Add --ocr-only mode for standalone batch OCR without upload/embed

Scans entire library for scanned PDFs (pdffonts detection), OCRs them
in-place with ocrmypdf (tur+eng, 3 retries per file). No AnythingLLM
API needed — works offline. Supports --persona, --cluster, --priority
filters and --dry-run preview.

Usage:
  python3 setup.py --ocr-only --dry-run    # preview
  python3 setup.py --ocr-only              # OCR all scanned PDFs
  python3 setup.py --ocr-only --cluster intel

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

setup.py

@@ -928,6 +928,91 @@ def show_status(config):
 # MAIN
 # ──────────────────────────────────────────────────────────
 
+def ocr_only(config, persona_list=None, priority_filter=None,
+             dry_run=False, max_size_mb=100, workers=4):
+    """Scan entire library for scanned PDFs and OCR them in-place.
+    No upload, no embedding — just makes PDFs text-searchable."""
+    log.info("═══ OCR-Only Mode ═══")
+
+    book_library = Path(config["storage"]["book_library"])
+    progress = load_progress()
+
+    # Scan & classify
+    log.info("Scanning folders for scanned PDFs...")
+    text_files, scanned_files, persona_folders, _ = collect_all_files(
+        config, book_library, persona_list, priority_filter, max_size_mb,
+    )
+
+    total_scanned = sum(len(v) for v in scanned_files.values())
+    already_ocr = len(progress.get("ocr_done", []))
+    already_failed = len(progress.get("ocr_failed", []))
+
+    # Count pending
+    pending_count = 0
+    for files in scanned_files.values():
+        for f in files:
+            sf = str(f)
+            if sf not in progress.get("ocr_done", []) and sf not in progress.get("ocr_failed", []):
+                pending_count += 1
+
+    log.info(f"Scanned PDFs found:  {total_scanned}")
+    log.info(f"Already OCR'd:       {already_ocr}")
+    log.info(f"Previously failed:   {already_failed}")
+    log.info(f"Pending OCR:         {pending_count}")
+
+    if dry_run:
+        for fn, files in sorted(scanned_files.items()):
+            pending = [f for f in files
+                       if str(f) not in progress.get("ocr_done", [])
+                       and str(f) not in progress.get("ocr_failed", [])]
+            if pending:
+                log.info(f"  {fn}: {len(pending)} PDFs to OCR")
+        return
+
+    if pending_count == 0:
+        log.info("Nothing to OCR.")
+        return
+
+    ocr_ok = ocr_fail = 0
+    for fn, files in sorted(scanned_files.items()):
+        pending = [f for f in files
+                   if str(f) not in progress.get("ocr_done", [])
+                   and str(f) not in progress.get("ocr_failed", [])]
+        if not pending:
+            continue
+
+        log.info(f"→ {fn}: {len(pending)} PDFs")
+
+        for i, pdf in enumerate(pending):
+            size_mb = pdf.stat().st_size / (1024 * 1024)
+            log.info(f"  [{ocr_ok + ocr_fail + 1}/{pending_count}] "
+                     f"{pdf.name} ({size_mb:.1f}MB)")
+
+            success = False
+            for attempt in range(3):
+                if ocr_pdf(pdf, language="tur+eng", dpi=200):
+                    success = True
+                    break
+                log.warning(f"    OCR attempt {attempt+1}/3 failed, retrying...")
+                time.sleep(2)
+
+            if success:
+                progress.setdefault("ocr_done", []).append(str(pdf))
+                ocr_ok += 1
+                log.info(f"    ✓ OCR OK")
+            else:
+                progress.setdefault("ocr_failed", []).append(str(pdf))
+                ocr_fail += 1
+                log.error(f"    ✗ OCR FAILED after 3 attempts")
+
+            if (ocr_ok + ocr_fail) % 5 == 0:
+                save_progress(progress)
+
+        save_progress(progress)
+
+    log.info(f"── OCR complete: {ocr_ok} OK, {ocr_fail} failed ──")
+
+
 def reassign_workspaces(config, persona_list=None, reset=False, dry_run=False):
     """Re-assign already-uploaded docs to workspaces without scanning/uploading.
     Skips the slow folder scan — uses upload_progress.json directly."""
@@ -1003,6 +1088,7 @@ def main():
     parser.add_argument("--create-workspaces", action="store_true")
     parser.add_argument("--upload-documents", action="store_true")
     parser.add_argument("--all", action="store_true", help="Run all steps")
+    parser.add_argument("--ocr-only", action="store_true", help="Only scan and OCR scanned PDFs (no upload/embed)")
     parser.add_argument("--reassign", action="store_true", help="Re-assign uploaded docs to workspaces (no scan/upload)")
     parser.add_argument("--reset", action="store_true", help="Reset assignment tracking before reassign")
     parser.add_argument("--status", action="store_true")
@@ -1020,7 +1106,7 @@ def main():
     config = load_config()
 
     if not any([args.storage_setup, args.create_workspaces,
-                args.upload_documents, args.reassign, args.all, args.status]):
+                args.upload_documents, args.ocr_only, args.reassign, args.all, args.status]):
         parser.print_help()
         return
 
@@ -1033,6 +1119,10 @@ def main():
         storage_setup(config, dry_run=args.dry_run)
     if args.create_workspaces or args.all:
         create_workspaces(config, persona_list=persona_list, dry_run=args.dry_run)
+    if args.ocr_only:
+        ocr_only(config, persona_list=persona_list,
+                 priority_filter=args.priority,
+                 dry_run=args.dry_run, max_size_mb=args.max_size)
     if args.reassign:
         reassign_workspaces(config, persona_list=persona_list,
                             reset=args.reset, dry_run=args.dry_run)