Flexible OCR Processing Modes

[icereed]

Release Highlights – v0.19.0

New: Flexible OCR Processing Modes

paperless-gpt now offers three distinct OCR processing modes, allowing you to optimize document processing based on your OCR provider and performance needs:

• Image Mode (default)
  Converts each PDF page into an image before OCR.
  Best for: Maximum compatibility with all OCR providers.
  Configure via: OCR_PROCESS_MODE: "image"

• PDF Mode
  Processes individual PDF pages directly, without image conversion.
  Best for: Preserving PDF structure and improving speed or accuracy with native PDF-compatible providers.
  Configure via: OCR_PROCESS_MODE: "pdf"

• Whole PDF Mode
  Sends the entire PDF as a single document for OCR.
  Best for: Providers optimized for multi-page processing and reduced API calls.
  Configure via: OCR_PROCESS_MODE: "whole_pdf"
  Note: Large PDFs may exceed your provider's API limits—switch to pdf mode if issues occur.

Enhancements & Fixes

• New Feature: Native PDF processing support for OCR #406
• Fix: Azure OpenAI LLM integration #398
• Dependency Updates:
  • Core libraries: React, TypeScript, Node.js types, ESLint
  • Backend: Gin, GORM, Google API modules, ocrchestra digest updates
  • Build & test tools: Go 1.24.3, Docker images, Testcontainers

Full Changelog: v0.18.0...v0.19.0

---

This discussion was created from the release Flexible OCR Processing Modes.