- Delete notebooks/ (personal scratch artifacts; gitignored)
- Move nbformat to --extra migrate; lazy imports in migrate.py and
cli.py so base CLI loads without it
- Restore langchain-text-splitters in core (chunking dependency)
PaddleOCR with preprocessing, scispaCy NER, figure/table detection,
citation extraction, and chunked Markdown output with frontmatter.
Includes watch mode and notebook reprocessing.