Update app.py

app.py

@@ -2643,6 +2643,58 @@ def _split_fused_date_transaction_header(grid):
         new_grid.append(split_row)
     return new_grid
 
+def _split_combined_summary_daily_balance_grid(grid):
+    """
+    Some OCR outputs fuse an account summary table and a Daily Balance block
+    into one HTML table. Split them into two tables when structure is clear.
+    Generic guards:
+    - A row containing "Daily Balance" exists
+    - There are rows after it with date-like + amount-like tokens
+    """
+    if not grid or len(grid) < 4:
+        return None
+
+    def _norm(s):
+        return re.sub(r"\s+", " ", str(s or "").strip().lower())
+
+    daily_idx = None
+    for i, row in enumerate(grid):
+        row_text = " ".join(_norm(c) for c in row if str(c or "").strip())
+        if "daily balance" in row_text:
+            daily_idx = i
+            break
+    if daily_idx is None or daily_idx <= 1 or daily_idx >= len(grid) - 2:
+        return None
+
+    date_re = re.compile(r"^\d{1,2}/\d{2}(?:/\d{2,4})?$")
+    money_re = re.compile(r"^-?\$?\d{1,3}(?:,\d{3})*(?:\.\d{2})-?$")
+
+    daily_rows = []
+    for row in grid[daily_idx + 1:]:
+        tokens = []
+        for cell in row:
+            ct = str(cell or "").strip()
+            if not ct:
+                continue
+            tokens.extend(re.split(r"\s+", ct))
+        dates = [t for t in tokens if date_re.match(t)]
+        amts = [t for t in tokens if money_re.match(t)]
+        if not dates or not amts:
+            continue
+        for d, a in zip(dates, amts):
+            daily_rows.append([d, a])
+
+    # Need meaningful daily table to split safely.
+    if len(daily_rows) < 2:
+        return None
+
+    summary = [r for r in grid[:daily_idx] if any(str(c or "").strip() for c in r)]
+    if len(summary) < 2:
+        return None
+
+    daily = [["Date", "Ledger Balance"]] + daily_rows
+    return [summary, daily]
+
 def _normalize_fused_date_posted_amount_table(grid):
     """
     Normalize 2-column OCR tables with fused header like:
@@ -2799,8 +2851,12 @@ def normalize_html_tables(text: str) -> str:
 
             # Fix 3: fused key-value rows in summary sections (no-space guard — safe on all PDFs)
             grid = _fix_fused_keyvalue_rows(grid)
-
-            out.append(_grid_to_html(grid) if grid else table_html)
+            # Fix 12: split fused summary+daily-balance combined table, if detected.
+            split_tables = _split_combined_summary_daily_balance_grid(grid)
+            if split_tables:
+                out.append("\n\n".join(_grid_to_html(g) for g in split_tables if g))
+            else:
+                out.append(_grid_to_html(grid) if grid else table_html)
         except Exception:
             out.append(table_html)
         last = m.end()