sorcerer-forecasts 0.1.3

Add your description here

Sorcerer Forecasts

A Python library for fetching and processing weather forecast data with automatic caching and region selection. Designed for simulations and applications that need efficient access to spatio-temporal forecast data.

Features

• Automatic Forecast Management: The service automatically determines when to fetch new forecast data based on your query location and time
• Local Caching: Downloaded forecasts are cached locally as NetCDF files for fast repeated access
• Smart Region Selection: Automatically selects the appropriate forecast region based on query coordinates
• 4D Querying: Query forecasts at specific points in space (latitude, longitude, altitude) and time
• Simulation-Friendly: Ideal for running simulations - no need to manually check if new data is needed

Installation

pip install sorcerer-forecasts

Quick Start

from datetime import datetime
from sorcerer_forecasts import ForecastService
from sorcerer_forecasts.sources import Stratocast

# Initialize the forecast source with your API key
source = Stratocast(api_key='YOUR_API_KEY')

# Create the forecast service with local caching
forecast_service = ForecastService(source=source, cache_dir='./.cache')

# Query forecast data at a specific 4D point
forecast = forecast_service.get({
    'time': datetime.fromisoformat('2025-08-26T00:00:00Z'),
    'latitude': 40,
    'longitude': 30,
    'altitude': 14625  # meters
})

# Access forecast variables
if forecast:
    print(f"Pressure: {forecast['pres']}")
    print(f"U wind: {forecast['u']}")
    print(f"V wind: {forecast['v']}")
    print(f"Height: {forecast['h']}")

How It Works

Automatic Forecast Management

The ForecastService intelligently manages forecast data:

1. First Request: When you query a point, the service checks if it has the relevant forecast in memory
2. Cache Check: If not in memory, it checks the local cache directory for a saved NetCDF file
3. Fetch if Needed: Only fetches from the remote source if the data isn't available locally
4. Reuse Loaded Data: Subsequent queries within the same forecast region and time period use the already-loaded data

This makes it perfect for simulations where a vehicle might be moving through space and time - the service will:

• Reuse the same forecast data when the vehicle moves within one time step
• Automatically fetch new forecasts only when crossing into a new time period or region
• Keep multiple forecasts in memory when needed

Example: Running a Simulation

from time import time
from datetime import datetime, timedelta

source = Stratocast(api_key='API_KEY')
forecast_service = ForecastService(source=source, cache_dir='./.cache')

# Simulate 10 time steps
base_time = datetime.fromisoformat('2025-08-26T00:00:00Z')

for i in range(10):
    start = time()

    # Query moves slightly in space and time
    forecast = forecast_service.get({
        'time': base_time + timedelta(minutes=i*15),
        'latitude': 40 + i * 0.1,
        'longitude': 30 + i * 0.1,
        'altitude': 14625
    })

    end = time()
    print(f"Step {i}: {end - start:.3f}s")
    # First query will be slower (fetching), subsequent queries within
    # the same forecast will be nearly instant

Download a Forecast Region

Sometimes you need the full forecast region (xarray Dataset) for a given time and location—for example to analyze offline or save to disk. Use get_region:

from datetime import datetime
from sorcerer_forecasts import ForecastService
from sorcerer_forecasts.sources import Stratocast

source = Stratocast(api_key='YOUR_API_KEY')
forecast_service = ForecastService(source=source, cache_dir='./.cache')

region = forecast_service.get_region({
    'time': datetime.fromisoformat('2025-09-12T00:00:00Z'),
    'latitude': 37.77,
    'longitude': -122.39,
    'altitude': 22
})

if region:
    # Save the full dataset wherever you like
    region['dataset'].to_netcdf(f"./tmp/{region['forecast_id']}.nc")

Notes:

• The location determines which forecast region is fetched; the service downloads the region that contains the point.
• If you don't have access/permissions for that region, get_region returns None.
• If a matching cached NetCDF exists, it is loaded from disk.
• Otherwise the dataset is fetched and, if cache_dir is set, written to the cache.

Supported Regions

The library automatically selects the appropriate forecast region based on your coordinates (requires proper permissions):

• CONUS: Continental United States
• EU-Central: Central Europe
• AF-East: Eastern Africa
• Region4: Extended North America
• Global: Worldwide coverage

Caching

Cached forecasts are stored as NetCDF files in the specified cache directory. The cache structure is:

cache_dir/
  - YYYYMMDD.tHHz.stratocast.0p25.ml80.wind.{region}.nc

Cache files persist between sessions, so restarting your application won't require re-downloading previously fetched forecasts.

API Reference

ForecastService

ForecastService(source: ForecastSource, cache_dir: str | None = None)

• source: A forecast source implementation (e.g., Stratocast)
• cache_dir: Directory for caching forecast files (optional)

Methods

• get(location: Point4) -> ForecastData | None: Retrieve forecast data at a 4D point
  • Returns None if the location is outside available forecast bounds
• get_region(location: Point4) -> ForecastRegion | None: Retrieve the full forecast dataset and its identifier
  • Region is chosen based on the provided location (point-in-region lookup)
  • Returns None if the dataset cannot be fetched
  • Returns None if you don't have access/permissions for that region
  • Uses cache when available; writes to cache when cache_dir is set

Point4 Dictionary Structure

{
    'time': datetime,      # UTC datetime
    'latitude': float,     # Degrees
    'longitude': float,    # Degrees
    'altitude': float      # Meters
}

ForecastRegion Dictionary Structure

{
    'forecast_id': str,        # Cache-friendly identifier used in filenames
    'dataset': xarray.Dataset   # Full forecast region dataset
}

License

See LICENSE file for details.