2.3 Miner (Inference node)

• Service name: px-miner

• Tech stack: FastAPI + ONNXRuntime (CPU)

• Port: 6060+

• Docker image: ghcr.io/privacyx-org/px-miner

Responsibilities

• Runs local inference using ONNX models (default: ResNet50 v2).

• Processes base64-encoded images or direct URLs.

• Returns structured detections and class probabilities.

Model implementation

The miner supports two backends:

Mode	Description
stub	Lightweight testing mode (no model required).
onnx	Real ONNX inference using ResNet50 v2 (CPU).

File structure

services/miner/
├─ app/api.py            # REST endpoints (FastAPI)
├─ impl_onnx.py          # ONNX-based inference
├─ models/
│  ├─ detector.onnx
│  └─ imagenet_classes.txt

Endpoints

Endpoint	Method	Description
/health	GET	Returns miner health & model metadata
/info	GET	Returns ONNX provider & env
/detect/image	POST	Executes ONNX inference
/infer/image	POST	Legacy endpoint for scheduler
/detect/video	POST	(Stub) Returns sample classification