charieswei
/
TradingAgents
mirror of https://github.com/TauricResearch/TradingAgents.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
TradingAgents/README.md

12 KiB
Raw Blame History

Litadel: Multi-Agents LLM Financial Trading Framework

Copyright Notice: Litadel is a successor of TradingAgents. This project builds upon and extends the original TradingAgents framework.

🚀 Overview | 💻 Dashboard | ⚡ Getting Started | 🎯 Usage | 🤖 How It Works | 🤝 Contributing | 📄 Citation

Overview

Litadel is a comprehensive AI-powered trading analysis platform that delivers professional-grade market insights across equities, commodities, and cryptocurrencies. Get actionable BUY/SELL/HOLD recommendations backed by multi-agent analysis covering fundamentals, technicals, news sentiment, and risk assessment.

Litadel framework is designed for research and educational purposes. Trading performance may vary based on many factors. It is not intended as financial, investment, or trading advice.

What You Get

Three Ways to Analyze:

  • 🌐 Web Dashboard - Modern, real-time interface with live tracking, interactive charts, and comprehensive reports
  • 💻 Interactive CLI - Rich terminal experience with live agent progress and automatic report generation
  • 📦 Python Package - Integrate multi-agent analysis directly into your own applications

Multi-Asset Coverage:

  • 📈 Equities - Full fundamental, technical, and sentiment analysis for stocks
  • 🛢️ Commodities - Specialized analysis for oil, metals, agricultural products, and more
  • ₿ Cryptocurrencies - Real-time crypto market analysis with sentiment tracking

Professional Analysis:

  • Real-time market data with automatic caching
  • Multi-perspective analysis with bull vs. bear debates
  • Comprehensive reports covering all aspects of market conditions
  • Clear trading recommendations with confidence scores

Dashboard

Your Trading Command Center

The web dashboard provides a complete control center for managing your trading analyses with real-time monitoring, interactive visualizations, and comprehensive reporting.

Analysis Management

Browse all your analyses with smart filtering and grouping. Track active analyses in real-time and review historical decisions with detailed statistics.

Real-Time Analysis Tracking

Watch your analysis unfold in real-time as AI agents collaborate to evaluate market conditions. See live progress updates, agent pipeline status, and streaming reports as they're generated.

Comprehensive Analysis Reports

Each completed analysis provides detailed insights with:

  • Trading Decision - Clear BUY/SELL/HOLD recommendation with confidence score
  • Interactive Price Charts - Candlestick charts with analysis date markers and 60-day history
  • Market Metrics - Current price, daily change, volume, and 52-week ranges
  • Specialist Reports - Detailed analysis from market, news, sentiment, and fundamental perspectives
  • Research Debate - Bull vs. bear perspectives with investment recommendations
  • Risk Assessment - Comprehensive risk evaluation and portfolio impact analysis

Key Features

  • Real-Time WebSocket Updates - Live progress tracking without page refreshes
  • Interactive Charts - Visualize price action with candlestick or line charts
  • Export Capabilities - Download complete analysis data as JSON
  • Analysis History - Browse and compare past analyses by ticker and date
  • Secure API Access - API key authentication with configurable endpoints
  • Responsive Design - Works seamlessly on desktop and tablet devices

Getting Started

Installation

Clone Litadel:

git clone https://github.com/deepweather/Litadel.git
cd Litadel

Create a virtual environment:

conda create -n litadel python=3.13
conda activate litadel

Install dependencies:

pip install -r requirements.txt

API Keys Setup

You will need API keys for LLM providers and market data. The default configuration uses OpenAI for agents and Alpha Vantage for market data.

Create a .env file in the project root:

cp .env.example .env
# Edit .env with your actual API keys

Or export them directly:

export OPENAI_API_KEY=$YOUR_OPENAI_API_KEY
export ALPHA_VANTAGE_API_KEY=$YOUR_ALPHA_VANTAGE_API_KEY

Note: Litadel partners with Alpha Vantage to provide robust API support. Get a free API key here—Litadel users receive increased rate limits (60 requests/minute, no daily limits) through Alpha Vantage's open-source support program.

Usage

Web Dashboard (Recommended)

The web interface provides the most comprehensive experience with real-time tracking, interactive charts, and complete analysis history.

1. Start the API Server:

python -m api.main

On first run, the system will automatically create a database and generate an API key. Save this key—you'll need it for the web interface.

2. Start the Frontend:

cd frontend
npm install
npm run dev

3. Access the Dashboard:

Open your browser to http://localhost:5173 and enter your API key in Settings. You're ready to create your first analysis!

Interactive CLI

For a terminal-based experience with live agent progress tracking:

python -m cli.main

Select your ticker, analysis date, analyst team, LLM models, and research depth through the interactive prompts.

Watch as agents collaborate in real-time, with live updates showing their reasoning and tool usage:

Results are automatically saved to results/<TICKER>/<DATE>/ with detailed logs and markdown reports.

Python Package

Integrate Litadel's multi-agent analysis directly into your own applications, trading bots, or research pipelines.

Basic Usage:

from tradingagents.graph.trading_graph import TradingAgentsGraph
from tradingagents.default_config import DEFAULT_CONFIG

# Initialize the trading agents
ta = TradingAgentsGraph(debug=True, config=DEFAULT_CONFIG.copy())

# Run analysis and get trading decision
_, decision = ta.propagate("NVDA", "2024-05-10")
print(decision)

Custom Configuration:

Customize LLM models, debate rounds, and data sources to match your needs:

from tradingagents.graph.trading_graph import TradingAgentsGraph
from tradingagents.default_config import DEFAULT_CONFIG

# Create custom configuration
config = DEFAULT_CONFIG.copy()
config["deep_think_llm"] = "o1-mini"           # Deep reasoning model
config["quick_think_llm"] = "gpt-4o-mini"      # Fast operations model
config["max_debate_rounds"] = 3                # More thorough research debates

# Configure data sources
config["data_vendors"] = {
    "core_stock_apis": "yfinance",             # Price data
    "technical_indicators": "yfinance",        # Technical analysis
    "fundamental_data": "alpha_vantage",       # Company fundamentals
    "news_data": "alpha_vantage",              # News and sentiment
}

# Run with custom config
ta = TradingAgentsGraph(debug=True, config=config)
_, decision = ta.propagate("AAPL", "2024-05-10")

Cost Optimization:

For testing and development, we recommend using gpt-4o-mini and o1-mini to minimize costs, as the multi-agent framework makes numerous API calls during analysis. For production use with higher accuracy requirements, consider gpt-4o and o1-preview.

Data Sources:

The default configuration uses YFinance for price/technical data and Alpha Vantage for fundamentals/news. You can switch to OpenAI for web-based data fetching or use local cached data for offline experimentation. See tradingagents/default_config.py for all available options.

How It Works

Litadel uses a multi-agent architecture that mirrors the structure of professional trading firms. Specialized AI agents collaborate to provide comprehensive market analysis.

Analyst Team

Four specialized analysts evaluate different aspects of market conditions:

  • Technical Analyst - Analyzes price patterns, trends, and technical indicators (MACD, RSI, moving averages)
  • Fundamentals Analyst - Evaluates company financials, earnings, balance sheets, and intrinsic value
  • News Analyst - Monitors global news, macroeconomic indicators, and their market impact
  • Sentiment Analyst - Analyzes social media and public sentiment to gauge market mood

Researcher Team

Bull and bear researchers critically assess analyst insights through structured debates, balancing potential gains against risks.

Trader Agent

Synthesizes all reports and research to make informed trading decisions with clear timing and position sizing recommendations.

Risk Management and Portfolio Manager

Evaluates portfolio risk by assessing market volatility, liquidity, and other risk factors. The risk team provides final assessment and approval for trading decisions.

What's New in Litadel

Completed Features

  • ✅ Web Dashboard - Full-featured web interface with real-time analysis tracking
  • ✅ REST API - Complete API for programmatic access with WebSocket support
  • ✅ Multi-Asset Support - Equities, commodities, and cryptocurrencies
  • ✅ Interactive Charts - Real-time candlestick and line charts with market data
  • ✅ Analysis History - Persistent storage and browsing of all analyses
  • ✅ Export Capabilities - Download complete analysis data as JSON

Roadmap

  • 🚧 Automated Trading Mode - Continuous automated trading execution
  • 🚧 Portfolio Management - Multi-asset portfolio tracking and optimization
  • 🚧 Backtesting Engine - Historical performance analysis with TauricDB
  • 🚧 OpenAI Agents SDK Migration - Enhanced parallelization and maintainability

Contributing

We welcome contributions from the community! Whether it's fixing a bug, improving documentation, or suggesting a new feature, your input helps make this project better.

Citation

Please reference our work if you find Litadel provides you with some help :)

Litadel citation:

@software{gabler2025litadel,
      title={Litadel: Multi-Agents LLM Financial Trading Framework}, 
      author={Marvin Gabler},
      year={2025},
      url={https://github.com/deepweather/Litadel},
      note={Extended framework based on TradingAgents}
}

Original TradingAgents citation:

@misc{xiao2025tradingagentsmultiagentsllmfinancial,
      title={TradingAgents: Multi-Agents LLM Financial Trading Framework}, 
      author={Yijia Xiao and Edward Sun and Di Luo and Wei Wang},
      year={2025},
      eprint={2412.20138},
      archivePrefix={arXiv},
      primaryClass={q-fin.TR},
      url={https://arxiv.org/abs/2412.20138}, 
}