tradingagents 0.0.0

TradingAgents: Multi-Agents LLM Financial Trading Framework

TradingAgents

🚀 TradingAgents is a multi-agent LLM financial trading framework that leverages large language models to simulate analyst teams, research debates, and portfolio management decisions for stock trading analysis.

Other Languages: English | 繁體中文 | 简体中文

✨ Highlights

• Built on LangGraph and AG2 (AutoGen) for robust multi-agent orchestration
• Multi-agent architecture: Analyst Team → Research Team → Trader → Risk Management → Portfolio Management
• Support for multiple LLM providers: OpenAI, Anthropic, Google Gemini, xAI (Grok), OpenRouter, Ollama
• Market data powered by yfinance for OHLCV, fundamentals, technical indicators, news, and insider transactions
• Pydantic-based configuration with strict typing and validation
• Analysis results automatically saved to results/ with organized subfolders
• Modern src/ layout with full type-annotated code
• Fast dependency management via uv
• Pre-commit suite: ruff, mdformat, codespell, mypy, uv hooks
• Pytest with coverage; MkDocs Material documentation

🚀 Quick Start

git clone https://github.com/Mai0313/TradingAgents.git
cd TradingAgents
make uv-install               # Install uv (only needed once)
uv sync                       # Install dependencies
cp .env.example .env          # Configure your API keys

Configure API Keys

Edit .env and set your LLM provider keys:

# LLM Providers (set the one you use)
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GOOGLE_API_KEY=AIza...
XAI_API_KEY=...
OPENROUTER_API_KEY=...

Usage

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

config = TradingAgentsConfig(
    llm_provider="openai",
    deep_think_llm="gpt-5.2",
    quick_think_llm="gpt-5-mini",
    max_debate_rounds=1,
)

ta = TradingAgentsGraph(debug=True, config=config)
_, decision = ta.propagate("NVDA", "2024-05-10")
print(decision)

📁 Project Structure

src/
└── tradingagents/
    ├── agents/           # Agent implementations
    │   ├── analysts/     # Market, News, Social, Fundamentals analysts
    │   ├── managers/     # Research & Portfolio managers
    │   ├── researchers/  # Bull & Bear researchers
    │   ├── risk_mgmt/    # Risk management agents
    │   ├── trader/       # Trader agent
    │   └── utils/        # Shared agent utilities
    ├── dataflows/        # Data ingestion via yfinance
    ├── graph/            # LangGraph trading graph setup
    ├── llm_clients/      # LLM provider clients (OpenAI, Anthropic, Google, xAI, OpenRouter, Ollama)
    └── default_config.py # Default configuration

🤖 Agent Workflow

1. Analyst Team — Each selected analyst independently researches market data, news, sentiment, and fundamentals
2. Research Team — Bull and Bear researchers debate; Research Manager makes a final investment decision
3. Trader — Formulates a trade plan based on research
4. Risk Management — Three risk analysts (aggressive, neutral, conservative) debate risk
5. Portfolio Manager — Makes the final trade decision based on all inputs

Results are saved to results/<TICKER>/<DATE>/ with per-team sub-folders and a consolidated complete_report.md.

🤝 Contributing

For development instructions including documentation, testing, and Docker services, please see CONTRIBUTING.md.

• Open issues/PRs
• Follow the coding style (ruff, type hints)
• Use Conventional Commit messages and descriptive PR titles

📄 License

MIT — see LICENSE.