Update app.py

app.py

@@ -23,85 +23,117 @@ try:
 except AttributeError:
     RESAMPLE = Image.LANCZOS
 
-PROMPT_FREIGHT_JSON = """
-Please analyze the freight rate tables in the file I provide. This file may contain **multiple airlines' tariffs**. Your task is to extract **one JSON object per airline**, using the following schema:
-
-[
-  {
-    "shipping_line": "...",
-    "shipping_line_code": "...",
-    "shipping_line_reason": "Why this carrier is chosen?",
-    "fee_type": "Air Freight",
-    "valid_from": ...,
-    "valid_to": ...,
-    "charges": [
-      {
-        "frequency": "...",
-        "package_type": "...",
-        "aircraft_type": "...",
-        "direction": "Export or Import or null",
-        "origin": "...",
-        "destination": "...",
-        "charge_name": "...",
-        "charge_code": "...",
-        "charge_code_reason": "...",
-        "cargo_type": "...",
-        "currency": "...",
-        "transit": "...",
-        "transit_time": "...",
-        "weight_breaks": {
-          "M": ...,
-          "N": ...,
-          "+45kg": ...,
-          "+100kg": ...,
-          "+300kg": ...,
-          "+500kg": ...,
-          "+1000kg": ...,
-          "other": {
-            key: value
-          },
-          "weight_breaks_reason":"Why chosen weight_breaks?"
-        },
-        "remark": "..."
-      }
-    ],
-    "local_charges": [
-      {
-        "charge_name": "...",
-        "charge_code": "...",
-        "unit": "...",
-        "amount": ...,
-        "remark": "..."
-      }
-    ]
+PROMPT_MULTI_AIRLINE_JSON = """
+You are an expert in air freight rate extraction and normalization.
+
+The document contains rate information for multiple airlines.  
+Please analyze all content (tables, headers, notes) and return **a list of JSON objects**, each representing a separate airline.
+
+Each airline should follow this schema:
+
+{
+  "shipping_line": "...",
+  "shipping_line_code": "...",
+  "shipping_line_reason": "Why this carrier is chosen?",
+  "fee_type": "Air Freight",
+  "valid_from": "...",
+  "valid_to": "...",
+  "charges": [ ... ],             # List of charge objects (see below)
+  "local_charges": [ ... ]        # Optional local charges if available
+}
+
+Each `charges` object must follow this schema:
+
+{
+  "frequency": "...",
+  "package_type": "...",                  # e.g. Carton, Pallet, Skid
+  "aircraft_type": "...",
+  "direction": "Export / Import / null",
+  "origin": "...",
+  "destination": "...",
+  "charge_name": "...",
+  "charge_code": "GCR / PER / DGR / etc.",
+  "charge_code_reason": "...",
+  "cargo_type": "...",
+  "currency": "...",
+  "transit": "...",
+  "transit_time": "...",
+  "weight_breaks": {
+    "M": ...,
+    "N": ...,
+    "+45kg": ...,
+    "+100kg": ...,
+    "+300kg": ...,
+    "+500kg": ...,
+    "+1000kg": ...,
+    "other": { key: value },
+    "weight_breaks_reason": "Why chosen weight_breaks?"
   },
-  ...
-]
-
-### Date rules
-- valid_from format:
-  - `DD/MM/YYYY` (if full date)
-  - `01/MM/YYYY` (if month+year only)
-  - `01/01/YYYY` (if year only)
+  "remark": "..."
+}
+
+Each `local_charges` object:
+
+{
+  "charge_name": "...",
+  "charge_code": "...",
+  "unit": "...",
+  "amount": ...,
+  "remark": "..."
+}
+
+---
+
+### ✈️ Airline Separation Logic:
+- If multiple airlines are detected in the document, separate each section and return a distinct JSON object per airline.
+- Infer `shipping_line` and `shipping_line_code` from the header (e.g. "AIR CHINA CARGO (CA)" → name = "AIR CHINA CARGO", code = "CA").
+- Each JSON object must include only data relevant to that airline.
+
+---
+
+### 💡 Date rules:
+- valid_from:
+  - `DD/MM/YYYY` if exact
+  - `01/MM/YYYY` if only month/year
+  - `01/01/YYYY` if only year
   - `UFN` if missing
 - valid_to:
   - exact `DD/MM/YYYY` if present
   - else `UFN`
 
-### STRICT RULES:
-- Return a JSON **array** of airline objects (not just one).
-- All rates must exactly match the corresponding weight break columns (M,N,45kg, 100kg, 300kg, 500kg, 1000kg, etc.). Set null if N/A. No assumptions or interpolations.
-- If the table shows "RQ" or similar, set value as "RQST".
-- Group same-price destinations into one record separated by "/".
-- Always use IATA code for origin and destination.
-- Flight number (e.g. ZH118) is not charge code.
-- Frequency: D[1-7]; 'Daily' = D1234567. Join multiple (e.g. D3,D4→D34).
-- If local charges exist, list them.
-- If validity is missing, set null.
-- Direction: Export if origin is Vietnam (SGN, HAN, DAD...), else Import.
-- Provide short plain English reasons for "shipping_line_reason" & "charge_code_reason".
-- Replace commas in remarks with semicolons.
-- **Only return valid JSON. No text explanation. No markdown.**
+---
+
+### 📦 Package and Surcharge Logic:
+Apply these when the remark or note indicates such rules:
+
+1. **Default case**: If no package mentioned → `"Carton"` is the default.
+2. **“Carton = Pallet”**: Duplicate rates with `package_type="Pallet"`.
+3. **“SKID shipment: add 10 cents (GEN & PER)”**: Add new charges with `+0.10 USD/kg` for GEN/PER, with `package_type="Pallet"` or `"Skid"`.
+4. **EU vs Non-EU surcharges**: If different pallet surcharges by region → split charges accordingly.
+5. **“All-in” or “inclusive of MY and SC”**: Record `FSC` and `WSC` as `local_charges` with `"NIL"` amount.
+6. **Flight number is not a charge code**. Always use standard cargo code (GCR, PER, etc.).
+
+---
+
+### ⚙️ Other Business Rules:
+- RQ / Request → "RQST"
+- Combine same-rate destinations using `/`
+- Always use **IATA code** for origin/destination
+- Direction = Export if origin is in Vietnam (SGN, HAN, DAD), else Import
+- Frequency:
+  - D[1-7] = day of week
+  - "Daily" = D1234567
+- Remarks: Replace `,` with `;`
+- Add meaningful `"shipping_line_reason"` and `"charge_code_reason"`
+
+---
+
+### 🚨 STRICT OUTPUT:
+- Return **a JSON array**, where each item is a full airline object
+- Do NOT return markdown or explanation
+- All fields must be valid
+- All numbers = numeric types
+- Use `null` if value missing
 
 """
 
@@ -253,7 +285,7 @@ def run_process(file, question, model_choice, temperature, top_p, external_api_u
             check_result = check_pdf_structure(file_bytes)
             print(f"[PDF Check] {filename}: {check_result}")
 
-            if check_result == "có" and 1=2: # bỏ qua if này test thử prompt nhiều hãng
+            if check_result == "có" and 1==2: # bỏ qua if này test thử prompt nhiều hãng
                 try:
                     print("➡️ PDF có nhiều cột/nhiều trang → dùng pdfplumber extract trước rồi Gemini.")
                     all_dfs = []