Cache System

How It Works

APScheduler (60 min)
    └── run_compute(force=False)
        ├── fetch weather + air quality data
        ├── build_ensemble_forecast_df(days=7)
        │       → src/airsafenet_ai/data/forecast_cache.csv
        ├── history_range_payload(days=30)
        │       → src/airsafenet_ai/data/history_cache.csv
        ├── get_ensemble_current_snapshot(profile)
        │       → src/airsafenet_ai/data/current_cache.json
        ├── compute_district_heatmap()
        │       → src/airsafenet_ai/data/district_cache.csv
        └── check_anomaly()
                → src/airsafenet_ai/data/anomaly_log.json

Cache Freshness

The cache is considered fresh for 70 minutes after the last compute. If stale, endpoints still return the last cached data with a cache_stale: true flag.

Manual Trigger

# Via API (requires admin key)
curl -X POST http://localhost:8000/admin/compute \
  -H "X-Admin-Key: your_admin_key"

# Or via .NET backend
curl -X POST http://localhost:7276/api/admin/compute \
  -H "Authorization: Bearer your_jwt_token"

Cache Files

File	Content	Updated
forecast_cache.csv	7-day hourly forecast × 4 profiles	Every 60 min
history_cache.csv	30-day hourly history × 4 profiles	Every 60 min
current_cache.json	Current snapshot × 4 profiles	Every 60 min
district_cache.csv	22-district current AQI	Every 60 min
anomaly_log.json	Spike detection events	Per event