Update app.py

app.py

@@ -2263,6 +2263,18 @@ def _parse_nfcu_page(pdf_path: str, page_num: int) -> str:
             "Errors",
             "Payments",
             "SAVINGS DIVIDENDS",
+            "REMITTANCE RECEIVED",
+            "DEPOSIT VOUCHER",
+            "ACCOUNT NUMBER",
+            "MARK \"X\"",
+            "ADDRESS/ORDER",
+            "ITEMS ON REVERSE",
+            "CHANGE OF ADDRESS",
+            "PLEASE PRINT",
+            "RANK/RATE",
+            "PO BOX",
+            "MERRIFIELD",
+            "Questions about this Statement",
         )
 
         def parse_account_rows(start_idx):
@@ -2280,6 +2292,18 @@ def _parse_nfcu_page(pdf_path: str, page_num: int) -> str:
                     break
                 if any(text.startswith(s) for s in stop_markers):
                     break
+                # Some boilerplate lines include prefixes/suffixes; match by containment too.
+                upper_text = text.upper()
+                if (
+                    "REMITTANCE RECEIVED" in upper_text
+                    or "DEPOSIT VOUCHER" in upper_text
+                    or "ACCOUNT NUMBER" in upper_text
+                    or "ADDRESS/ORDER" in upper_text
+                    or "ITEMS ON REVERSE" in upper_text
+                    or "CHANGE OF ADDRESS" in upper_text
+                    or "PLEASE PRINT" in upper_text
+                ):
+                    break
                 if text in ("Checking", "Savings"):
                     break
                 if text.startswith("Date Transaction Detail"):
@@ -2318,6 +2342,28 @@ def _parse_nfcu_page(pdf_path: str, page_num: int) -> str:
                     if pending is not None:
                         mvals = [tok for tok in toks if money_re.match(tok)]
                         desc_tokens = [tok for tok in toks if not money_re.match(tok)]
+
+                        # Stop if we have already captured a balance and now hit
+                        # known non-transaction boilerplate lines.
+                        upper = text.upper()
+                        if pending["balance"] and (
+                            "REMITTANCE" in upper
+                            or "DEPOSIT VOUCHER" in upper
+                            or "ACCOUNT NUMBER" in upper
+                            or "ADDRESS/ORDER" in upper
+                            or "ITEMS ON REVERSE" in upper
+                            or "PO BOX" in upper
+                        ):
+                            break
+
+                        # Beginning/Ending balance rows are single logical rows.
+                        # Do not absorb unrelated continuation text.
+                        if (
+                            ("BEGINNING BALANCE" in pending["desc"].upper() or "ENDING BALANCE" in pending["desc"].upper())
+                            and not mvals
+                            and not (toks and date_re.match(toks[0]))
+                        ):
+                            break
                         if desc_tokens:
                             pending["desc"] = (pending["desc"] + " " + " ".join(desc_tokens)).strip()
                         if len(mvals) >= 2 and (not pending["amount"] or not pending["balance"]):
@@ -2571,18 +2617,8 @@ def run_ocr(uploaded_file):
 
             parts = []
 
-            # Fix NF-1: Navy Federal full text-layer parse.
-            # Must run FIRST — before header/OCR/footer — so that when NFCU is
-            # detected, the entire page output is replaced with clean text-layer
-            # data and none of the garbled OCR pipeline runs at all.
-            # Safe no-op for all other banks (_parse_nfcu_page returns "").
-            # NFCU-only path; no probing on non-NFCU files.
-            if is_pdf and is_nfcu_doc:
-                nfcu_html = _parse_nfcu_page(path, page_num)
-                if nfcu_html:
-                    all_pages.append(nfcu_html)
-                    continue  # skip ALL OCR processing for this page
-                log.warning("NFCU parser returned empty output for page %d", page_num + 1)
+            # Keep GLM-OCR as the primary parser for all pages.
+            # Navy-specific helpers run only as additive post-processing.
 
             # Header inclusion: PDF text -> band words -> OCR band
             hdr = ""
@@ -2600,7 +2636,7 @@ def run_ocr(uploaded_file):
                 stabilized = stabilize_tables_and_text(page_md.strip())
                 # Fix 4: restore any rows OCR dropped by cross-checking the PDF text layer.
                 # Safe no-op for scanned PDFs (no text layer) and non-transaction pages.
-                if is_pdf:
+                if is_pdf and not is_nfcu_doc:
                     stabilized = _patch_ocr_with_textlayer(stabilized, path, page_num)
 
                 # Fix 4D: append Johnson Bank Checks + Daily Account Balance