AI交易机器人

src/
├── polymarket/              # Pure API clients (domain-agnostic)
│   ├── gamma.py            # Gamma API (events/markets discovery)
│   ├── clob.py             # CLOB API (pricing/orderbook)
│   ├── models.py           # Pydantic models
│   └── utils.py            # Shared utilities
│
├── mcp/                    # MCP servers for Cursor integration
│   ├── base.py            # Base server with generic tools
│   └── cricket.py         # Cricket-specific server
│
├── bot/                   # Trading bot framework (domain-agnostic)
│   ├── base.py           # Base bot with main loop
│   ├── config.py         # Configuration loader
│   ├── database.py       # SQLite persistence
│   └── portfolio.py      # Portfolio/position management
│
├── domains/              # Domain-specific implementations
│   └── cricket/
│       ├── bot.py       # Cricket bot runner
│       ├── scanner.py   # Market scanner
│       └── prompts.py   # LLM prompts
│
└── strategies/           # Pluggable trading strategies
    ├── base.py          # Base strategy interface
    └── llm.py           # LLM-powered strategy

# Clone the repository
git clone https://github.com/artvandelay/polymarket-agents.git
cd polymarket-agents

# Create and activate virtual environment
python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Install dependencies
pip install mcp httpx pydantic websockets python-dotenv pyyaml aiosqlite

cp .env.example .env
# Edit .env and set OPENROUTER_API_KEY

trading:
  starting_capital: 1000
  interval_minutes: 5
  duration_hours: 2
  max_position_size_pct: 30

strategies:
  ai:
    model: anthropic/claude-3.5-sonnet
    temperature: 0.7
    min_confidence: 0.6

python -m src.domains.cricket.bot --duration 0.05 --interval 1

python -m src.domains.cricket.bot

tmux new-session -s trading-bot
python -m src.domains.cricket.bot
# Detach: Ctrl+B, then D
# Reattach: tmux attach -t trading-bot

# Start the server (typically called by Cursor)
polymarket-cricket

============================================================
Trading Cycle #1 @ 15:56:56
============================================================
Portfolio: $1,000.00 | Cash: $1,000.00 | Open: 0 positions | P&L: +0.00%

[SCAN] Found 2 active matches

[ANALYZE] T20 World Cup: England vs Nepal (Game 1)
    Volume: $202,000 | Liquidity: $79,000
    [DECISION] PASS
       Reasoning: England at 95.6¢ appears efficient. Potential 
       return (4.4%) doesn't justify the risk. Preserving capital 
       for markets with clearer edges.

[ANALYZE] T20 World Cup: Afghanistan vs New Zealand (Game 1)
    Volume: $437,000 | Liquidity: $73,000
    [DECISION] PASS
       Confidence: 90%
       Reasoning: 99.9¢ offers no realistic edge. Afghanistan at 
       0.1¢ is too risky. Best to avoid this market.

============================================================
End of Cycle #1
Portfolio Value: $1,000.00
Cash: $1,000.00
Open Positions: 0
Total P&L: +0.00%

Next scan at 15:57:56

timestamp INTEGER PRIMARY KEY
cash REAL
total_value REAL
pnl REAL
num_open_positions INTEGER

id INTEGER PRIMARY KEY
token_id TEXT           -- Polymarket token ID
match_slug TEXT         -- Match identifier
outcome TEXT            -- Team/outcome name
side TEXT               -- YES or NO
entry_price REAL        -- Entry price (0-1)
shares REAL             -- Number of shares
cost_basis REAL         -- Total cost
entry_time INTEGER      -- Unix timestamp
exit_price REAL         -- Exit price (if closed)
exit_time INTEGER       -- Exit timestamp
pnl REAL                -- Profit/loss
reasoning TEXT          -- AI reasoning
status TEXT             -- OPEN or CLOSED

id INTEGER PRIMARY KEY
timestamp INTEGER
cycle INTEGER           -- Cycle number
match_slug TEXT
action TEXT             -- BUY/SELL/HOLD/PASS
reasoning TEXT          -- AI reasoning
confidence REAL         -- 0-1
edge REAL               -- Estimated edge %
market_data TEXT        -- JSON snapshot

from src.strategies.base import BaseStrategy, TradeDecision

class MomentumStrategy(BaseStrategy):
    @property
    def name(self) -> str:
        return "Momentum Strategy"
    
    async def analyze(self, snapshot, portfolio, existing_position):
        momentum = self._calculate_momentum(snapshot)
        
        if momentum > self.config["threshold"]:
            return TradeDecision(
                action="BUY",
                outcome="Team with momentum",
                position_size=100,
                confidence=0.7,
                reasoning=f"Strong momentum: +{momentum}% in last hour"
            )
        
        return TradeDecision(action="PASS", reasoning="No momentum")