configure .env
cp .env.example .env
# Add GROQ_API_KEY for AI agents (optional - falls back to deterministic)

Frontend Setup

cd frontend
npm install
npm run dev

Configuration

Create backend/.env with:

# Required for AI debate agents (get free key at https://console.groq.com/)
GROQ_API_KEY=your_groq_key

# Market Data APIs (at least one recommended)
POLYGON_API_KEY=your_polygon_key          # Primary - 5 calls/min free tier
ALPHA_VANTAGE_KEY=your_alpha_vantage_key  # Secondary - 5 calls/min free tier
FINNHUB_KEY=your_finnhub_key            # Optional - sentiment data

# Shoonya API (optional - for live Indian market data)
SHOONYA_API_KEY=
SHOONYA_USER_ID=
SHOONYA_PASSWORD=

Caching System

The platform uses a hybrid caching system for market data:

Layer	Storage	TTL	Purpose
Memory	In-memory (dict)	5 minutes	Fast access for repeated requests
File	backend/cache/*.csv	24 hours	Persist across restarts
Incremental	Date-range queries	Per request	Only fetch new data points

Cache Behavior

1. First request: Fetches full data from API, saves to memory + file
2. Subsequent requests: Uses cached data if recent
3. Old cache: Fetches only new data points since last cache date
4. All APIs fail: Generates realistic mock data as fallback

Cache Directory

backend/cache/
├── RELIANCE_NS.csv    # Cached market data
├── TCS_NS.csv
├── INFY_NS.csv
└── ...

Cache files are gitignored and stored persistently.

API Endpoints

Method	Endpoint	Description
GET	/health	Health check
POST	/api/portfolio	Generate portfolio with debate agents
GET	/api/live	Live market data
POST	/api/auth/login	Login and receive access token
POST	/api/auth/register	Register and receive access token
GET/POST/PUT/DELETE	/api/portfolios	Saved portfolio management
GET/POST/PUT/DELETE	/api/scenarios	Saved scenario management
GET/POST/PUT/DELETE	/api/watchlists	Watchlist management
GET/POST/PUT/DELETE	/api/goals	Goal tracking
GET/POST/PUT/DELETE	/api/alerts	Alert management
GET/PUT	/api/notifications	Notifications and read state
GET/POST	/api/risk-questionnaire	Risk assessment and submission

Example Request

POST /api/portfolio
{
  "age": 30,
  "risk_profile": "moderate",
  "investment_amount": 100000,
  "horizon_years": 5
}

Supported Assets

• NIFTYBEES.NS (Nifty Index ETF)
• GOLDBEES.NS (Gold ETF)
• LIQUIDBEES.NS (Money Market ETF)
• RELIANCE.NS, INFY.NS, TCS.NS
• HDFCBANK.NS, ICICIBANK.NS, TATAMOTORS.NS
• ^NSEI (Nifty 50 Index)

Architecture

User Input → Market Data (with caching) → Feature Engineering → Portfolio Optimization → Risk Model → Monte Carlo → Debate Agents → Consensus → Output

Data Source Priority

1. Polygon.io - Primary (reliable for Indian stocks)
2. Alpha Vantage - Secondary
3. Yahoo Finance - Tertiary fallback
4. Mock Data - Ultimate fallback (always works)

Debate Flow

1. Three agents (Aggressive, Historical, Risk) analyze portfolio in parallel
2. Each returns JSON with recommendations and confidence scores
3. Consensus Engine averages recommendations, highlights disagreements
4. Final allocation adjusted based on weighted agent outputs

Production Metrics

• Expected Return: Calculated from historical returns (annualized)
• Volatility: Annualized standard deviation
• Sharpe Ratio: (Return - Risk-Free Rate) / Volatility
• Max Drawdown: Largest peak-to-trough decline
• Beta: Correlation with Nifty 50 benchmark
• VaR 95%: Value at Risk at 95% confidence

License

MIT

Status

Current app date: May 16, 2026. The authenticated portfolio-management flows are active and the backend login/protected routes are verified with the demo account.