This release adds sealed OCR configuration modes, improves OCR result compatibility for formulas and adjacent table captions, and prepares pdf-craft for VGE worktree-based development.
What's Changed
Features
- Sealed OCR configuration modes in #371
- Added public OCR configuration objects for local DeepSeek OCR, vendor DeepSeek OCR, and vendor Unlimited OCR
- Keeps
doc-page-extractorconstruction behind pdf-craft's public configuration surface - Updated README and development documentation for the supported OCR modes
Bug Fixes
-
Inline LaTeX formula parsing in #373
- Preserves inline LaTeX before Markdown and HTML parsing can split formula text
- Adds regression coverage for formula-heavy OCR text
-
Adjacent table caption joining in ad3c064
- Improves compatibility when OCR returns table titles, captions, or adjacent descriptive blocks as separate records
- Adds tests for table caption and description joining behavior
-
VGE worktree environment inheritance in #372
- Copies the source workspace
.envinto new worktrees when available - Keeps vendor OCR credentials in private worktree configuration rather than committed files
- Copies the source workspace
Documentation and Release Workflow
- Added VGE-oriented Agent documentation and worktree guidance in 82e4caf
- Added the manual GitHub Actions release workflow in #375
- Updated related project, OOMOL links, changelog, and maintenance coverage in #358, #359, #366, and #367
Pull Requests Included
Full Changelog: v1.0.13...v1.0.14