AI-driven weather intelligence for agricultural commodity futures, predicting price moves 48-72 hours early. It targets traders and researchers by analyzing satellite weather data and historical patterns, providing probabilistic price impact signals for six key global markets. The system leverages advanced AI and diverse data sources to deliver early insights, potentially enhancing trading strategies.

How It Works

The system ingests real-time NOAA forecasts, satellite imagery (GOES, Sentinel-2), and vegetation indices. Claude Sonnet 4 analyzes this data alongside 40 years of historical price patterns. A main orchestrator, operating on a 15-minute cycle, uses eight ML models, including an ensemble and an LSTM with attention, to generate probabilistic price impact signals. This approach enables predictions significantly earlier than official reports, identifying opportunities based on weather events like freezes, droughts, and hurricanes.

Quick Start & Requirements

Install dependencies via pip install -r requirements.txt. Set ANTHROPIC_API_KEY environment variable; NOAA_API_KEY and PLANET_API_KEY are optional. A PostgreSQL database is required. Run the demo with python demo.py or the full system with python main.py. Backtesting is available via python backtest.py.

Highlighted Details

• Backtest Performance (2023-2025): Achieved +223% total return (2.5x leverage), 70% win rate, 2.6:1 win/loss, 1.8 Sharpe ratio, -21% max drawdown.
• Markets Covered: Orange Juice (FCOJ-A), Coffee (KC), Cocoa (CC), Sugar #11 (SB), Corn (ZC), and Wheat (ZW).
• System Architecture: Features a Weather Collector, Claude Sonnet 4, PostgreSQL, and an orchestrator running eight ML models.
• Data Sources: NOAA, GOES/Sentinel-2/MODIS satellite imagery, INMET Brazil, Ghana Met Agency, CME/ICE futures, USDA reports.

Maintenance & Community

Project website: agriquant.ai. Demo: demo.agriquant.ai. Contact: hello@agriquant.ai. X: @AgriQuant_AI.

Licensing & Compatibility

The README does not specify a software license, requiring clarification for commercial use or integration.

Limitations & Caveats

Backtested performance assumes ideal trading conditions (perfect fills, no slippage, hindsight) and is amplified by leverage. Early signals (>72 hours) and trades held through USDA report releases were less reliable. Minor weather events may not trigger significant price impacts. This is not financial advice.