v0.5.0
New Features:
-
VLM Checkpoint Conversion
• Added checkpoint support to the VLM converter: resumes interrupted conversions from the last successfully processed page instead of restarting from scratch. -
Markdown Enrichment
• Completed the markdown enrichment pipeline: regex-based artifact correction, chunk-and-send LLM correction flow, and optional summary generation.
• Markdown enrichment is now available in bulk mode, mirroring single-file behaviour.
Improvements:
- Bug Fixes & Refactoring
• Resolved various bugs affecting stability and correctness across the pipeline.
• Internal refactoring for better maintainability.