Update app.py

app.py

@@ -207,6 +207,20 @@ def ocr_zone(image_path, y_start_frac, y_end_frac):
     return ""
 
 
+def fix_account_number(page_md, hdr):
+    """If page_md has 'Account Number:' with no value after it,
+    find the account number from header text and inject it."""
+    import re
+    if not page_md or not hdr:
+        return page_md
+    if not re.search(r'Account Number:\s*\S', page_md):
+        match = re.search(r'(\d{10,})', hdr)
+        if match:
+            acct = match.group(1)
+            page_md = page_md.replace('Account Number:', 'Account Number: ' + acct)
+    return page_md
+
+
 def close_unclosed_html(md):
     """Close any unclosed <table>, <tbody>, <tr> tags to prevent layout bleed."""
     if not md:
@@ -301,6 +315,7 @@ def run_ocr(uploaded_file):
                     parts.append(hdr.strip())
 
             if page_md:
+                page_md = fix_account_number(page_md, hdr)
                 parts.append(close_unclosed_html(page_md))
 
             # Footer extraction removed — MaaS API already captures all footer