OffsetAI: Autonomous Financial Research Agent

A next-generation autonomous agent that performs deep financial research, risk analysis, and hedge strategy validation in real-time.

stocks-demo-1769491413416.1.1.mp4

🚀 Features

• Autonomous Agent Pipeline: Utilizes a Monitor -> Researcher -> Critic -> Strategist workflow using LangGraph.
• Real-Time Progress: Watch the agent "think" live on the dashboard as it moves through analysis steps.
• Deep Reasoning: Powered by Gemini 2.0 Flash for high-speed, reasoning-heavy financial analysis.
• RAG Memory: Integrated Pinecone vector database to recall historical context and regulatory news.
• Privacy-First MCP: Uses Model Context Protocol (MCP) to securely analyze your local portfolio (portfolio.csv) without uploading sensitive data to the cloud.
• Live Market Data: Real-time stock quotes via Alpaca API.
• Dynamic Dashboard: A premium, dark-mode UI built with Next.js 14 and Framer Motion.

🛠 Tech Stack

• Frontend: Next.js 14 (App Router), TypeScript, Framer Motion, Lucid React.
• Backend: FastAPI (Python), LangGraph (Agent Orchestration), Pydantic.
• AI & Data: Google Gemini 2.0, Pinecone (Vector DB), Alpaca (Market Data).
• Database: Supabase (PostgreSQL + Auth).
• Security: MCP (Model Context Protocol) for local file access.

⚡ Quick Start

Prerequisites

• Python 3.11+
• Node.js 18+
• Api Keys: Supabase, Alpaca, Gemini, Pinecone.

1. Backend Setup

# Navigate to project root
cd stocks

# install dependencies
pip install -r backend/requirements.txt

# Run the backend server
python -m backend.app.main

Server runs on http://localhost:8000

2. Frontend Setup

# Open a new terminal
cd stocks

# Install packages
npm install

# Start the dashboard
npm run dev

UI runs on http://localhost:3000

🧠 Architecture

The system uses a State Graph architecture:

1. Monitor: Detects price movements or user triggers.
2. Researcher: Scrapes news and queries Vector DB (RAG) for context.
3. Critic: Validates findings against Prediction Markets (Polymarket).
4. MCP: Securely calculates risk exposure against your local portfolio file.
5. Strategist: Synthesizes all data to generate a BUY/SELL/HEDGE signal.

🛡 Security

• .env files are strictly git-ignored.
• Portfolio data is processed locally via MCP where possible.
• Auth is handled securely via Supabase.

---

Disclaimer: This project is for educational purposes only. Not financial advice.