Update app.py

app.py

@@ -51,6 +51,10 @@ Primary goals for reconcile rate:
             and the first is mostly credit/deposit rows while the second is mostly debit/withdrawal-style
             rows (generic keywords), drop the second — fixes OCR pasting the withdrawals block again.
 - Fix 23b: Loose multiset keys use normalized float amounts so comma/spacing variants still match.
+- Fix 26: Drop small DA3 fragments (3–14 rows) that are a multiset subset of an earlier larger DA3
+            table — matches PDF layouts that re-print a header and repeat a few rows.
+- Fix 27: Strip a lone “…(continued)” account banner line (no DATE/DESCRIPTION/AMOUNT) between
+            page separators — duplicate header noise without the long flatline.
 """
 
 # Patch asyncio first (before Gradio imports it) to suppress Python 3.13 cleanup noise
@@ -3476,6 +3480,37 @@ def _strip_orphan_continued_da3_flatline(text: str) -> str:
     return "".join(out)
 
 
+def _strip_standalone_continued_banner_line(text: str) -> str:
+    """
+    Fix 27: Remove a single line that is only an account/section “(continued)” banner
+    (no transaction columns), often duplicated between page-separator and the next heading/table.
+    """
+    if not text or "(continued)" not in text.lower():
+        return text
+    lines = text.splitlines(keepends=True)
+    out = []
+    for line in lines:
+        core = line.rstrip("\r\n")
+        if not core.strip():
+            out.append(line)
+            continue
+        c = core.lower()
+        if "$" in core or "=" in core:
+            out.append(line)
+            continue
+        if (
+            "(continued)" in c
+            and "date" not in c
+            and "description" not in c
+            and "amount" not in c
+            and len(re.findall(r"\d{2}/\d{2}", core)) == 0
+            and len(core) < 240
+        ):
+            continue
+        out.append(line)
+    return "".join(out)
+
+
 def _patch_da3_msft_g_ref_amounts_across_tables(text: str) -> str:
     """
     Fix 21: Build MICROSOFT#G… -> strict Amount from any DA3 row, then patch other DA3 rows
@@ -3723,18 +3758,29 @@ def _dedupe_subset_transaction_tables_html(text: str) -> str:
                 ci = _row_counter_loose_da3(gi)
                 cj = _row_counter_loose_da3(gj)
                 ni, nj = sum(ci.values()), sum(cj.values())
-                if ni >= 6 and nj >= 6:
+                if ni >= 1 and nj >= 1:
                     oi_j = _overlap_ratio_counter(ci, cj)
                     oj_i = _overlap_ratio_counter(cj, ci)
-                    if oi_j >= 0.88 and oj_i >= 0.88:
+                    # Short re-printed fragment (often 3–5 rows after a repeated header on same PDF page)
+                    if (
+                        nj >= 8
+                        and 3 <= ni <= 14
+                        and ni < nj
+                        and _multiset_leq(ci, cj)
+                        and oi_j >= 0.9
+                    ):
                         drop_idx.add(i)
                         break
-                    if _multiset_leq(ci, cj) and ni < nj and oi_j >= 0.9:
-                        drop_idx.add(i)
-                        break
-                    if _multiset_leq(cj, ci) and nj < ni and oj_i >= 0.9:
-                        drop_idx.add(j)
-                        continue
+                    if ni >= 6 and nj >= 6:
+                        if oi_j >= 0.88 and oj_i >= 0.88:
+                            drop_idx.add(i)
+                            break
+                        if _multiset_leq(ci, cj) and ni < nj and oi_j >= 0.9:
+                            drop_idx.add(i)
+                            break
+                        if _multiset_leq(cj, ci) and nj < ni and oj_i >= 0.9:
+                            drop_idx.add(j)
+                            continue
 
         if not drop_idx:
             return t
@@ -3905,6 +3951,7 @@ def stabilize_tables_and_text(page_md: str) -> str:
     stabilized = "\n\n".join(out_blocks)
     stabilized = convert_plaintext_bank_sections(stabilized)
     stabilized = _strip_orphan_continued_da3_flatline(stabilized)
+    stabilized = _strip_standalone_continued_banner_line(stabilized)
     stabilized = normalize_html_tables(stabilized)
     stabilized = _trim_adjacent_da3_suffix_duplicating_next_table_prefix(stabilized)
     stabilized = _split_ucb_deposits_electronic_sections_html(stabilized)
@@ -3912,6 +3959,7 @@ def stabilize_tables_and_text(page_md: str) -> str:
     stabilized = _dedupe_subset_transaction_tables_html(stabilized)
     stabilized = _drop_da3_misplaced_debit_after_deposit_heading_tables(stabilized)
     stabilized = _strip_orphan_continued_da3_flatline(stabilized)
+    stabilized = _strip_standalone_continued_banner_line(stabilized)
     return close_unclosed_html(stabilized)
 
 # --------------------------
@@ -4036,6 +4084,7 @@ def run_ocr(uploaded_file):
                 if is_pdf and not is_nfcu_doc:
                     stabilized = _patch_ocr_with_textlayer(stabilized, path, page_num)
                     stabilized = _strip_orphan_continued_da3_flatline(stabilized)
+                    stabilized = _strip_standalone_continued_banner_line(stabilized)
                     # _patch_ocr_with_textlayer may append Case B rows from the PDF text layer
                     # with fused UCB-style cells; run the same HTML table pass again (Fix 13/14).
                     stabilized = normalize_html_tables(stabilized)
@@ -4044,6 +4093,7 @@ def run_ocr(uploaded_file):
                     stabilized = _patch_da3_msft_g_ref_amounts_across_tables(stabilized)
                     stabilized = _dedupe_subset_transaction_tables_html(stabilized)
                     stabilized = _drop_da3_misplaced_debit_after_deposit_heading_tables(stabilized)
+                    stabilized = _strip_standalone_continued_banner_line(stabilized)
                 elif is_pdf and is_nfcu_doc:
                     # Navy-only additive fix: rebuild malformed summary table from
                     # text layer, but keep GLM body/header/footer unchanged.
@@ -4125,12 +4175,14 @@ def run_ocr(uploaded_file):
         # Cross-page: UCB section split + same transaction tables dedupe once on full doc.
         if all_pages and merged and not merged.startswith("Error") and "<table" in merged.lower():
             merged = _strip_orphan_continued_da3_flatline(merged)
+            merged = _strip_standalone_continued_banner_line(merged)
             merged = _trim_adjacent_da3_suffix_duplicating_next_table_prefix(merged)
             merged = _split_ucb_deposits_electronic_sections_html(merged)
             merged = _patch_da3_msft_g_ref_amounts_across_tables(merged)
             merged = _dedupe_subset_transaction_tables_html(merged)
             merged = _drop_da3_misplaced_debit_after_deposit_heading_tables(merged)
             merged = _strip_orphan_continued_da3_flatline(merged)
+            merged = _strip_standalone_continued_banner_line(merged)
         return merged
 
     except Exception as e:
@@ -4155,4 +4207,3 @@ with gr.Blocks(title="GLM-OCR") as demo:
 
 if __name__ == "__main__":
     demo.launch()
-