v0.3.1

[0.3.1]

Added

• {DOCUMENT_IMAGE} Placeholder Support in Pattern-2
  • Added new {DOCUMENT_IMAGE} placeholder for precise image positioning in classification and extraction prompts
  • Enables strategic placement of document images within prompt templates for enhanced multimodal understanding
  • Supports both single images and multi-page documents (up to 20 images per Bedrock constraints)
  • Full backward compatibility - existing prompts without placeholder continue to work unchanged

Fixed

• When encountering excessive Bedrock throttling, service returned 'unclassified' instead of retrying, when using multi-modal page level classification method.
• Minor documentation issues.