Digital Buffett - AI Financial Advisor

A sophisticated AI financial advisor system that mimics Warren Buffett's investment philosophy and decision-making process. Built with Flask backend and React frontend, featuring RAG (Retrieval-Augmented Generation), financial analysis tools, and belief tracking.

Features

Core Capabilities

• Chat Interface: Interactive conversation with AI Warren Buffett
• Financial Tools: DCF calculator, P/E analyzer, margin of safety calculator, Buffett score
• RAG System: Retrieval from Buffett's shareholder letters (1977-2021)
• Belief Tracking: Temporal belief system for market context
• Chain-of-Thought: Step-by-step reasoning display
• Portfolio Analysis: Multi-stock evaluation using Buffett methodology

Technical Architecture

• Backend: Flask with SQLAlchemy, OpenAI GPT-4 integration
• Frontend: React with Tailwind CSS and shadcn/ui components
• Database: SQLite for conversations, beliefs, and recommendations
• AI Integration: OpenAI API with custom prompting and tool calling
• Deployment: Render-ready with Docker support

Quick Start

Prerequisites

• Python 3.11+
• OpenAI API key
• Node.js 18+ (for development)

Environment Setup

1. Clone the repository
2. Set environment variable:

   export OPENAI_API_KEY="your-openai-api-key"

Local Development

1. Install Python dependencies:

   pip install -r requirements.txt

2. Run the Flask application:

   python src/main.py

3. Access the application at http://localhost:5000

Deployment on Render

Method 1: Direct GitHub Deployment

1. Fork this repository to your GitHub account
2. Connect your GitHub account to Render
3. Create a new Web Service on Render
4. Select this repository
5. Configure environment variables:
   • OPENAI_API_KEY: Your OpenAI API key
   • FLASK_ENV: production
6. Deploy

Method 2: Manual Upload

1. Download the ZIP file
2. Upload to Render as a new Web Service
3. Set build command: pip install -r requirements.txt
4. Set start command: python src/main.py
5. Configure environment variables as above

Project Structure

buffett-agent/
├── src/
│   ├── models/
│   │   └── conversation.py      # Database models
│   ├── routes/
│   │   └── buffett_routes.py    # API endpoints
│   ├── static/                  # Built React frontend
│   ├── database/
│   │   └── app.db              # SQLite database
│   ├── buffett_agent.py        # Main AI agent
│   ├── financial_tools.py      # Financial calculations
│   ├── rag_system.py          # Retrieval system
│   ├── belief_system.py       # Belief tracking
│   └── main.py                 # Flask application
├── requirements.txt            # Python dependencies
├── render.yaml                # Render configuration
└── README.md                  # This file

API Endpoints

Chat & Analysis

• POST /api/chat - Send message to Buffett agent
• GET /api/conversation/{session_id} - Get conversation history

Financial Tools

• POST /api/tools/dcf - Calculate DCF valuation
• POST /api/tools/pe - Analyze P/E ratios
• POST /api/tools/margin - Calculate margin of safety
• POST /api/tools/buffett-score - Calculate Buffett score

Belief System

• GET /api/beliefs - Get current beliefs
• POST /api/beliefs/update - Update beliefs from news

Portfolio

• POST /api/portfolio/analyze - Analyze portfolio

Usage Examples

Chat Interface

Ask questions like:

• "Should I invest in Apple at current prices?"
• "How do you calculate intrinsic value?"
• "What's your view on the current market?"

Financial Tools

Use the tools tab to:

• Calculate DCF for specific cash flows
• Analyze P/E ratios vs industry averages
• Determine margin of safety for investments
• Get overall Buffett score for companies

Belief Tracking

Monitor how market beliefs influence decisions:

• Fed policy changes
• Inflation trends
• Market sentiment shifts

Configuration

Environment Variables

• OPENAI_API_KEY: Required for AI functionality
• FLASK_ENV: Set to 'production' for deployment
• PORT: Port number (default: 5000)

Customization

• Modify src/rag_system.py to add more Buffett content
• Update src/financial_tools.py for additional calculations
• Extend src/belief_system.py for more market factors

Technical Details

AI System

• Uses OpenAI GPT-4 with custom system prompts
• Implements Chain-of-Thought reasoning
• RAG system with semantic search
• Tool calling for financial calculations

Database Schema

• Conversations: User messages and agent responses
• Beliefs: Temporal belief states with confidence scores
• Portfolio: Investment recommendations and analysis

Security

• Input validation and sanitization
• Rate limiting considerations
• API key protection
• CORS configuration for frontend

Troubleshooting

Common Issues

1. OpenAI API Key: Ensure valid API key is set
2. Database: SQLite file permissions
3. CORS: Frontend-backend communication
4. Dependencies: Check Python version compatibility

Logs

Check application logs for detailed error information:

python src/main.py

---

Project Improvements and Refactoring Summary on July 24th

This section documents the major improvements, refactoring, and bug fixes made to the project, with a focus on clarity, robustness, and maintainability. Each file/module is described in detail, highlighting the changes compared to the initial version.

---

1. src/buffett_agent.py

• Refactored agent orchestration:
  The BuffettAgent class now acts as the central orchestrator, coordinating between the RAG system, financial tools, and the belief system.
• Ticker extraction with LLM:
  Integrated LangChainTools.parse_tickers_with_langchain for robust ticker extraction from user queries, ensuring only valid tickers are processed.
• Dynamic news-driven belief updates:
  Added logic to fetch and process news for extracted tickers, updating the belief system automatically, evaluated by LLMs, and generated the key, summary, and confidence.
• Cleaner data flow:
  Ensured that all financial tool inputs are dynamically fetched and validated, reducing reliance on hardcoded or sample data.

---

2. src/belief_system.py

• Belief system modularization:
  The BeliefTracker class now manages beliefs, confidence levels, and causal relationships in a more modular way.
• LLM integration for belief updates:
  Refactored to use a pluggable LLM interface (LangChainOpenAI), allowing for flexible and up-to-date language model usage.
• Batch news processing:
  Improved the update_beliefs_from_news method to handle batch news updates, with robust parsing and error handling.

---

3. src/financial_tools.py

• Simplified DCF calculation:
  Replaced the original multi-period DCF with a Gordon Growth Model (simple_gordon_dcf), using only one forward FCF and the formula FCF1/(r-g).
• Per-share intrinsic value:
  Updated margin of safety calculations to use per-share intrinsic value (DCF divided by shares outstanding).

---

4. src/resources/fin_data.py

• Robust financial data acquisition:
  The MultiDataAgent class now fetches data from multiple sources (yfinance, finnhub, simfin, Alpha Vantage) with improved error handling.
• PE analysis improvements:
  Added methods to fetch historical PE ranges, industry average PE, and aggregate all necessary data for PE analysis.
• Margin of safety automation:
  Automated the process of fetching current price, FCF, and shares outstanding for margin of safety calculations.

---

5. src/parse_tools.py

• Centralized LLM parsing utilities:
  Created a new module to house all LLM-based parsing functions, such as ticker extraction and news-to-belief conversion.
• Modern LangChain usage:
  Updated to use llm.invoke() instead of deprecated call patterns, ensuring compatibility with the latest LangChain versions.
• Robust output parsing:
  Implemented fallback logic for JSON extraction and normalization, improving reliability when parsing LLM outputs.

---

6. src/resources/data_template.py

• Class-based data templates:
  Refactored from static methods to an instantiable DataTemplate class, supporting dynamic data population and validation.
• API alignment:
  Ensured that all data structures align with API requirements for seamless integration.

---

7. src/main.py

• Flask app configuration:
  Enhanced development experience by enabling debug mode for real-time error tracking and hot reloading

---

8. General Improvements

• Modernized LangChain usage:
  Removed deprecated patterns and ensured all LangChain calls are up-to-date.
• Ticker validation:
  Added a utility to validate tickers using yfinance, filtering out invalid or non-tradable symbols.
• Prompt engineering:
  Improved prompts for LLM-based extraction tasks, clarifying output formats and increasing reliability.
• Error handling:
  Enhanced error handling and logging throughout the codebase, especially for external API calls and LLM integration.
• Documentation:
  Added or improved code comments (in both English and Chinese) for better maintainability and onboarding.

---

How to Use This Section

• Use this summary to quickly understand the evolution of the codebase and the rationale behind major changes.
• Refer to the detailed comments in each file for implementation specifics and usage examples.
• For further questions or contributions, see the Contributing Guidelines and API Documentation.