Smart Document ScannerVision + OCR Workbench
Checking API
Reference

API and demo reference

What the hosted demo actually does, and the REST contract the full pipeline exposes.

How demo mode works

The demo ships as a static export on Vercel. There is no Python, OpenCV, or ONNX runtime on the server, so nothing is processed remotely.

  • Your file stays in the browser. It is never uploaded.
  • Results come from a stored sample run of the real pipeline.
  • Six sample records load on the first visit.
  • New records are saved to this browser's localStorage.

To run the real pipeline, build the container with docker build -t smart-document-scanner . && docker run -p 8000:8000 smart-document-scanner and open http://localhost:8000. The same image runs on Railway, or locally with the Python entrypoint.

REST API

Five endpoints cover processing, job polling, history, and health.

Endpoint reference
MethodPathDescription
POST/api/processAccept multipart fields file, doc_type, fast_denoise, and cpp_acceleration, then queue a job.
GET/api/jobs/{id}Return job status, stage progress, and the full result once processing completes.
DELETE/api/jobs/{id}Remove a job and its stored artifacts from the server history.
GET/api/historyList recent jobs, newest first, for the history screen.
GET/api/healthReport service health plus the OCR engine and accelerator in use.

Example request

Send the image as multipart form data. Fast denoise and C++ acceleration are optional booleans.

POST /api/process
curl -X POST http://localhost:8000/api/process \
  -F "file=@document.jpg" \
  -F "doc_type=auto" \
  -F "fast_denoise=true" \
  -F "cpp_acceleration=true"

Response shape

Every job result carries detection facts, extraction coverage, parsed fields, and artifact URLs.

Job result (JSON, abbreviated)
{
  "job_id": "9f2c1e70",
  "document_detected": true,
  "rotation_angle": 1.4,
  "detection_area_ratio": 0.62,
  "detection_method": "contour",
  "detection_score": 0.91,
  "detection_selection": "largest_quad",
  "corners": [[82, 64], [1518, 71], [1509, 1562], [75, 1551]],
  "image_width": 1600,
  "image_height": 1200,
  "processing_time_ms": 1840,
  "ocr_confidence": 0.88,
  "extracted_field_count": 4,
  "recognised_field_count": 4,
  "extraction_coverage": 1.0,
  "fields": {
    "document_type": "receipt",
    "classification": { "document_type": "receipt", "score": 0.93, "signals": ["total"] },
    "fields": [
      { "name": "merchant", "value": "Kopi Senja", "confidence": 0.9, "source": "line" }
    ]
  },
  "raw_text": "KOPI SENJA\nTotal 48.000",
  "warnings": [],
  "assets": {
    "original": "/api/jobs/9f2c1e70/assets/original.jpg",
    "detected": "/api/jobs/9f2c1e70/assets/detected.jpg",
    "corrected": "/api/jobs/9f2c1e70/assets/corrected.jpg",
    "enhanced": "/api/jobs/9f2c1e70/assets/enhanced.jpg"
  }
}
Open scannerView history