How upstairs-aura/drill_rig_backend works
This is a FastAPI backend for drill-rig predictive maintenance. Sensor feature batches enter through the ingest API and are persisted; asset dashboard endpoints expose current metrics, threshold alerts, health history, and system status. Prediction requests load the latest reading, run available forest-based inference, and record results for history. Separate training scripts build supervised forest, unsupervised isolation-forest, and sequence-based LSTM models. The README is empty, so capability coverage is based on the source index and excerpts; the LSTM trainer is not shown wired into runtime inference.
Main entry points: database.py DB sessions · dashboard.py Asset dashboard · ingest.py Sensor ingestion · features.py Feature conversion
Components 17
API and dashboard
- FastAPI application
app/main.py - Sensor ingestion
app/routes/ingest.py - Ingest validation
app/schemas.py - Asset dashboard
app/routes/dashboard.py - Metrics and alerts
app/routes/dashboard.py - Health and status
app/routes/dashboard.py - Prediction endpoints
app/routes/dashboard.py - Prediction history
app/routes/dashboard.py
Persistence
- DB sessions
app/database.py - Maintenance records
app/models.py - Relational database
Inference
- Model inference
app/ml/predict.py - Feature conversion
app/ml/features.py
Model training
- Random forest training
app/ml/train_forest.py - Isolation forest training
app/ml/train_isolation.py - LSTM training
app/ml/train_lstm.py
Other components
- API client
Connections 26
- API client → FastAPI application: calls API
- FastAPI application → Sensor ingestion: mounts router
app/main.py - FastAPI application → Asset dashboard: mounts router
app/main.py - API client → Sensor ingestion: submits batch
app/routes/ingest.py - Sensor ingestion → Ingest validation: validates payload
app/routes/ingest.py - Sensor ingestion → Maintenance records: creates records
app/routes/ingest.py - Sensor ingestion → DB sessions: gets session
app/routes/ingest.py - Asset dashboard → Metrics and alerts: serves metrics
app/routes/dashboard.py - Asset dashboard → Health and status: serves health
app/routes/dashboard.py - Asset dashboard → Prediction endpoints: serves predictions
app/routes/dashboard.py - Asset dashboard → Prediction history: serves history
app/routes/dashboard.py - Metrics and alerts → DB sessions: queries readings
app/routes/dashboard.py - Health and status → DB sessions: queries readings
app/routes/dashboard.py - Prediction endpoints → DB sessions: loads latest
app/routes/dashboard.py - Prediction endpoints → Model inference: runs inference
app/routes/dashboard.py - Prediction endpoints → DB sessions: saves result
app/routes/dashboard.py - Prediction history → DB sessions: queries results
app/routes/dashboard.py - Maintenance records → DB sessions: uses metadata
app/models.py - DB sessions → Relational database: connects sessions
- Model inference → Feature conversion: converts record
app/ml/predict.py - Random forest training → DB sessions: loads training data
app/ml/train_forest.py - Random forest training → Feature conversion: builds features
app/ml/train_forest.py - Isolation forest training → DB sessions: loads records
app/ml/train_isolation.py - Isolation forest training → Feature conversion: builds features
app/ml/train_isolation.py - LSTM training → DB sessions: loads records
app/ml/train_lstm.py - LSTM training → Feature conversion: uses feature columns
app/ml/train_lstm.py