📈 StockMind

AI-Powered Multi-Agent Market Simulation & Sentiment Analysis Platform

A neo-brutalist "War Room" where thousands of LLM agents trade in real-time.

---

What is StockMind?

StockMind is a multi-agent financial simulation system that models hedge funds, retail traders, market makers, and regulators — each with distinct strategies, risk profiles, and AI-driven decision logic.

Watch thousands of agents trade in real-time. Trigger market scenarios. Observe emergent behavior. Understand how sentiment, news, and agent psychology shape market dynamics.

This is not a trading app. It's a simulation engine for market research, strategy testing, and AI behavior analysis.

---

Agent Types

Agent	Strategy	Risk Profile	Behavior
🏦 Hedge Fund	Quantitative / Arbitrage	High (0.8)	Maximize alpha via complex timing
👤 Retail Trader	Emotional / Momentum	Low (0.3)	FOMO-driven herding behavior
📰 News Agent	Information Broker	N/A	Classifies & broadcasts sentiment
🏪 Market Maker	Delta Neutral / Spread	Med (0.2)	Provides liquidity, maintains spreads
⚖️ Regulator	Market Stability	N/A	Enforces circuit breakers (20% halt)

---

Key Features

⚡ Real-Time Simulation Engine

• FastAPI + AsyncIO backend orchestrates thousands of concurrent agent decisions
• WebSocket streaming delivers live market dynamics to the dashboard
• Sub-second latency from agent decision → chart update

🧠 AI-Driven Sentiment Analysis

• Groq LLaMA 3.3 70B processes real-time news ingestion
• Sentiment scores directly influence agent behavior and price movements
• Heatmaps visualize sentiment distribution across market sectors

📚 RAG-Powered Decision Making

• NVIDIA NIM embeddings (nv-embedqa-e5-v5) index historical market events
• Agents query this knowledge base to inform decisions — e.g., "this pattern looks like 2008"
• Retrieval-augmented reasoning makes agent behavior more realistic

🎮 Scenario Testing

Pre-configured market scenarios:

• 🚀 Bull Run
• 🐻 Bear Crash
• 🐸 Meme Stock Frenzy
• ⚡ Flash Crash

🖥️ Neo-Brutalist War Room Dashboard

• High-contrast dark mode with electric lime, purple, and pink accents
• Live Recharts visualizations: price charts, volume bars, sentiment heatmaps
• Agent activity feed — see each agent's decision in real-time

---

Tech Stack

Layer	Technology
Frontend	Next.js 15, React 19, TypeScript, Tailwind CSS, Recharts
Backend	FastAPI, Python, AsyncIO
Real-time	WebSockets
LLM	Groq LLaMA 3.3 70B
RAG	NVIDIA NIM Embeddings
Market Data	yfinance
News/Sentiment	NewsAPI
Database	Supabase PostgreSQL
Auth	JWT with token rotation

---

Architecture

                    ┌─────────────────────────────┐
                    │   Next.js War Room Dashboard  │
                    │  (Neo-brutalist real-time UI) │
                    └──────────────┬───────────────┘
                                   │ WebSocket
                    ┌──────────────▼───────────────┐
                    │      FastAPI Backend           │
                    │   Simulation Engine (AsyncIO)  │
                    └──┬──────────┬───────────┬─────┘
                       │          │           │
              ┌────────▼──┐  ┌───▼────┐  ┌──▼──────────┐
              │ Agent Swarm│  │  Groq  │  │  NVIDIA NIM  │
              │(1000s LLM) │  │  LLM   │  │  RAG Engine  │
              └────────────┘  └────────┘  └─────────────┘
                       │
              ┌────────▼──────────────┐
              │  yfinance + NewsAPI    │
              │  (Real market data)    │
              └───────────────────────┘

---

Getting Started

git clone https://github.com/pritpatel2412/StockMind
cd StockMind

# Frontend
npm install
npm run dev

# Backend
cd stockmind-backend
pip install -r requirements.txt
uvicorn main:app --reload

Required env vars: GROQ_API_KEY, NVIDIA_NIM_API_KEY, NEWS_API_KEY, SUPABASE_URL, SUPABASE_SERVICE_ROLE_KEY

---

Built By

Prit Patel — B.Tech CSE @ CHARUSAT University GitHub · LinkedIn