Extraction Options

Customize how PROOF extracts tables from your files.

Model selection

Choose between basic and premium processing models:

Model	Flag	Default	Notes
Basic	ocr_model=basic	—	$0.001/page, 100 free pages/day, no audio/video
Premium	ocr_model=premium	✓	$0.005/page, supports all file types

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@document.pdf" \
  -F "ocr_model=basic"

OCR engine

For the premium model, select the OCR engine for text extraction:

Engine	Flag	Default	Notes
EasyOCR	ocr_engine=easyocr	✓	Balanced speed and accuracy
Tesseract	ocr_engine=tesseract	—	Fast, widely supported
RapidOCR	ocr_engine=rapidocr	—	Optimized for CJK text

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@document.pdf" \
  -F "ocr_model=premium" \
  -F "ocr_engine=rapidocr"

Columns

Specify column names to guide the model's extraction. If omitted, default columns are used.

Comma-separated

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@document.pdf" \
  -F "columns=날짜,계정과목,적요,입금액,출금액,잔액"

JSON array

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@document.pdf" \
  -F 'columns=["date","account","description","debit","credit","balance"]'

Prompt

Add extra instructions for the model to customize extraction behavior.

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@document.pdf" \
  -F "prompt=Extract only rows where the amount is greater than 1,000,000"

Common prompt examples:

• "Merge multi-line cells into single cells"
• "Ignore header rows and only extract data rows"
• "Use YYYY-MM-DD date format"
• "Include a row number column"

DPI

Control the rendering resolution for PDF pages. Higher DPI improves accuracy for small text but increases processing time.

DPI	Use case
300	Default, good for most documents
150	Lower detail, faster processing
600	Very fine print, receipts

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@document.pdf" \
  -F "dpi=300"

Docling refinement

docling_refinement applies an optional LLM post-processing step to Docling-preprocessed documents (PDF, Office, HWP) to improve layout accuracy. Costs 3 md ($0.003) per page.

Available via both the web app and the v1 API (docling_refinement form field on POST /jobs/upload).

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@document.docx" \
  -F "docling_refinement=true"

Relative paths (for archives)

When uploading archives, you can specify relative paths to preserve directory structure:

curl -X POST https://your-domain.com/api/v1/jobs/upload \
  -H "X-API-Key: chu_live_xxxxxxxx" \
  -F "files=@archive.zip" \
  -F 'relative_paths=["folder/doc1.pdf","folder/doc2.pdf"]'