Available now
PDF, DOCX and OCR detection
The Document Reader converts supported files into readable text before analysis. It prefers fast native extraction and uses OCR for scanned or layout-heavy material.
Extraction pipeline
- Validate
The web route verifies type, size and access before forwarding the file.
- Extract natively
PDF text layers, DOCX structure and plain text formats use native readers first.
- Apply OCR when needed
Scans, images and complex layouts can use Docling to recover readable text.
- Return normalized text
The detector receives plain text plus extraction metadata, not a visual imitation of the page.
Limits and access
- Document upload is available from Plus.
- The maximum upload size is 20 MB.
- The extraction service returns up to 120,000 characters before the selected tool applies its own limit.
- OCR jobs use short-lived in-memory state and expire after approximately 15 minutes.
Check extraction quality
Review the extracted text before interpreting a detection score. Missing columns, scrambled reading order or OCR substitutions change the linguistic structure available to the detector.