Update src/streamlit_app.py

src/streamlit_app.py

@@ -1,6 +1,6 @@
 # =========================
 # Invoice Extractor (Qwen3-VL via RunPod vLLM) - Batch Mode with Tax Validation
-# UPDATED: Fixed tax percentage parsing - no thousands separator logic
+# UPDATED: Fixed tax percentage parsing + time format quantities + scrolling issue
 # =========================
 import os
 from pathlib import Path
@@ -88,6 +88,68 @@ def ensure_state(k: str, default):
     if k not in st.session_state:
         st.session_state[k] = default
 
+def parse_time_to_minutes(x) -> float:
+    """
+    Parse time format quantities to minutes.
+    
+    Examples:
+        "0:35"  → 35.0   (0 hours, 35 minutes = 35 minutes)
+        "1:30"  → 90.0   (1 hour, 30 minutes = 90 minutes)
+        "2:15"  → 135.0  (2 hours, 15 minutes = 135 minutes)
+        "0:05"  → 5.0    (5 minutes)
+        "123"   → 123.0  (regular number, not time format)
+        "1.5"   → 1.5    (regular decimal, not time format)
+    """
+    if x is None:
+        return 0.0
+    if isinstance(x, (int, float)):
+        return float(x)
+    
+    s = str(x).strip()
+    if s == "":
+        return 0.0
+    
+    # Check if it's in time format (H:MM or HH:MM)
+    time_pattern = r'^(\d+):(\d{1,2})$'
+    match = re.match(time_pattern, s)
+    
+    if match:
+        hours = int(match.group(1))
+        minutes = int(match.group(2))
+        total_minutes = (hours * 60) + minutes
+        return float(total_minutes)
+    
+    # Not time format, treat as regular number
+    return 0.0
+
+def clean_quantity(x) -> float:
+    """
+    Parse quantity - handles both time format (H:MM) and regular numbers.
+    
+    Examples:
+        "0:35"     → 35.0   (time format: 35 minutes)
+        "1:30"     → 90.0   (time format: 90 minutes)
+        "123"      → 123.0  (regular number)
+        "1,234.56" → 1234.56 (US format with decimals)
+        "1.234,56" → 1234.56 (EUR format with decimals)
+    """
+    if x is None:
+        return 0.0
+    if isinstance(x, (int, float)):
+        return float(x)
+    
+    s = str(x).strip()
+    if s == "":
+        return 0.0
+    
+    # First check if it's time format (H:MM or HH:MM)
+    time_value = parse_time_to_minutes(s)
+    if time_value > 0.0:
+        return time_value
+    
+    # Not time format, use regular number parsing
+    return clean_float(s)
+
 def clean_tax_percentage(x) -> float:
     """
     Parse tax percentage - ALWAYS treats periods as decimals, never as thousands.
@@ -797,7 +859,7 @@ def parse_vllm_json(raw_json_text):
             result["Itemized Data"].append({
                 "Description": item.get("descriptions", ""),
                 "SKU": item.get("SKU", ""),
-                "Quantity": clean_amount(item.get("quantity", "0")),
+                "Quantity": clean_quantity(item.get("quantity", "0")),  # ✅ USE clean_quantity for time format support
                 "Unit Price": clean_amount(item.get("unit_price", "0")),
                 "Amount": clean_amount(item.get("amount", "0")),
                 "Tax": clean_amount(raw_tax),
@@ -1059,7 +1121,7 @@ def map_prediction_to_ui(pred):
 
         return {
             "Description": str(desc).strip(),
-            "Quantity": float(clean_number(qty)),
+            "Quantity": float(clean_quantity(qty)),  # ✅ USE clean_quantity for time format support
             "Unit Price": float(clean_number(unit_price)),
             "Amount": float(clean_number(amount)),
             "Tax": float(clean_number(tax)),
@@ -1190,7 +1252,7 @@ def map_prediction_to_ui(pred):
         if desc or amt or qty or unit_price:
             items_rows.append({
                 "Description": str(desc or ""),
-                "Quantity": float(clean_number(qty)),
+                "Quantity": float(clean_quantity(qty)),  # ✅ USE clean_quantity for time format support
                 "Unit Price": float(clean_number(unit_price)),
                 "Amount": float(clean_number(amt)),
                 "Tax": float(clean_number(pick_first("tax", "tax_amount") or 0.0)),
@@ -1617,10 +1679,11 @@ elif len(st.session_state.batch_results) > 0:
                 if st.session_state.get(f"Currency_{selected_hash}") == 'Other':
                     st.text_input("Specify Currency", key=f"Currency_Custom_{selected_hash}")
 
-                st.number_input("Subtotal", key=f"Subtotal_{selected_hash}", format="%.2f")
-                st.number_input("Tax %", key=f"Tax Percentage_{selected_hash}", format="%.4f")
-                st.number_input("Total Tax", key=f"Total Tax_{selected_hash}", format="%.2f")
-                st.number_input("Total Amount", key=f"Total Amount_{selected_hash}", format="%.2f")
+                # ✅ FIX: Add step=None to prevent scroll interference
+                st.number_input("Subtotal", key=f"Subtotal_{selected_hash}", format="%.2f", step=None)
+                st.number_input("Tax %", key=f"Tax Percentage_{selected_hash}", format="%.4f", step=None)
+                st.number_input("Total Tax", key=f"Total Tax_{selected_hash}", format="%.2f", step=None)
+                st.number_input("Total Amount", key=f"Total Amount_{selected_hash}", format="%.2f", step=None)
 
             with tabs[1]:
                 st.text_input("Sender Name", key=f"Sender Name_{selected_hash}")