A desktop notifier for new runs

Weather models publish on a schedule. This little daemon polls your slices, watches each one's last_run_time, and fires a native desktop notification when a fresh run lands — so you know to re-download.

The script

It polls /v1/slices every 10 minutes, remembers the run it last announced per slice, and notifies via osascript on macOS or notify-send on Linux. The first pass just primes the state (no spurious alerts on startup).

#!/usr/bin/env python3
"""Pop a desktop notification when any of your slices gets a new model run."""
import os, sys, time, subprocess, requests

S = requests.Session()
S.headers["Authorization"] = f"Bearer {os.environ['WF_TOKEN']}"
URL = "https://api.weatherfiles.com/v1/slices"
POLL_SECONDS = 600

def notify(title, msg):
    if sys.platform == "darwin":
        subprocess.run(["osascript", "-e",
                        f'display notification "{msg}" with title "{title}"'])
    else:  # Linux / BSD with libnotify
        subprocess.run(["notify-send", title, msg])

seen = {}  # token -> last_run_time we've already announced
while True:
    try:
        for sl in S.get(URL, timeout=30).json():
            tok, run = sl["token"], sl["last_run_time"]
            name = sl["label"] or sl["model_id"]
            if tok in seen and run and run != seen[tok]:
                notify("WeatherFiles", f"New run for {name}: {run}")
            seen[tok] = run            # first pass primes; later changes notify
    except requests.RequestException:
        pass                           # transient - try again next poll
    time.sleep(POLL_SECONDS)

Run it

• Linux: ensure libnotify (notify-send) is installed; run it from your session, or as a user systemd service.
• macOS: no dependency — osascript is built in.
• Polling every 10 minutes is gentle on the API; tighten only if you need to.

Want to act on the notification automatically? Swap notify() for a call into the slice client's download() to fetch the new run on the spot.