- Implement tests for converter properties, ensuring correct input and output formats.
- Add tests for chip parsing, covering various image, pdf_image, and pdfa chip formats.
- Include tests for the isArchiveOutput and getOutputFileName functions.
- Mock execFile for testing the convert function, verifying command execution and output file creation.
- Ensure all chips are parseable and validate the completeness of chip categories.
- Updated README.md to include new example files for minimal and production configurations with detailed instructions.
- Added compose.production-alt.example.yml for an annotated production setup.
- Introduced new example files: compose.minimal.example.yml, compose.production.example.yml, nginx.example.conf, and traefik.example.yml for better clarity and usability.
- Removed outdated configuration files and consolidated documentation for OCR language support.
- Adjusted Nginx and Traefik configurations to reflect best practices and added necessary comments for user guidance.
- Minor formatting and consistency improvements across documentation files.
- Introduced `compose.ocr-languages.yml` for installing additional Tesseract OCR language packs during container startup.
- Updated documentation to include new OCR language extension options and usage scenarios.
- Enhanced existing documentation with references to the new OCR language extension guide.
- Improved test coverage for the deark converter and formats, ensuring accurate format handling and error management.
- Refactored code to streamline file system operations and improve readability.
Breaking change from Inkscape 1.0+:
- Old syntax: inkscape input.png -o output.svg (triggers GTK init)
- New syntax: inkscape input.png --export-type=svg --export-filename=output.svg
This is the correct headless-safe approach, no need for xvfb.
Ref: https://inkscape.org/doc/inkscape-man.html
- Inkscape requires X11 display connection, causing 'GtkStyleContext without display' error
- Solution: use xvfb-run to create virtual X11 display
- Fallback: try direct inkscape if xvfb-run fails
- Add xvfb package to Dockerfile
- Update inkscape tests for new xvfb-run behavior
Fixes: Gtk-ERROR: Can't create a GtkStyleContext without a display connection
- Add existsSync checks in download.tsx before file access
- Fix normalizeFiletype.ts: md-t/md-i return original format (not tar.gz)
- Add outputMode: 'archive' support in main.ts with auto .tar suffix
- PDFMathTranslate outputs translated-<lang>.pdf + bilingual-<lang>.pdf (no original)
- MinerU outputs complete folder in .tar with logging
- Hide preview icon for .tar files in results.tsx
- Update Bun version: 1.2.2 -> 1.3.6
- Add Step 19: auto-deploy to remote server after Docker push
All 159 tests pass.
- Introduced PDFMathTranslate converter to handle PDF translations while preserving mathematical formulas.
- Updated Dockerfile to include necessary dependencies and pre-download models during build.
- Enhanced README.md to document the new converter and its features.
- Added tests for PDFMathTranslate to ensure correct functionality and error handling.
- Change archive format from zip to tar.gz
- Use 'mineru' CLI command instead of 'magic-pdf'
- Install mineru[all] via pipx in Dockerfile
- Remove zip dependency from Dockerfile
- Update normalizeFiletype to output tar.gz extension
- Update all tests for tar.gz output format
- Simplify README MinerU section
- Add MinerU converter engine (src/converters/mineru.ts)
- Support md-t (table as Markdown) and md-i (table as image) output formats
- Input formats: pdf, ppt, pptx, xls, xlsx, doc, docx
- Output as ZIP archive containing Markdown and images
- Update Dockerfile to install magic-pdf via pipx
- Add zip utility for archive creation
- Update normalizeFiletype to map md-t/md-i to zip extension
- Add comprehensive tests for MinerU converter
- Update README with MinerU documentation
