PortfolioAgent

A multi-asset portfolio management system powered by LLMs (Large Language Models) that provides intelligent asset allocation recommendations across multiple asset classes.

Overview

PortfolioAgent is a framework that combines traditional financial models with AI-powered insights to optimize investment portfolios. The system integrates data from various financial markets and uses LLMs to analyze this data and produce investment recommendations.

Features

• Multi-asset Class Coverage:

  • Equities (Sector ETFs)
  • Fixed Income (Treasury and Credit ETFs)
  • Commodities
  • Foreign Exchange

• Intelligent Analysis:

  • Uses OpenAI's GPT models to analyze market data
  • Combines traditional financial metrics with AI insights
  • Produces structured recommendations with confidence scores

• Advanced Backtesting:

  • Black-Litterman model integration
  • Performance evaluation over historical data

Project Structure

├── src/                     # Source code
│   ├── agent/               # Agent implementations
│   │   ├── DataCollector.py     # Data collection utilities
│   │   ├── PortfolioAgent.py    # Base agent class
│   │   ├── equity_agent.py      # Equity-specific agent
│   │   ├── fixedIncome_agent.py # Fixed income agent
│   │   ├── fx_agent.py          # Foreign exchange agent
│   │   ├── commodity_agent.py   # Commodity agent
│   │   └── manager_agent.py     # Manager agent coordinating other agents

│   ├── backtest/            # Backtesting framework
│   │   └── blackLitterman_test.py # Black-Litterman model implementation
│   ├── config/              # Configuration settings
│   │   └── settings.py      # Portfolio definitions and settings
│   ├── evaluator/           # Performance evaluation tools
│   └── util/                # Utility functions
├── data/                    # Data storage directory
├── report/                  # Analysis reports
├── slides/                  # Presentation materials
└── literature/              # Research papers and references

Requirements

• Python 3.8+
• OpenAI API Key
• FRED API Key (for macroeconomic data)
• WRDS Login
• Dependencies listed in requirements.txt

Setup

1. Clone the repository
2. Create a .env file with your API keys:

   OPENAI_API_KEY=your_openai_api_key
   FRED_API_KEY=your_fred_api_key
   WRDS_USERNAME=your_wrds_username 
   

3. Install required dependencies:

   pip install -r requirements.txt
   

Usage

The system operates through specialized agents for each asset class that:

1. Collect relevant market data via DataCollector classes
2. Process and prepare data for LLM analysis
3. Generate investment recommendations using structured LLM prompts
4. Combine recommendations across asset classes via the manager_agent

Acknowledgements

This project builds on research in portfolio optimization, specifically the Black-Litterman model, and leverages large language models for financial analysis.