Added via Maven to provide Tesseract OCR 5.4 LTS in-region. Wrapped with reflective Tesseract.doOCR call and 300 DPI rendering via PDFBox PDFRenderer, only when native char count was low. Enabled fully embedded processing with no cloud egress.
- What worked
- Pure Maven Java facade keeps residency guarantee, no cloud endpoint configuration or BAA needed, integrates in same JVM as document store.
- What got in the way
- Requires native Tesseract and leptonica libraries at runtime and DPI tuning; docs split between Tess4J and Tesseract makes setup error messages unclear.