dhanush-10-1/spdrha

How dhanush-10-1/spdrha works

SPDRHA is a road-hazard detection product: a phone gathers motion and GPS data, classifies sensor windows on-device, stores non-normal detections, and synchronizes them to a FastAPI backend. The backend also exposes anomaly and health APIs and has a separate model-training workflow. The README identifies the root app as active; sampled runtime excerpts are chiefly from the parallel irads-mobile app, so unsampled root-app and backend internals are represented only where the README or SOURCE INDEX supports them.

Main entry points: aiService.ts Inference and policy · main.py FastAPI application · anomalies.py Anomaly API · syncService.ts Upload and retry

Components 27

Mobile collection

Active mobile app App.tsx
Motion and GPS capture src/sensors/SensorManager.ts
Sensor reading types src/sensors/types.ts

On-device detection

Window builder src/ai/windowManager.ts
Inference and policy src/ai/aiService.ts
TFLite inference src/ai/tflite_inference.ts
Input normalization src/ai/normalization.ts
Inference lifecycle src/ai/inferenceLifecycle.ts
Persistence policy src/ai/anomalyPersistencePolicy.ts

Persistence and sync

Anomaly event store src/storage/database.ts
Device identity src/storage/deviceIdentity.ts
Upload and retry src/storage/syncService.ts

Backend services

FastAPI application irads-backend/backend/main.py
Anomaly API irads-backend/backend/routers/anomalies.py
Health API irads-backend/backend/routers/health.py
Development auth irads-backend/backend/routers/dev_auth.py
Backend database connection irads-backend/backend/db/connection.py
Kafka producer irads-backend/backend/kafka/producer.py
Anomaly consumer irads-backend/consumer/anomaly_consumer.py
Consumer DB writer irads-backend/consumer/db_writer.py

Model development

Training preprocessing irads-backend/ml/preprocessing
Training dataset irads-backend/ml/training/dataset.py
Model training irads-backend/ml/training/train.py
Training model irads-backend/ml/training/model.py

Other components

Phone user
Phone sensors and GPS
Backend HTTP API

Connections 26

Phone user → Active mobile app: uses README.md
Phone sensors and GPS → Motion and GPS capture: provides readings README.md
Motion and GPS capture → Active mobile app: collects data for App.tsx
Active mobile app → Inference and policy: submits readings
Inference and policy → Window builder: builds windows src/ai/aiService.ts
Inference and policy → TFLite inference: predicts with src/ai/aiService.ts
TFLite inference → Input normalization: normalizes input src/ai/tflite_inference.ts
Inference and policy → Inference lifecycle: manages inference src/ai/aiService.ts
Inference and policy → Persistence policy: checks persistence src/ai/aiService.ts
Inference and policy → Anomaly event store: saves hazards src/ai/aiService.ts
Inference and policy → Device identity: gets device ID src/ai/aiService.ts
Inference and policy → Upload and retry: starts sync src/ai/aiService.ts
Upload and retry → Anomaly event store: reads and marks src/storage/syncService.ts
Upload and retry → Backend HTTP API: posts events src/storage/syncService.ts
Backend HTTP API → Anomaly API: sends events README.md
FastAPI application → Anomaly API: registers router irads-backend/backend/main.py
FastAPI application → Health API: registers router irads-backend/backend/main.py
FastAPI application → Development auth: registers router irads-backend/backend/main.py
FastAPI application → Backend database connection: opens and closes irads-backend/backend/main.py
FastAPI application → Kafka producer: starts and stops irads-backend/backend/main.py
Anomaly API → Backend database connection: accesses database irads-backend/backend/routers/anomalies.py
Anomaly API → Kafka producer: uses producer irads-backend/backend/routers/anomalies.py
Anomaly consumer → Consumer DB writer: delegates writes
Training preprocessing → Training dataset: prepares data for irads-backend/ml/training/dataset.py
Model training → Training dataset: uses dataset irads-backend/ml/training/train.py
Model training → Training model: uses model irads-backend/ml/training/train.py

More TypeScript diagrams: flowr-analysis/flowr · googlecloudplatform/docker-clamav-malware-scanner · igorkowalczyk/igorkowalczyk.dev · krish120003/zercel · libre-webui/libre-webui

Drawn by GitDiagram on October 26, 2026 from the repository's file tree, README and source files. Read as Markdown ·