Organized

docs/raw-notes/features/ocr-rename.md

@@ -0,0 +1,46 @@
+# Feature detail: OCR attachment → subject rename
+
+**Parent roadmap:** [[../features-for-redmine_tickets-sync]]  
+**Status:** MVP 2026-07-15  
+**Module:** `redmine_ticket_ocr.py` · CLI: `ocr-rename`
+
+## Intent
+
+Scanned docs land in Helpdesk as tickets titled “Scanned document…”.  
+Ops need: open attachment → understand contents → rename subject (+ optional notes).
+
+## Commands
+
+```bash
+# One ticket: download OCR-able attachments, propose subject, queue private note (+ subject draft)
+./bin/rtl ocr-rename --issue-id 12345
+
+# Scan cache for generic scan subjects
+./bin/rtl ocr-rename --where '@scanned-document' --only-generic-subjects --max-tickets 20
+
+# Apply subject now (LAN/prod safety gates); still queues private OCR note by default
+./bin/rtd ocr-rename --issue-id 12345 --apply-subject
+
+# Extract only (no drafts)
+./bin/rtl ocr-rename --issue-id 12345 --no-queue-subject --no-queue-note --json
+```
+
+## Pipeline
+
+1. Load ticket + attachments (`include=attachments` refresh if needed)
+2. Download PDF/image candidates (multi-URL fallback)
+3. Extract text: `pdftotext` → else `pdftoppm` + `tesseract`
+4. Heuristic `suggest_subject_from_text` (doc type, person/company, date)
+5. Queue drafts:
+   - `fields` subject change (unless `--no-queue-subject`)
+   - private `internal_note` with OCR snippet (unless `--no-queue-note`)
+
+## Saved filter
+
+`saved-filters/scanned-document.expr` → `@scanned-document`
+
+## Still open
+
+- [ ] Stronger LLM subject summarization (optional; keep heuristics default)
+- [ ] Cron wrapper that only processes new generic-scan tickets
+- [ ] HR-project default project config example