How mr-saab-29/causal-energy-intelligence works
The repository is a carbon-aware energy forecasting and workload-scheduling platform. Its principal workflow refreshes energy and weather data, derives modeling features, forecasts energy signals, estimates carbon intensity, ranks cleaner scheduling hours, and exposes dashboard recommendations. It also documents future recommendations, causal analysis, and operational monitoring. The README supplies the broad pipeline; sampled source confirms ingestion, dashboard orchestration, and monitoring details, so uncertain internal calls are not inferred.
Main entry points: models Forecast models · features Feature engineering · definitions.py Refresh orchestration · workload_shift.py Clean-hour ranking
Components 19
Data foundation
- Local ingestion
src/data/local_ingest.py - Supabase ingestion
src/data/supabase_ingest.py - Transformed data store
src/data/pipelines - Feature engineering
src/features
Forecasting
- Forecast models
src/models
Scheduling decisions
- Carbon estimates
src/carbon - Clean-hour ranking
src/optimization/workload_shift.py - Causal analysis
src/causal - Causal recommendations
src/causal/recommendations.py - Future recommendations
src/models/future_recommendations.py
Product delivery
- Dashboard data contract
scripts/build_dashboard_data.py - Decision dashboard
frontend/src/main.jsx - Platform API
api/main.py
Operations
- Refresh orchestration
orchestration/definitions.py - Pipeline health
src/data/pipeline_health.py - Forecast monitoring
src/monitoring/forecast_monitor.py - MLflow tracking
src/tracking/mlflow_tracking.py
Other components
- Energy and weather sources
- Dashboard user
Connections 24
- Energy and weather sources → Local ingestion: fetches
- Energy and weather sources → Supabase ingestion: fetches
- Local ingestion → Feature engineering: rebuilds
- Supabase ingestion → Transformed data store: loads
- Supabase ingestion → Feature engineering: exports
- Transformed data store → Feature engineering: feeds
- Feature engineering → Forecast models: trains and predicts
- Forecast models → Carbon estimates: supplies forecasts
- Carbon estimates → Clean-hour ranking: informs ranking
- Clean-hour ranking → Causal recommendations: provides rankings
- Causal analysis → Causal recommendations: supports analysis
- Forecast models → Future recommendations: supports recommendations
- Clean-hour ranking → Dashboard data contract: supplies results
- Future recommendations → Dashboard data contract: supplies results
- Dashboard data contract → Decision dashboard: serves data
- Dashboard user → Decision dashboard: views
- Refresh orchestration → Local ingestion: dispatches refresh
- Refresh orchestration → Pipeline health: runs checks
- Refresh orchestration → Future recommendations: dispatches generation
- Refresh orchestration → Dashboard data contract: dispatches build
- Feature engineering → Pipeline health: validated by
- Forecast models → Forecast monitoring: quality monitored
- Clean-hour ranking → Forecast monitoring: drift monitored
- Forecast models → MLflow tracking: tracked by