Blog

Short notes about how the MRZ Toolkit is built and why. None of these pages replace the code or the engineering handbook — they are a reader-friendly layer on top.

Posts

• The OCR engine fallback ladder

  MRZ model first, stock `eng` data only as a last resort — how the client scan keeps working when the primary model cannot load.

  app/blog/ocr-engine-fallback
• OCR data tests are fallback-only

  Why OCR data tests should only be used when everything else fails, with functionality maintained and the user involved.

  app/ocr-fallback (top level, not under /blog)
• OCR read log: user involvement when the fallback wins

  What the scan UI shows when the weak `eng` fallback engine produced the result, and the “Try the MRZ model again” retry.

  app/log/ocr-retry
• Release note: client OCR fallback ladder

  Shipping behaviour for the OCR engine ladder: primary MRZ model, cache then refresh, stock `eng` last resort, never cached failures.

  app/release-notes/ocr-fallback
• News: OCR data tests used only as fallback

  Functionality is maintained on the fallback engine and the user is told which engine read their document.

  app/news/ocr-data-tests-fallback

Want these moved under a single `/blog` tree? The posts live in different folders on purpose so they can be reorganised — see the front matter of each page.