Update app.py

app.py

@@ -118,6 +118,11 @@ MIN_CROP_PIXELS = 112 * 112
 PAGE_PNG_COMPRESS_LEVEL = 3
 # JPEG quality for small header/footer crops sent to the API.
 ZONE_JPEG_QUALITY = 95
+PAGE_JPEG_QUALITY = 88
+MAX_PAGE_UPLOAD_BYTES = 9_200_000
+# Keep page PNGs under MaaS input constraints to avoid 400 errors.
+MAX_PAGE_LONG_SIDE = 3600
+MAX_PAGE_TOTAL_PIXELS = 7_800_000
 
 MIN_PDF_TEXT_CHARS_NATIVE_LAYER = 1500
 
@@ -1363,6 +1368,298 @@ def normalize_amount_only_transaction_tables(md: str) -> str:
     )
 
 
+def infer_credit_debit_from_balance_deltas(md: str) -> str:
+    """
+    When OCR leaves Credit/Debit empty but a running Balance column is present,
+    infer signed movement from consecutive balance deltas (same for all banks).
+    """
+    if not md or "<table" not in md.lower():
+        return md
+
+    def repl_table(m: re.Match) -> str:
+        full = m.group(0)
+        rows = _extract_rows_plain_from_table(full)
+        if len(rows) < 3:
+            return full
+
+        hdr = [c.strip() for c in rows[0]]
+        hdr_l = [h.lower() for h in hdr]
+        date_i = next((i for i, h in enumerate(hdr_l) if "date" in h), None)
+        desc_i = next(
+            (i for i, h in enumerate(hdr_l) if "description" in h or "memo" in h or "details" in h),
+            None,
+        )
+        credit_i = next((i for i, h in enumerate(hdr_l) if "credit" in h), None)
+        debit_i = next((i for i, h in enumerate(hdr_l) if "debit" in h), None)
+        balance_i = next((i for i, h in enumerate(hdr_l) if "balance" in h), None)
+
+        if date_i is None or desc_i is None or credit_i is None or debit_i is None or balance_i is None:
+            return full
+
+        max_len = max(len(hdr), max((len(r) for r in rows), default=0))
+        body_bals: List[Optional[float]] = []
+        body_rows: List[List[str]] = []
+        for r in rows[1:]:
+            padded = (r + [""] * (max_len - len(r)))[:max_len]
+            dt = (padded[date_i] or "").strip()
+            if not _is_date_like(dt):
+                continue
+            cr = _parse_amount_or_none((padded[credit_i] or "").strip()) if credit_i < len(padded) else None
+            db = _parse_amount_or_none((padded[debit_i] or "").strip()) if debit_i < len(padded) else None
+            bal = _parse_amount_or_none((padded[balance_i] or "").strip()) if balance_i < len(padded) else None
+            body_rows.append(padded)
+            body_bals.append(bal)
+
+        if len(body_rows) < 3:
+            return full
+
+        empty_cd = 0
+        for i, pr in enumerate(body_rows):
+            cr = _parse_amount_or_none((pr[credit_i] or "").strip()) if credit_i < len(pr) else None
+            db = _parse_amount_or_none((pr[debit_i] or "").strip()) if debit_i < len(pr) else None
+            if cr is None and db is None:
+                empty_cd += 1
+        if empty_cd < max(3, int(0.6 * len(body_rows))):
+            return full
+
+        new_body: List[List[str]] = []
+        for i, pr in enumerate(body_rows):
+            row = pr[:]
+            cr0 = _parse_amount_or_none((row[credit_i] or "").strip()) if credit_i < len(row) else None
+            db0 = _parse_amount_or_none((row[debit_i] or "").strip()) if debit_i < len(row) else None
+            if cr0 is not None or db0 is not None:
+                new_body.append(row)
+                continue
+            if i == 0 or body_bals[i] is None:
+                new_body.append(row)
+                continue
+            prev_b = body_bals[i - 1]
+            cur_b = body_bals[i]
+            if prev_b is None or cur_b is None:
+                new_body.append(row)
+                continue
+            delta = cur_b - prev_b
+            if abs(delta) < 1e-6:
+                new_body.append(row)
+                continue
+            tol = max(0.55, 0.025 * abs(delta))
+            if abs(delta) <= tol:
+                new_body.append(row)
+                continue
+            if delta > 0:
+                row[credit_i] = _fmt_money(delta)
+            else:
+                row[debit_i] = _fmt_money(-delta)
+            new_body.append(row)
+
+        filled = 0
+        for i, pr in enumerate(new_body):
+            cr = _parse_amount_or_none((pr[credit_i] or "").strip()) if credit_i < len(pr) else None
+            db = _parse_amount_or_none((pr[debit_i] or "").strip()) if debit_i < len(pr) else None
+            if cr is not None or db is not None:
+                filled += 1
+        if filled < max(2, int(0.35 * len(new_body))):
+            return full
+
+        html_rows = ["<tr>" + "".join(f"<th>{html.escape(c)}</th>" for c in hdr) + "</tr>"]
+        for pr in new_body:
+            cells = (pr + [""] * (len(hdr) - len(pr)))[: len(hdr)]
+            html_rows.append(
+                "<tr>" + "".join(f"<td>{html.escape((c or '').strip())}</td>" for c in cells) + "</tr>"
+            )
+        return "<table>\n" + "\n".join(html_rows) + "\n</table>"
+
+    return re.sub(
+        r"<table\b[^>]*>.*?</table>",
+        repl_table,
+        md,
+        flags=re.IGNORECASE | re.DOTALL,
+    )
+
+
+def sanitize_transaction_tables(md: str) -> str:
+    """
+    Final defensive cleanup for transaction-like tables:
+    - remove in-body header repeats and subtotal rows
+    - deduplicate repeated rows
+    - correct likely credit/debit direction using row text and balance deltas
+    """
+    if not md or "<table" not in md.lower():
+        return md
+
+    credit_kw = re.compile(
+        r"\b(deposit|deposits|credit|credits|recd|received|return|refund|interest|rtp\s*rcvd|pmt\s*cr)\b",
+        re.IGNORECASE,
+    )
+    debit_kw = re.compile(
+        r"\b(debit|debits|withdrawal|withdrawals|check|checks|payment|purchase|charge|fee|ach)\b",
+        re.IGNORECASE,
+    )
+
+    def parse_money(txt: str):
+        t = (txt or "").strip()
+        if not t:
+            return None
+        return _parse_amount_or_none(t)
+
+    def fmt_money(v: float) -> str:
+        return f"{abs(v):,.2f}"
+
+    def repl_table(m: re.Match) -> str:
+        full = m.group(0)
+        rows = _extract_rows_plain_from_table(full)
+        if len(rows) < 2:
+            return full
+
+        hdr = [c.strip() for c in rows[0]]
+        hdr_l = [h.lower() for h in hdr]
+        date_i = next((i for i, h in enumerate(hdr_l) if "date" in h), None)
+        desc_i = next((i for i, h in enumerate(hdr_l) if "description" in h or "memo" in h or "details" in h), None)
+        credit_i = next((i for i, h in enumerate(hdr_l) if "credit" in h), None)
+        debit_i = next((i for i, h in enumerate(hdr_l) if "debit" in h), None)
+        amount_i = next((i for i, h in enumerate(hdr_l) if "amount" in h), None)
+        balance_i = next((i for i, h in enumerate(hdr_l) if "balance" in h), None)
+        is_txn = (date_i is not None) and (desc_i is not None) and (
+            credit_i is not None or debit_i is not None or amount_i is not None
+        )
+        if not is_txn:
+            return full
+        # Without a real balance column, do not rewrite the table (avoids blank balances).
+        if balance_i is None:
+            return full
+
+        max_cols = max(len(hdr), max((len(r) for r in rows), default=len(hdr)))
+        if max_cols < 4:
+            max_cols = 4
+
+        new_hdr = hdr[:]
+        if amount_i is not None and credit_i is None and debit_i is None:
+            new_hdr[amount_i] = "Debit"
+            debit_i = amount_i
+            amount_i = None
+        if credit_i is None:
+            new_hdr.append("Credit")
+            credit_i = len(new_hdr) - 1
+        if debit_i is None:
+            new_hdr.append("Debit")
+            debit_i = len(new_hdr) - 1
+
+        records = []
+        seen = set()
+        for r in rows[1:]:
+            padded = r + [""] * (len(new_hdr) - len(r))
+            low = [c.strip().lower() for c in padded]
+            joined = " ".join(low)
+
+            # Remove duplicate header rows inside body and subtotal/total lines.
+            if any(
+                x in low
+                for x in ("date", "posting date", "description", "amount", "credit", "debit", "balance")
+            ):
+                continue
+            if "subtotal" in joined or joined.startswith("total "):
+                continue
+
+            dt = padded[date_i].strip() if date_i is not None and date_i < len(padded) else ""
+            if not _is_date_like(dt):
+                continue
+            desc = padded[desc_i].strip() if desc_i is not None and desc_i < len(padded) else ""
+            bal = parse_money(padded[balance_i]) if balance_i < len(padded) else None
+            cr = parse_money(padded[credit_i]) if credit_i < len(padded) else None
+            db = parse_money(padded[debit_i]) if debit_i < len(padded) else None
+
+            if cr is None and db is None:
+                continue
+
+            amt = cr if cr is not None else db
+            direction = "credit" if cr is not None else "debit"
+
+            # Row-level lexical correction
+            if amt is not None:
+                if credit_kw.search(desc) and not debit_kw.search(desc):
+                    direction = "credit"
+                elif debit_kw.search(desc) and not credit_kw.search(desc):
+                    direction = "debit"
+
+            key = (dt, desc, f"{amt:.2f}" if isinstance(amt, (int, float)) else "", f"{bal:.2f}" if isinstance(bal, (int, float)) else "")
+            if key in seen:
+                continue
+            seen.add(key)
+            records.append(
+                {
+                    "row": padded[: len(new_hdr)],
+                    "amount": abs(amt) if amt is not None else None,
+                    "balance": bal,
+                    "lex_dir": direction,
+                }
+            )
+
+        if not records:
+            return full
+
+        # Choose balance interpretation direction (forward vs reverse row order)
+        # based on which one produces more valid delta-to-amount matches.
+        def delta_match(delta: float, amt_val: float) -> str:
+            tol = max(0.75, 0.03 * abs(amt_val))
+            if abs(delta - abs(amt_val)) <= tol:
+                return "credit"
+            if abs(delta + abs(amt_val)) <= tol:
+                return "debit"
+            return ""
+
+        fwd_matches = 0
+        rev_matches = 0
+        prev_bal = None
+        for rec in records:
+            bal = rec["balance"]
+            amt_val = rec["amount"]
+            if bal is None or amt_val is None:
+                continue
+            if prev_bal is not None:
+                if delta_match(bal - prev_bal, amt_val):
+                    fwd_matches += 1
+                if delta_match(prev_bal - bal, amt_val):
+                    rev_matches += 1
+            prev_bal = bal
+
+        use_reverse = rev_matches > fwd_matches
+
+        body = []
+        prev_bal = None
+        for rec in records:
+            out = rec["row"][:]
+            amt_val = rec["amount"]
+            bal = rec["balance"]
+            direction = rec["lex_dir"]
+
+            if amt_val is not None and bal is not None and prev_bal is not None:
+                delta = (prev_bal - bal) if use_reverse else (bal - prev_bal)
+                d = delta_match(delta, amt_val)
+                if d:
+                    direction = d
+            if bal is not None:
+                prev_bal = bal
+
+            out[credit_i] = fmt_money(amt_val) if (amt_val is not None and direction == "credit") else ""
+            out[debit_i] = fmt_money(amt_val) if (amt_val is not None and direction == "debit") else ""
+            out[balance_i] = f"{bal:,.2f}" if bal is not None else out[balance_i]
+            body.append(out)
+
+        if not body:
+            return full
+        html_rows = ["<tr>" + "".join(f"<th>{html.escape(c)}</th>" for c in new_hdr) + "</tr>"]
+        for rr in body:
+            html_rows.append("<tr>" + "".join(f"<td>{html.escape(c)}</td>" for c in rr) + "</tr>")
+        return "<table>\n" + "\n".join(html_rows) + "\n</table>"
+
+    return re.sub(
+        r"<table\b[^>]*>.*?</table>",
+        repl_table,
+        md,
+        flags=re.IGNORECASE | re.DOTALL,
+    )
+
+
 def _extract_daily_balance_by_date(md: str):
     """
     Build a date->balance map from daily-balance style tables.
@@ -1761,8 +2058,44 @@ def render_pdf_pages_to_images(pdf_path: str) -> Tuple[List[str], List[int]]:
             canvas.paste(img, (pad_l, pad_t))
             img = canvas
 
-        img_path = os.path.join(tempfile.gettempdir(), f"glmocr_page_{os.getpid()}_{i}.png")
+        # Deterministic guardrail: downscale oversized pages before upload.
+        iw, ih = img.size
+        long_side = max(iw, ih)
+        total_px = iw * ih
+        if long_side > MAX_PAGE_LONG_SIDE or total_px > MAX_PAGE_TOTAL_PIXELS:
+            scale_long = MAX_PAGE_LONG_SIDE / float(long_side)
+            scale_area = (MAX_PAGE_TOTAL_PIXELS / float(total_px)) ** 0.5
+            scale = min(scale_long, scale_area, 1.0)
+            nw = max(900, int(iw * scale))
+            nh = max(900, int(ih * scale))
+            img = img.resize((nw, nh), Image.Resampling.LANCZOS)
+
+        base_path = os.path.join(tempfile.gettempdir(), f"glmocr_page_{os.getpid()}_{i}")
+        img_path = f"{base_path}.png"
         img.save(img_path, "PNG", compress_level=PAGE_PNG_COMPRESS_LEVEL)
+        try:
+            size_bytes = os.path.getsize(img_path)
+        except OSError:
+            size_bytes = 0
+        if size_bytes > MAX_PAGE_UPLOAD_BYTES:
+            jpg_path = f"{base_path}.jpg"
+            quality = PAGE_JPEG_QUALITY
+            work = img
+            for _ in range(5):
+                work.save(jpg_path, "JPEG", quality=quality, optimize=True)
+                try:
+                    jsize = os.path.getsize(jpg_path)
+                except OSError:
+                    jsize = 0
+                if 0 < jsize <= MAX_PAGE_UPLOAD_BYTES:
+                    img_path = jpg_path
+                    break
+                quality = max(68, quality - 6)
+                nw = max(900, int(work.width * 0.93))
+                nh = max(900, int(work.height * 0.93))
+                work = work.resize((nw, nh), Image.Resampling.LANCZOS)
+            else:
+                img_path = jpg_path
         page_images.append(img_path)
         page_heights.append(img.height)
 
@@ -1874,6 +2207,10 @@ def run_ocr(uploaded_file):
             merged = strip_malformed_leading_table_rows(merged)
             merged = strip_subtotal_rows_from_transaction_tables(merged)
             merged = mask_non_monetary_long_numbers(merged)
+            merged = normalize_ambiguous_amount_balance_tables(merged)
+            merged = normalize_amount_only_transaction_tables(merged)
+            merged = infer_credit_debit_from_balance_deltas(merged)
+            merged = sanitize_transaction_tables(merged)
             merged = stabilize_table_markup(merged, rounds=4)
             merged = enrich_transaction_tables_with_daily_balances(merged)
             # Keep transaction amounts untouched; add explicit snapshot for metadata extraction.
@@ -1892,7 +2229,11 @@ def run_ocr(uploaded_file):
     finally:
         for p in page_images:
             try:
-                if isinstance(p, str) and p.endswith(".png") and "glmocr_page_" in os.path.basename(p):
+                if (
+                isinstance(p, str)
+                and "glmocr_page_" in os.path.basename(p)
+                and (p.endswith(".png") or p.endswith(".jpg") or p.endswith(".jpeg"))
+            ):
                     os.unlink(p)
             except Exception:
                 pass