live-trade-bench 0.0.1b0

Live evaluation of trading agents

Live Evaluation of Trading Agents

Overview

Trading agent evaluation in the live environment. We target at avoiding overfitting on back test and build an arena for LLM-based trading agents.

Features

• AI Agents: GPT-4 powered trading decisions
• Multi-Asset: Stocks and prediction markets
• Real-time Data: Live market feeds
• Portfolio Management: Automated tracking and execution

Quick Start

# Install
poetry install

# Stock trading
from live_trade_bench import LLMStockAgent, create_stock_account
agent = LLMStockAgent("Trader")
account = create_stock_account(10000.0)

# Prediction markets
from live_trade_bench import LLMPolyMarketAgent, fetch_trending_markets
agent = LLMPolyMarketAgent("Predictor")
markets = fetch_trending_markets(5)

Structure

live_trade_bench/
├── agents/                     # AI trading agents
│   ├── base_agent.py          # Base LLM agent class
│   ├── stock_agent.py         # Stock trading agent
│   ├── polymarket_agent.py    # Prediction market agent
│   ├── stock_system.py        # Stock trading system
│   └── polymarket_system.py   # Polymarket trading system
├── accounts/                   # Portfolio management
│   ├── base_account.py        # Base account class
│   ├── stock_account.py       # Stock portfolio & execution
│   ├── polymarket_account.py  # Prediction market portfolio
│   ├── action.py              # Trading action definitions
│   └── utils.py               # Account utilities
├── fetchers/                   # Real-time data sources
│   ├── base_fetcher.py        # Base fetcher class
│   ├── stock_fetcher.py       # Yahoo Finance integration
│   ├── polymarket_fetcher.py  # Polymarket API
│   ├── news_fetcher.py        # Financial news
│   ├── option_fetcher.py      # Options data
│   └── reddit_fetcher.py      # Social sentiment
└── utils/                      # LLM & utilities
    ├── llm_client.py          # LLM integration
    └── logger.py              # Logging utilities

Examples

See examples/ directory for demo scripts.

License

MIT