Update app.py

app.py

@@ -2480,7 +2480,7 @@ def _extract_nfcu_summary_table(pdf_path: str, page_num: int) -> str:
         money_re = re.compile(r"^\$?[\d,]+\.\d{2}-?$")
         acct_no_re = re.compile(r"^\d{8,12}$")
         stop_re = re.compile(
-            r"(business checking\s*-|mbr business savings\s*-|date\s+transaction\s+detail|items paid|checking|savings)",
+            r"(business checking\s*-|mbr business savings\s*-|date\s+transaction\s+detail|items paid)",
             re.IGNORECASE,
         )
 
@@ -2600,26 +2600,29 @@ def _split_fused_date_transaction_header(grid):
     if ncols < 3:
         return grid
 
-    header = [str(c or "").strip() for c in (grid[0] + [""] * (ncols - len(grid[0])))]
-    first_h = re.sub(r"\s+", " ", header[0]).strip().lower()
-    if first_h != "date transaction detail":
-        return grid
-
-    hnorm = [re.sub(r"\s+", " ", str(c or "").strip().lower()) for c in header]
-    has_amount = any("amount" in h for h in hnorm)
-    has_balance = any("balance" in h for h in hnorm)
-    if not (has_amount and has_balance):
+    header_row_idx = None
+    fused_col_idx = None
+    for ri in range(min(4, len(grid))):
+        r = [str(c or "").strip() for c in (grid[ri] + [""] * (ncols - len(grid[ri])))]
+        for ci, c in enumerate(r):
+            cc = re.sub(r"\s+", " ", c).strip().lower()
+            if cc == "date transaction detail":
+                row_text = " ".join(re.sub(r"\s+", " ", x).strip().lower() for x in r)
+                if "amount" in row_text and "balance" in row_text:
+                    header_row_idx = ri
+                    fused_col_idx = ci
+                    break
+        if header_row_idx is not None:
+            break
+    if header_row_idx is None or fused_col_idx is None:
         return grid
 
     date_tok = re.compile(r"^\d{1,2}(?:[-/])\d{2}(?:[-/]\d{2,4})?$")
 
     new_grid = []
-    new_header = ["Date", "Transaction Detail"] + header[1:]
-    new_grid.append(new_header)
-
-    for row in grid[1:]:
+    for ri, row in enumerate(grid):
         cells = [str(c or "").strip() for c in (row + [""] * (ncols - len(row)))]
-        first = cells[0]
+        first = cells[fused_col_idx]
         dval = ""
         tval = first
         if first:
@@ -2627,7 +2630,17 @@ def _split_fused_date_transaction_header(grid):
             if parts and date_tok.match(parts[0]):
                 dval = parts[0]
                 tval = " ".join(parts[1:]).strip()
-        new_grid.append([dval, tval] + cells[1:])
+        if ri == header_row_idx:
+            split_row = (
+                cells[:fused_col_idx]
+                + ["Date", "Transaction Detail"]
+                + cells[fused_col_idx + 1:]
+            )
+        elif ri < header_row_idx:
+            split_row = cells[:fused_col_idx] + [cells[fused_col_idx], ""] + cells[fused_col_idx + 1:]
+        else:
+            split_row = cells[:fused_col_idx] + [dval, tval] + cells[fused_col_idx + 1:]
+        new_grid.append(split_row)
     return new_grid
 
 def normalize_html_tables(text: str) -> str: