Extract page number as metadata instead of silently removing it

_filter_footer_words now returns page number info (text, y_pct, number)
instead of just removing footer words. The page number is included in
the grid result as `page_number` and displayed in the frontend summary
bar as "S. 233".

This preserves page numbers for later page concatenation in the
customer frontend while still removing them from the grid content.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

admin-lehrer/components/grid-editor/GridEditor.tsx

@@ -174,6 +174,11 @@ export function GridEditor({ sessionId, onNext }: GridEditorProps) {
             Woerterbuch ({Math.round(grid.dictionary_detection.confidence * 100)}%)
           </span>
         )}
+        {grid.page_number?.text && (
+          <span className="px-1.5 py-0.5 rounded bg-gray-100 dark:bg-gray-700 text-gray-600 dark:text-gray-300 border border-gray-200 dark:border-gray-600">
+            S. {grid.page_number.text}
+          </span>
+        )}
         <span className="text-gray-400">
           {grid.duration_seconds.toFixed(1)}s
         </span>

admin-lehrer/components/grid-editor/types.ts

@@ -20,6 +20,13 @@ export interface DictionaryDetection {
   headword_col_index: number | null
 }
 
+/** Page number extracted from footer region of the scan. */
+export interface PageNumber {
+  text: string
+  y_pct: number
+  number?: number
+}
+
 /** A complete structured grid with zones, ready for the Excel-like editor. */
 export interface StructuredGrid {
   session_id: string
@@ -31,6 +38,7 @@ export interface StructuredGrid {
   formatting: GridFormatting
   layout_metrics?: LayoutMetrics
   dictionary_detection?: DictionaryDetection
+  page_number?: PageNumber | null
   duration_seconds: number
   edited?: boolean
   layout_dividers?: LayoutDividers

klausur-service/backend/grid_editor_api.py

@@ -124,8 +124,9 @@ async def _build_grid_core(
 
     # 2c. Filter footer rows (page numbers at the very bottom).
     # Isolated short text in the bottom 5% of the page is typically a
-    # page number ("64", "S. 12") and not real content.
-    _filter_footer_words(all_words, img_h, logger, session_id)
+    # page number ("64", "S. 12") and not real content.  The page number
+    # is extracted as metadata for the frontend header display.
+    page_number_info = _filter_footer_words(all_words, img_h, logger, session_id)
 
     # 2c2. Filter OCR junk from header illustrations.
     # Low-confidence short fragments above the first real content row.
@@ -1668,6 +1669,7 @@ async def _build_grid_core(
             "ipa_applied": bool(ipa_target_cols) if not skip_ipa else False,
             "syllables_applied": syllable_insertions > 0,
         },
+        "page_number": page_number_info,
         "duration_seconds": round(duration, 2),
     }
 

klausur-service/backend/grid_editor_helpers.py

@@ -1301,29 +1301,42 @@ def _filter_footer_words(
     img_h: int,
     log: Any,
     session_id: str,
-) -> None:
+) -> Optional[Dict]:
     """Remove isolated words in the bottom 5% of the page (page numbers).
 
-    Modifies *words* in place.
+    Modifies *words* in place and returns a page_number metadata dict
+    if a page number was extracted, or None.
     """
     if not words or img_h <= 0:
-        return
+        return None
     footer_y = img_h * 0.95
     footer_words = [
         w for w in words
         if w["top"] + w.get("height", 0) / 2 > footer_y
     ]
     if not footer_words:
-        return
+        return None
     # Only remove if footer has very few words (≤ 3) with short text
     total_text = "".join((w.get("text") or "").strip() for w in footer_words)
     if len(footer_words) <= 3 and len(total_text) <= 10:
+        # Extract page number metadata before removing
+        page_number_info = {
+            "text": total_text.strip(),
+            "y_pct": round(footer_words[0]["top"] / img_h * 100, 1),
+        }
+        # Try to parse as integer
+        digits = "".join(c for c in total_text if c.isdigit())
+        if digits:
+            page_number_info["number"] = int(digits)
+
         footer_set = set(id(w) for w in footer_words)
         words[:] = [w for w in words if id(w) not in footer_set]
         log.info(
-            "build-grid session %s: removed %d footer words ('%s')",
-            session_id, len(footer_words), total_text,
+            "build-grid session %s: extracted page number '%s' and removed %d footer words",
+            session_id, total_text, len(footer_words),
         )
+        return page_number_info
+    return None
 
 
 def _filter_header_junk(