Nephyr Weather

Weather signal intelligence for prediction markets.

Fetches 31-member GFS ensemble forecasts from Open-Meteo and compares model probabilities against live market prices on Polymarket and Kalshi to detect tradeable edges.

Install

pip install nephyr-weather          # core + httpx
pip install "nephyr-weather[mcp]"   # + MCP server
pip install "nephyr-weather[api]"   # + FastAPI REST server
pip install "nephyr-weather[all]"   # everything

Quick start

import asyncio
from nephyr_weather import OpenMeteoClient, signal_from_forecast, MarketInfo, detect_edge

async def main():
    # 1. Fetch GFS ensemble forecast
    async with OpenMeteoClient() as client:
        forecast = await client.fetch_ensemble("NYC", "2026-03-29")

    # 2. Generate probability signal
    signal = signal_from_forecast(forecast, threshold_f=75.0, direction="above")
    print(f"Model P(NYC high > 75°F): {signal.model_probability:.1%}")

    # 3. Compare to market price
    market = MarketInfo(
        platform="kalshi",
        ticker="KXHIGHNY-26MAR29-T75",
        city="NYC",
        threshold_f=75.0,
        direction="above",
        market_price=0.55,
    )
    result = detect_edge(signal, market)
    print(f"Edge: {result.edge:+.1%}, EV: {result.ev:.4f}, Tradeable: {result.is_tradeable}")

asyncio.run(main())

Supported cities

NYC, Chicago, Miami, Austin, LA, Denver, Boston, London, Seoul, Shanghai, Hong Kong, Atlanta

Modules

MCP server

nephyr-weather-mcp

Tools: get_weather_signals, get_ensemble_forecast, detect_edges, list_active_markets, get_city_info

REST API

uvicorn api.app:app --reload

Endpoints: POST /v1/signals, POST /v1/forecast, POST /v1/edges, GET /v1/markets, GET /v1/cities, GET /v1/health

Pricing

License

MIT — CLM Studios