How gandharv2323/hydrox-tesseract-26 works
HydroX is a pump-monitoring digital twin: sensor readings are simulated or replayed, processed into live health/anomaly/fault/RUL state, then exposed through REST and WebSocket interfaces. Its advanced inference path validates readings, accumulates sliding windows, derives hybrid features, and combines anomaly, fault, and prognostic estimates with calibration and consistency logic. Training, data preparation, and evaluation are important supporting product capabilities. Some runtime internals and API wiring are documented but not fully excerpted; uncertain links are omitted.
Main entry points: train_all.py Model Training · inference_engine.py Inference Orchestrator · main.py API And Broadcast · representation.py Hybrid Representation
Components 26
Live Monitoring
- Sensor Simulator
sensor_sim.py - Legacy Detector
ml_model.py - Pump Health Metrics
health_engine.py - API And Broadcast
main.py - API Server Wrapper
api/server.py
Advanced Inference
- Sliding Window
pipeline/buffer.py - Window Features
pipeline/features.py - Hybrid Representation
pipeline/representation.py - Inference Orchestrator
pipeline/inference_engine.py
Models And Calibration
- Fault Classifier
pipeline/fault_classifier.py - RUL Predictor
pipeline/rul_lstm.py - Shared Latent Encoder
models/shared_latent.py - Threshold Calibration
calibration/threshold.py - Learned Score Fusion
calibration/fusion_meta.py - Fallback Score Fusion
calibration/score_fusion.py
Data And Training
- Synthetic Data Generator
training/generate_data.py - Real Data Loader
data_pipeline/loader.py - Data Preprocessing
data_pipeline/preprocessing.py - Dataset Loader
dataset_loader.py - Model Training
training/train_all.py
Evaluation
- Model Metrics
evaluation/pipeline.py - Evaluation Reports
evaluation/report.py - Scenario Evaluation
evaluation/scenario_runner.py
Other components
- Monitoring Client
- Dashboard Consumer
- Unity Consumer
Connections 33
- Monitoring Client → API And Broadcast: requests state
README.md - Sensor Simulator → API And Broadcast: supplies readings
README.md - API And Broadcast → Legacy Detector: runs detector
README.md - API And Broadcast → Pump Health Metrics: computes metrics
README.md - API And Broadcast → Inference Orchestrator: runs inference
README.md - API And Broadcast → Dashboard Consumer: streams state
README.md - API And Broadcast → Unity Consumer: serves payload
README.md - API Server Wrapper → API And Broadcast: imports
api/server.py - Inference Orchestrator → Sliding Window: pushes readings
pipeline/inference_engine.py - Inference Orchestrator → Hybrid Representation: builds features
pipeline/inference_engine.py - Hybrid Representation → Window Features: extracts descriptors
pipeline/representation.py - Hybrid Representation → Shared Latent Encoder: encodes windows
pipeline/representation.py - Inference Orchestrator → Fault Classifier: classifies faults
pipeline/inference_engine.py - Inference Orchestrator → RUL Predictor: predicts RUL
pipeline/inference_engine.py - Inference Orchestrator → Threshold Calibration: loads threshold
pipeline/inference_engine.py - Inference Orchestrator → Learned Score Fusion: fuses scores
pipeline/inference_engine.py - Inference Orchestrator → Fallback Score Fusion: uses fallback
pipeline/inference_engine.py - Synthetic Data Generator → Sliding Window: builds windows
training/generate_data.py - Synthetic Data Generator → Window Features: extracts features
training/generate_data.py - Real Data Loader → Data Preprocessing: prepares data
- Dataset Loader → Sensor Simulator: uses simulation
dataset_loader.py - Model Training → Synthetic Data Generator: generates samples
training/train_all.py - Model Training → Real Data Loader: loads real data
training/train_all.py - Model Training → Data Preprocessing: prepares real data
training/train_all.py - Model Training → Hybrid Representation: builds training features
training/train_all.py - Model Training → Fault Classifier: trains classifier
training/train_all.py - Model Training → RUL Predictor: trains predictor
training/train_all.py - Model Training → Shared Latent Encoder: trains encoder
training/train_all.py - Model Training → Threshold Calibration: calibrates thresholds
training/train_all.py - Model Training → Learned Score Fusion: trains fusion
training/train_all.py - Evaluation Reports → Model Metrics: computes metrics
- Scenario Evaluation → Inference Orchestrator: runs scenarios
evaluation/scenario_runner.py - Scenario Evaluation → Sensor Simulator: drives scenarios
evaluation/scenario_runner.py