fix(p1): scheduler deprecation, ML persistence, CRISP vintage, forecast staleness #71
Merged
wasabi opened 3 days ago

Changes

Scheduler (WP4a)

  • Deprecate crontab entries covered by DB scheduler (keep health-alert.sh only)
  • Fix climate data refresh: remove hardcoded location-id, use --all

ML Anomaly Detection (WP4b)

  • Add load_models() to load cached Prophet/Isolation Forest models from disk
  • run_ml_check() loads cached models before fitting fresh (avoids retraining every invocation)
  • Models staleness check: skip if older than 7 days

CRISP Scoring (WP4c)

  • Activate vintage_year in financial risk: 2% risk increase per year of age, capped at 15%
  • vintage_year now stored in factors dict for audit trail

Forecast Engine (WP4d-e)

  • Add data freshness checks: warns when input tables exceed staleness thresholds
  • Staleness notes included in forecast output
  • Expand price projection from 4 to 8 crops (add coffee, avocado, tomato, banana)
Commits were merged into target branch
  • fix(p1): scheduler deprecation, ML persistence, CRISP vintage, forecast staleness
    - Deprecate crontab entries covered by DB scheduler (keep health-alert.sh only)
    - Fix climate data refresh: remove hardcoded location-id, use --all
    - Add load_models() to ML anomaly detector; run_ml_check() loads cached models
      before fitting fresh, avoiding expensive retraining on every invocation
    - Activate vintage_year in CRISP financial risk: 2%/year age penalty, capped at 15%
    - Expand forecast price projection from 4 to 8 crops (add coffee, avocado,
      tomato, banana with reasonable defaults)
    - Add data freshness checks to forecast engine: warns when input tables
      (price_observation, crop_cycle, expense_event, harvest_event, sensor_reading)
      exceed staleness thresholds; notes included in forecast output
    3 days ago
  • wasabi merged 3 days ago
1/1
Submitter wasabi
Target main
Source feat/p1-forecast-crud
Jobs
Merge Strategy
Create Merge Commit
Watchers (1)
Reference
pull request KI-71
Please wait...
Connection lost or session expired, reload to recover
Page is in error, reload to recover