[SimpleCaci]

Summary

• Add a self-contained HTML review report pairing each source page with recognized text.
• Calculate and display average OCR word confidence per page.
• Keep the existing text and PDF outputs unchanged and document the new workflow.

Progression

• Previous state: OCR exported text and PDF without an easy visual comparison step.
• New state: every conversion can produce a portable source/transcription review page.
• Next unlock or recommended follow-up: install Tesseract locally and capture a reviewed before/after sample.

Changes

• Switch OCR collection to structured Tesseract data so word confidence is available.
• Embed source images directly in a responsive, print-friendly HTML report.
• Add --no-review-report for text/PDF-only runs.
• Extend deterministic tests for confidence, escaping, and embedded images.

Impact

• Added: reviewable HTML output and per-page confidence.
• Improved: transparency around imperfect handwriting recognition.
• Unchanged: local-only processing and existing text/PDF exports.

Verification

• Automated tests — 4 passed
• Lint/type-check — Ruff passed
• Manual behavior check — Tesseract executable is not installed in this environment
• Security/privacy review — report stays local and embeds only user-selected source files

Risks and Trade-offs

• Confidence is Tesseract's word-level estimate, not a guarantee of transcription accuracy.
• Embedded source images can make the HTML report large.

Reviewer Focus

Review the structured OCR conversion and HTML escaping/data-URI generation in main.py.

Follow-ups

• Run the sample set with a local Tesseract installation.
• Add orientation and perspective correction separately.