Document OCR review UI and docs
Document OCR review UI and docs
Line 12: Line 12:


=== 10 September 2026 ===
=== 10 September 2026 ===
* '''OCR review queue extras''' — '''Preview Apply''' shows planned soldier-page changes before writing. ASN matching tries padded/stripped ASN forms and title hits. Multi-person enrich results are cached on the job. Old applied/discarded jobs are pruned automatically. Facebook OCR candidates (70–79%) can be seeded into [[Contribute documents]] for human review without re-running the engine (`python scripts/seed_fb_ocr_review_jobs.py`).
* '''OCR soft-launch hardening''' — Document previews load through the wiki (not the loopback OCR port), so reviewers off localhost can see the scan. Dates and ranks are normalized before Apply; long PDFs warn when truncated at 8 pages. The Contribute documents queue shows who uploaded, age, match counts, and “Only mine / Ready / In progress” filters. Apply can optionally write selected notes into Research notes, checkpoints each soldier page so a failed mass roster Apply can resume, and soldier-page uploads can mark “multi-person roster” (or rebuild into roster mode from Advanced).
* '''OCR soft-launch hardening''' — Document previews load through the wiki (not the loopback OCR port), so reviewers off localhost can see the scan. Dates and ranks are normalized before Apply; long PDFs warn when truncated at 8 pages. The Contribute documents queue shows who uploaded, age, match counts, and “Only mine / Ready / In progress” filters. Apply can optionally write selected notes into Research notes, checkpoints each soldier page so a failed mass roster Apply can resume, and soldier-page uploads can mark “multi-person roster” (or rebuild into roster mode from Advanced).