Update app.py

app.py

@@ -2470,20 +2470,29 @@ def _extract_nfcu_summary_table(pdf_path: str, page_num: int) -> str:
 
         anchor_idx = None
         for i, (_, _, text) in enumerate(lines):
-            if "summary of your deposit accounts" in text.lower():
+            tl = text.lower()
+            if "summary of your deposit accounts" in tl or ("summary" in tl and "deposit accounts" in tl):
                 anchor_idx = i
                 break
         if anchor_idx is None:
             return ""
 
-        money_re = re.compile(r"^[\d,]+\.\d{2}-?$")
-        date_re = re.compile(r"^\d{2}-\d{2}$")
+        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)",
             re.IGNORECASE,
         )
 
+        # NFCU summary table real columns from PDF:
+        # Account | Previous Balance | Deposits/Credits | Withdrawals/Debits | Ending Balance | YTD Dividends
         rows = []
+        pending_account = ""
+        account_name_re = re.compile(
+            r"^(business checking|mbr business savings|checking|savings|money market|certificate).*$",
+            re.IGNORECASE,
+        )
+
         for _, toks, text in lines[anchor_idx + 1:]:
             if not text:
                 continue
@@ -2491,35 +2500,51 @@ def _extract_nfcu_summary_table(pdf_path: str, page_num: int) -> str:
                 break
             if not toks:
                 continue
-            if not date_re.match(toks[0]):
+            t0 = toks[0].lower()
+            tl = text.lower()
+
+            # Ignore split header lines ("Previous ...", "Balance Credits ...")
+            if (
+                "previous" in tl
+                or "deposits" in tl
+                or "withdrawals" in tl
+                or "ending" in tl
+                or "dividends" in tl
+            ) and not any(money_re.match(t) for t in toks):
                 continue
 
-            date = toks[0]
-            rest = toks[1:]
-            mvals = [t for t in rest if money_re.match(t)]
-            desc_tokens = [t for t in rest if not money_re.match(t)]
-            desc = " ".join(desc_tokens).strip()
-            if not desc:
+            # Account name can appear on its own line, followed by acct# + numeric columns.
+            if account_name_re.match(text.strip()) and not any(money_re.match(t) for t in toks):
+                pending_account = text.strip()
                 continue
 
-            amount = ""
-            balance = ""
-            if len(mvals) >= 2:
-                amount = mvals[0]
-                balance = mvals[-1]
-            elif len(mvals) == 1:
-                # Beginning/ending balance rows usually carry the number in Balance.
-                if re.search(r"\b(beginning|ending)\s+balance\b", desc, re.IGNORECASE):
-                    balance = mvals[0]
-                else:
-                    amount = mvals[0]
+            mvals = [t.replace("$", "") for t in toks if money_re.match(t)]
+            acct_tokens = [t for t in toks if acct_no_re.match(t)]
 
-            rows.append([date, desc, "", amount, balance])
+            # Totals row: "Totals <five amounts>"
+            if t0 == "totals" and len(mvals) >= 5:
+                rows.append(["Totals", mvals[0], mvals[1], mvals[2], mvals[3], mvals[4]])
+                continue
+
+            # Account numeric row: "<acctno> <five amounts>"
+            if acct_tokens and len(mvals) >= 5:
+                acct = acct_tokens[0]
+                acct_label = (pending_account + " - " + acct).strip(" -") if pending_account else acct
+                rows.append([acct_label, mvals[0], mvals[1], mvals[2], mvals[3], mvals[4]])
+                pending_account = ""
+                continue
 
         if not rows:
             return ""
 
-        hdr = ["Date", "Transaction", "Detail", "Amount($)", "Balance($)"]
+        hdr = [
+            "Account",
+            "Previous Balance",
+            "Deposits/Credits",
+            "Withdrawals/Debits",
+            "Ending Balance",
+            "YTD Dividends",
+        ]
         return _grid_to_html([hdr] + rows)
     except Exception:
         return ""
@@ -2543,8 +2568,7 @@ def _replace_nfcu_summary_table(page_md: str, summary_table_html: str) -> str:
                 return page_md.replace(old_table, summary_table_html, 1)
 
     # Fallback path: heading may be malformed/absent in OCR text.
-    # For NFCU, summary table is the 5-column table with
-    # Date + Transaction + Detail + Amount + Balance headers.
+    # For NFCU summary, header should contain balance/deposit/withdrawal/ytd signals.
     for m in table_pattern.finditer(page_md):
         tbl = m.group(0)
         p = TableGridParser()
@@ -2554,16 +2578,58 @@ def _replace_nfcu_summary_table(page_md: str, summary_table_html: str) -> str:
             continue
         h = " ".join(str(c or "").strip().lower() for c in g[0])
         if (
-            "date" in h
-            and "transaction" in h
-            and "detail" in h
-            and "amount" in h
-            and "balance" in h
+            "previous" in h
+            and "deposits" in h
+            and "withdrawals" in h
+            and "ending" in h
+            and "dividends" in h
         ):
             return page_md.replace(tbl, summary_table_html, 1)
 
     return page_md
 
+def _split_fused_date_transaction_header(grid):
+    """
+    Fix fused header "Date Transaction Detail" -> separate Date + Transaction Detail.
+    Strict guard: only when header has that exact fused phrase and also has
+    Amount/Balance columns, so unrelated tables are untouched.
+    """
+    if not grid or len(grid) < 2:
+        return grid
+    ncols = max(len(r) for r in grid if isinstance(r, list))
+    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):
+        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:]:
+        cells = [str(c or "").strip() for c in (row + [""] * (ncols - len(row)))]
+        first = cells[0]
+        dval = ""
+        tval = first
+        if first:
+            parts = first.split()
+            if parts and date_tok.match(parts[0]):
+                dval = parts[0]
+                tval = " ".join(parts[1:]).strip()
+        new_grid.append([dval, tval] + cells[1:])
+    return new_grid
+
 def normalize_html_tables(text: str) -> str:
     """
     For every <table>...</table>:
@@ -2612,6 +2678,9 @@ def normalize_html_tables(text: str) -> str:
             # Fix 2: misplaced header row promotion (keyword-score guard — safe on clean PDFs)
             grid = _promote_misplaced_header_row(grid)
 
+            # Fix NF-2: split fused "Date Transaction Detail" header/cell.
+            grid = _split_fused_date_transaction_header(grid)
+
             # Fix 8: split fused multi-keyword header cell (First Horizon Bank)
             grid = _split_fused_multicolumn_header(grid)