ML Worker for object detection at scale. Pulls tasks from Pub/Sub, reads images from GCS, runs RFDETR, stores results to GCS, and notifies an internal API.

• Consume tasks from Pub/Sub subscription
• Download image from GCS (image_path)
• Run detection with RFDETR
• Store JSON results to results/<task_id>/detection_results.json in GCS
• POST result summary to internal API

• Pub/Sub driven worker
• GCS I/O (images, results)
• Internal API callback
• Docker + Kubernetes ready

• Python 3.11+
• Google Cloud: Pub/Sub + Storage
• GCS bucket

1. Clone and setup:

   git clone <repository-url>
   cd object-detection-worker

2. Install dependencies:

   pip install -r requirements.txt

3. Configure environment:

   cp .env.example .env
   # Edit .env with your configuration

4. Run tests (optional):

   pytest -q

5. Run the worker:

   python -m src.main

# Build image
docker build -t object-detection-worker .

# Run container
docker run --env-file .env object-detection-worker

# Apply all manifests
kubectl apply -f k8s/

Publish a message to Pub/Sub with:

{
  "task_id": "550e8400-e29b-41d4-a716-446655440000",
  "image_path": "images/photo.jpg"
}

Result JSON stored in GCS:

{
  "task_id": "550e8400-e29b-41d4-a716-446655440000",
  "detections": [
    {
      "class_id": 0,
      "class_name": "person",
      "confidence": 0.85,
      "bbox": {
        "x1": 100.0,
        "y1": 150.0,
        "x2": 200.0,
        "y2": 300.0
      }
    }
  ],
  "processed_at": "2024-01-01T12:00:00Z",
  "processing_time_ms": 1500
}

src/
├── main.py
├── domain/
│   ├── entities/
│   └── repositories/
└── infrastructure/
    ├── config.py
    ├── models/
    ├── repositories/
    └── services/

pytest -q

GitHub Actions workflow .github/workflows/deploy.yml:

• Job 1: run tests → pytest tests/ --verbose
• Job 2: build & push Docker to GAR
• Apply K8s manifests with updated image

K8s HPA manifests included under k8s/. Tune based on Pub/Sub queue depth and resource usage.

INFO-level logs for task start/end, errors, and processing time.