Best practices for parsing

​

Handling handwriting or small text

• Use enhance.agentic=[{"scope": "text"}].
• Best for: forms, signatures, handwritten notes.
• Tradeoff: slightly higher cost and latency in exchange for higher accuracy.

​

Working with non-Germanic languages

• Use settings.ocr_system="standard" (the default in v3).
• Needed for: languages outside English, Spanish, Italian, Portuguese, French, or German.
• Also improves parsing of Unicode and special characters.

​

Unexpected symbols in your output

• Cause: metadata embeddings in PDFs may contain corrupted or hidden text.
• Note: In v3, the extraction mode is automatically optimized and no longer configurable. If you continue to see unexpected symbols, please contact support.

​

Missing checkboxes or images

• Checkboxes: In v3, checkbox detection is automatically enabled and no longer needs configuration.
• Images: To return images for figures or tables, use settings.return_images:
  • settings.return_images=["figure"]: Returns figure images
  • settings.return_images=["table"]: Returns table images
  • settings.return_images=["figure", "table"]: Returns both

​

Complex tables causing problems?

• Option 1: Use agentic table mode
  • Enable enhance.agentic with [{"scope": "table"}].
  • Add prompts to guide how rows and columns should align using prompt field.
• Option 2: AI JSON format
  • Use formatting.table_output_format set to "ai_json".
  • Passes the table image to a model for structural analysis.
  • Tradeoff: higher latency, sometimes higher accuracy.

​

Spreadsheet related configurations

• spreadsheet.include=["cell_colors"]: adds Excel cell color details with LaTeX to the parse output.
• spreadsheet.clustering: splits up individual tables inside of multi-table spreadsheets (use "accurate", "fast", or "disabled").
• spreadsheet.split_large_tables: splits very large tables into smaller, manageable chunks for downstream processing.