Add manual OCR page exclusion by PDF page number

smoke_signal_tab.py

@@ -1238,7 +1238,7 @@ def _run_tesseract_batch(images):
     return outputs
 
 
-def run_ocr(page_selection: str = "", replace_book_queue: bool = True) -> tuple:
+def run_ocr(page_selection: str = "", page_exclusion: str = "", replace_book_queue: bool = True) -> tuple:
     """Run Surya OCR on all profiled PDFs."""
     df = load_manifest_df()
     debug = f"[DEBUG] SS_ROOT={SS_ROOT}\nMANIFEST_CSV={MANIFEST_CSV}\nCSV exists={MANIFEST_CSV.exists()}\n"
@@ -1269,11 +1269,26 @@ def run_ocr(page_selection: str = "", replace_book_queue: bool = True) -> tuple:
     selection_set, selection_error = _parse_page_selection(page_selection)
     if selection_error:
         return _ocr_status_html(), f"{debug}Invalid page selection: {selection_error}"
+    exclusion_spec = (page_exclusion or "").strip().lower()
+    exclusion_request = None if exclusion_spec in ("", "none", "-") else exclusion_spec
+    if exclusion_request is not None:
+        exclusion_set, exclusion_error = _parse_page_selection(exclusion_request)
+        if exclusion_error:
+            return _ocr_status_html(), f"{debug}Invalid exclusion selection: {exclusion_error}"
+    else:
+        exclusion_set = set()
     if selection_set is None:
         log.append(log_line("ℹ Page selection: all pages"))
     else:
         selected_preview = ",".join(str(p) for p in sorted(selection_set))
         log.append(log_line(f"ℹ Page selection: {selected_preview}"))
+    if exclusion_request is None:
+        log.append(log_line("ℹ Page exclusion: none"))
+    elif exclusion_set is None:
+        log.append(log_line("ℹ Page exclusion: all selected pages"))
+    else:
+        excluded_preview = ",".join(str(p) for p in sorted(exclusion_set))
+        log.append(log_line(f"ℹ Page exclusion: {excluded_preview}"))
 
     surya, surya_error, reused = _get_surya_runtime_with_timeout(SS_SURYA_LOAD_TIMEOUT_SEC)
     if surya:
@@ -1308,14 +1323,33 @@ def run_ocr(page_selection: str = "", replace_book_queue: bool = True) -> tuple:
         pages_all = profile.get("pages", [])
         page_numbers = [int(p.get("page_number", 0) or 0) for p in pages_all]
         max_page = max(page_numbers) if page_numbers else None
-        selected_pages_for_book = selection_set
-        if selection_set is not None:
+        if selection_set is None:
+            selected_pages_for_book = set(page_numbers)
+        else:
             selected_pages_for_book, sel_err = _parse_page_selection(page_selection, max_page=max_page)
             if sel_err:
                 log.append(log_line(f"⚠ {book_id}: {sel_err}"))
                 continue
-            selected_preview = ",".join(str(p) for p in sorted(selected_pages_for_book))
-            log.append(log_line(f"ℹ {book_id}: selected pages {selected_preview}"))
+            if selected_pages_for_book is None:
+                selected_pages_for_book = set(page_numbers)
+
+        if exclusion_request is None:
+            excluded_pages_for_book: set[int] = set()
+        else:
+            excluded_pages_for_book, excl_err = _parse_page_selection(exclusion_request, max_page=max_page)
+            if excl_err:
+                log.append(log_line(f"⚠ {book_id}: {excl_err}"))
+                continue
+            if excluded_pages_for_book is None:
+                excluded_pages_for_book = set(page_numbers)
+
+        selected_pages_for_book = set(selected_pages_for_book) - set(excluded_pages_for_book)
+        if not selected_pages_for_book:
+            log.append(log_line(f"⚠ {book_id}: no pages left after applying selection/exclusion"))
+            continue
+
+        selected_preview = ",".join(str(p) for p in sorted(selected_pages_for_book))
+        log.append(log_line(f"ℹ {book_id}: effective selected pages {selected_preview}"))
         processed_books.append(book_id)
 
         ocr_pages = []
@@ -1721,6 +1755,8 @@ def save_review_decision(idx: int, final_text: str, action: str, reviewer: str,
 
     decided_ids = set(d_df["region_id"].tolist()) if not d_df.empty else set()
     pending     = q_df[~q_df["region_id"].isin(decided_ids)]
+    if not pending.empty and {"book_id", "page"}.issubset(pending.columns):
+        pending = pending.sort_values(["book_id", "page"], ascending=[True, True], kind="stable")
     if pending.empty:
         return "All done!", *get_review_item(0)[1:]
 
@@ -2027,6 +2063,11 @@ def smoke_signal_tab():
                     placeholder="all or e.g. 7 or 3-8 or 1,4,9-12",
                     lines=1,
                 )
+                ocr_page_exclusion = gr.Textbox(
+                    label="Pages to Exclude (optional)",
+                    placeholder="e.g. 1,2,3,17,25 (PDF page numbers)",
+                    lines=1,
+                )
                 ocr_replace_queue = gr.Checkbox(
                     label="Replace existing review queue entries for processed books",
                     value=True,
@@ -2036,7 +2077,7 @@ def smoke_signal_tab():
 
                 ocr_run_event = ocr_btn.click(
                     run_ocr,
-                    inputs=[ocr_page_selection, ocr_replace_queue],
+                    inputs=[ocr_page_selection, ocr_page_exclusion, ocr_replace_queue],
                     outputs=[ocr_status, ocr_log],
                     show_progress="minimal",
                 )