Financial Analysis Multi-Agent System

Overview

This project implements a multi-agent system for comprehensive financial analysis of stocks. It combines web searching, stock data analysis, and report generation using three specialized AI agents working in collaboration.

Features

• Real-time stock data analysis using Yahoo Finance
• Web search integration for latest company information
• Automated report generation and saving
• Technical analysis including moving averages and trend analysis
• Stock price visualization with charts

Agents

1. Google Search Agent: Gathers recent information about companies
2. Stock Analysis Agent: Performs technical analysis on stock data
3. Report Agent: Compiles information into comprehensive reports

Prerequisites

pip install yfinance matplotlib pytz numpy pandas python-dotenv requests bs4 autogen

Environment Setup

Create a .env file in the project root with the following:

GOOGLE_SEARCH_ENGINE_ID=your_search_engine_id
GOOGLE_API_KEY=your_google_api_key
OPENAI_API_KEY=your_openai_api_key

Directory Structure

project/
├── .env
├── stock_analyzer.py
├── coding/
│   ├── stock_plots/
│   └── reports/
└── README.md

Usage

1. Ensure all environment variables are set in .env
2. Run the script:

python stock_analyzer.py

Functions

google_search()

Searches Google for recent information about a company or topic.

• Parameters:
  • query (str): Search query
  • num_results (int): Number of results to return (default: 2)
  • max_chars (int): Maximum characters per result (default: 500)

analyze_stock()

Performs technical analysis on a given stock ticker.

• Parameters:
  • ticker (str): Stock symbol (e.g., "AAL" for American Airlines)
• Returns:
  • Dictionary containing:
    • Current price
    • 52-week high/low
    • Moving averages
    • Trend analysis
    • Volatility metrics
    • Price chart (saved as PNG)

save_txt_tool()

Saves generated reports to text files.

• Parameters:
  • text (str): Report content
  • filename (str): Name for the report file (default: "report.txt")

Output

The system generates:

1. Stock analysis plots (saved in coding/stock_plots/)
2. Text reports (saved in coding/reports/)
3. Console output showing agent interactions

Example

To analyze American Airlines:

async def main():
    stream = team.run_stream(task="Write a financial report on American airlines")
    await Console(stream)

if __name__ == "__main__":
    asyncio.run(main())

Notes

• The system uses GPT-4 for all agents
• Each agent has a specific role and tools
• The analysis includes both technical and fundamental data
• Reports are saved automatically in the coding directory

Limitations

• Limited to publicly available stock data
• Requires valid API keys for all services
• Google search is limited to 2 results by default
• Plots are generated in non-interactive mode

Error Handling

• Validates API keys before execution
• Handles missing stock data gracefully
• Provides error messages for failed API requests
• Ensures proper file saving with error reporting

Contributing

Feel free to submit issues and enhancement requests!

License

MIT License