Alpha Arena

The platform enforces the following trading and risk constraints:

• Initial Capital: $10,000 USDT per model.
• Order Size: Maximum 20% of Net Asset Value (NAV) per trade.
• Position Limit: Maximum of 1 open symbol at a time.
• Stop Loss/Take Profit: Provided by the model, but the system enforces a hard liquidation threshold at -5%.
• Fees & Slippage: 0.05% (5 bp) fee; 10 bp slippage (in paper-trading).
• Decision Cadence: 5 minutes (only one new decision allowed per 5m window).
• Timeout Handling: If an LLM fails to respond within 8 seconds, the system defaults to a HOLD action.
• Kill-Switch: If Net Value drawdown exceeds 10%, the system immediately stops, flattens all positions, and disables new orders.

The project is organized into core logic, adapters for external services, and a main entry point:

• core/: Contains the primary business logic.
  • market.py: Manages market data.
  • decision.py: The decision engine.
• adapters/: Implements interfaces for external services.
  • llm_base.py: Base class for Large Language Models.
  • openai_adapter.py: Adapter for OpenAI.
  • claude_adapter.py: Adapter for Anthropic Claude.
  • exchange_api.py: Adapter for exchange APIs (Bitget).
• main.py: The main application entry point.
• requirements.txt: List of Python dependencies.
• env.example: Template for environment variables.

alpha-arena/
├── core/
│   ├── market.py
│   └── decision.py
├── adapters/
│   ├── llm_base.py
│   ├── openai_adapter.py
│   ├── claude_adapter.py
│   └── exchange_api.py
├── main.py
├── requirements.txt
└── env.example

To run the minimal MVP (which includes real-time price fetching and comparison of OpenAI vs Claude decisions), follow these steps:

1. Clone the repository.
2. Install dependencies via pip.
3. Configure your API keys in a .env file.
4. Execute the main entry point.

Note: The current version is v0.1.0.

# 1. 克隆项目
git clone https://github.com/AmadeusGB/alpha-arena.git
cd alpha-arena

# 2. 安装依赖
pip install -r requirements.txt

# 3. 配置API密钥
cp env.example .env
# 编辑.env文件，填入你的API密钥

# 4. 运行程序
python main.py

To run the Alpha Arena MVP, ensure you have Python 3.8+ installed. Follow these steps to clone the repository, install dependencies, and configure your environment.

1. Clone the repository

git clone https://github.com/AmadeusGB/alpha-arena.git
cd alpha-arena

2. Install dependencies

pip install -r requirements.txt

3. Configure API Keys

Alpha Arena requires API keys for OpenAI and Anthropic to power its decision engine. Use the provided template to create your .env file:

cp env.example .env

Then, edit the .env file and populate the following keys:

• OPENAI_API_KEY
• ANTHROPIC_API_KEY

# Clone and install
git clone https://github.com/AmadeusGB/alpha-arena.git
cd alpha-arena
pip install -r requirements.txt

# Setup environment
cp env.example .env
# Edit .env with your keys

# Run the application
python main.py

Once the environment is configured, execute the main entry point to start the market data retrieval and AI decision process.

python main.py

To enable detailed logging for debugging, set the DEBUG environment variable to 1 before running:

export DEBUG=1
python main.py

The following parameters govern the behavior of the Alpha Arena environment:

Alpha Arena uses python-dotenv for configuration. You must provide valid API keys in a .env file located in the project root.

Required keys:

• OPENAI_API_KEY: Your OpenAI API key for GPT-4 access.
• ANTHROPIC_API_KEY: Your Anthropic API key for Claude-3-Sonnet access.

If these are missing, the application will fail with an error similar to: ❌ OpenAI API密钥未设置，请设置OPENAI_API_KEY环境变量

# .env file content example
OPENAI_API_KEY=your_openai_api_key_here
ANTHROPIC_API_KEY=your_anthropic_api_key_here

The main.py file serves as the entrypoint for the Alpha Arena MVP, a system designed to compare AI trading decisions from different LLM providers. When executed, the orchestrator performs the following workflow:

1. Market Data Initialization: Initializes MarketData to fetch real-time prices.
2. LLM Adapter Initialization: Initializes adapters for different models (e.g., OpenAIAdapter, ClaudeAdapter).
3. Decision Generation: Uses DecisionMaker instances to generate trading decisions based on current market prices.
4. Comparison: Compares decisions from different models to identify consensus or divergence.

Ensure you have a .env file configured with the necessary API keys for your chosen providers before running.

python3 main.py

API Key Errors

Error: ❌ OpenAI API密钥未设置，请设置OPENAI_API_KEY环境变量 Solution: Verify that your .env file exists and contains the correct OPENAI_API_KEY.

Network Connection Issues

Error: ❌ 获取BTCUSDT价格失败: 网络请求失败 Solution: Check your internet connection and ensure you can reach the Bitget API endpoints.

Missing Dependencies

Error: ❌ 请安装openai: pip install openai Solution: Run the full dependency installation command:

pip install -r requirements.txt

To ensure comparability, all LLM agents receive a unified system prompt and must respond with a strictly validated JSON schema.

System Prompt Context: Agents are provided with Market Time (UTC), Account details (cash and positions), Universe (BTCUSDT, ETHUSDT), 60m of 1m-bars (OHLCV), Live Ticker data (bid, ask, mid, spread), and constraints (5m cadence, max 20% NAV exposure per trade, long only, max 1 open symbol).

Required JSON Output Schema: Every model must return a JSON object following this structure:

{
  "symbol": "BTCUSDT|ETHUSDT|null",
  "action": "BUY|SELL|HOLD",
  "position_size_pct": 0.0,
  "take_profit": 0.0,
  "stop_loss": 0.0,
  "confidence": 0.0,
  "rationale": "short text (<=200 chars)"
}