Update app.py

Make prep list and replies deterministic

app.py

@@ -4,7 +4,6 @@ import gradio as gr
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
-# MODEL_ID = "Qwen/Qwen2.5-0.5B-Instruct"
 MODEL_ID = "Qwen/Qwen2.5-1.5B-Instruct"
 
 ORDER_COLUMNS = [
@@ -23,32 +22,6 @@ tokenizer = AutoTokenizer.from_pretrained(MODEL_ID)
 model = AutoModelForCausalLM.from_pretrained(MODEL_ID, torch_dtype=torch.float32)
 model.eval()
 
-# SYSTEM_PROMPT = """
-# You extract customer orders from messy DMs for tiny sellers.
-# Return only valid JSON with this exact shape:
-# {
-#   "orders": [
-#     {
-#       "customer": "",
-#       "item": "",
-#       "quantity": "",
-#       "flavor": "",
-#       "pickup_time": "",
-#       "delivery_address": "",
-#       "payment_status": "",
-#       "notes": "",
-#       "missing_fields": []
-#     }
-#   ],
-#   "prep_list": [],
-#   "reply_drafts": [
-#     {"customer": "", "reply": ""}
-#   ]
-# }
-# Use empty strings for unknown values. Put missing details in missing_fields.
-# """
-
-
 
 SYSTEM_PROMPT = """
 You are a careful order extraction engine for tiny sellers.
@@ -69,25 +42,26 @@ Extract customer orders from messy DMs. Return only valid JSON with this exact s
     }
   ],
   "prep_list": [],
-  "reply_drafts": [
-    {"customer": "", "reply": ""}
-  ]
+  "reply_drafts": []
 }
 
 Critical rules:
-- Do not invent customer names.
-- Do not change customer names.
-- Do not merge different customers.
+- Treat each line as one separate customer message.
+- The text before the first ":" is the customer name.
+- Copy customer names exactly as written. Do not uppercase or lowercase them.
+- Never copy details from one customer's message into another customer's order.
 - Include every customer message that looks like an order or possible order.
-- Use only facts explicitly present in the messages.
+- Use only facts explicitly present in that customer's own message.
 - If a value is unknown, use an empty string.
-- Do not add order_id or total_cost unless the message mentions them.
-- For pickup orders, put the pickup time in pickup_time and leave delivery_address empty unless a real address is given.
-- If the customer is unsure, still include the order and put uncertainty in notes.
-- missing_fields should only include practical fields the seller needs to ask for, such as quantity, flavor, pickup_time, delivery_address, or payment_status.
-- Reply drafts should be short, friendly, and ask only for missing information.
+- Do not add order_id or total_cost.
+- For pickup orders, put pickup time in pickup_time. Put a pickup place or delivery address in delivery_address.
+- If the customer is unsure, still include the order and describe the uncertainty in notes.
+- missing_fields should only include fields the seller needs to ask for: quantity, flavor, pickup_time, delivery_address, payment_status.
+- Always set prep_list to [].
+- Always set reply_drafts to [].
 """
 
+
 EXAMPLE_INPUT = """Maya: Hi! Can I get 2 dozen cupcakes for Saturday morning? Half vanilla, half chocolate.
 Sam: Need 1 birthday cake, chocolate, for pickup Friday 5pm. I can pay Venmo.
 Lena: Do you still have lemon bars? I need some for tomorrow but not sure how many yet.
@@ -134,6 +108,79 @@ def format_replies(replies):
         lines.append(f"**{customer}**\n\n{text}")
     return "### Reply drafts\n\n" + "\n\n---\n\n".join(lines)
 
+def text_value(value):
+    if isinstance(value, list):
+        return ", ".join(str(v) for v in value if str(v).strip())
+    if value is None:
+        return ""
+    return str(value).strip()
+
+def missing_list(order):
+    raw = order.get("missing_fields", [])
+    if isinstance(raw, str):
+        fields = [part.strip() for part in raw.split(",") if part.strip()]
+    else:
+        fields = [str(part).strip() for part in raw if str(part).strip()]
+
+    item = text_value(order.get("item"))
+    if item and not text_value(order.get("quantity")):
+        fields.append("quantity")
+    if item and not text_value(order.get("pickup_time")) and not text_value(order.get("delivery_address")):
+        fields.append("pickup_time")
+    if item and not text_value(order.get("payment_status")):
+        fields.append("payment_status")
+
+    return sorted(set(fields))
+
+def build_prep_list(data):
+    items = []
+    for order in data.get("orders", []):
+        item = text_value(order.get("item"))
+        if not item:
+            continue
+
+        customer = text_value(order.get("customer")) or "customer"
+        quantity = text_value(order.get("quantity")) or "quantity to confirm"
+        flavor = text_value(order.get("flavor"))
+
+        line = f"{quantity} {item}"
+        if flavor:
+            line += f" ({flavor})"
+        line += f" - {customer}"
+        items.append(line)
+
+    return items
+
+def build_reply_drafts(data):
+    replies = []
+    labels = {
+        "quantity": "quantity",
+        "flavor": "flavor",
+        "pickup_time": "pickup or delivery time",
+        "delivery_address": "pickup place or delivery address",
+        "payment_status": "payment status",
+    }
+
+    for order in data.get("orders", []):
+        customer = text_value(order.get("customer")) or "there"
+        item = text_value(order.get("item")) or "order"
+        quantity = text_value(order.get("quantity"))
+        flavor = text_value(order.get("flavor"))
+        missing = [labels.get(field, field) for field in missing_list(order)]
+
+        if missing:
+            needed = ", ".join(missing)
+            reply = f"Thanks, {customer}! I have your {item} order. Could you confirm the {needed}?"
+        else:
+            summary = f"{quantity} {item}".strip()
+            if flavor:
+                summary += f" ({flavor})"
+            reply = f"Thanks, {customer}! Confirming your order: {summary}."
+
+        replies.append({"customer": customer, "reply": reply})
+
+    return replies
+
 def analyze_messages(messages):
     if not messages.strip():
         return pd.DataFrame(columns=ORDER_COLUMNS), "Paste some DMs first.", "", ""
@@ -172,8 +219,14 @@ def analyze_messages(messages):
         )
 
     orders_df = normalize_orders(data)
-    prep = format_list("Prep list", data.get("prep_list", []))
-    replies = format_replies(data.get("reply_drafts", []))
+    auto_prep = build_prep_list(data)
+    auto_replies = build_reply_drafts(data)
+    
+    data["prep_list"] = auto_prep
+    data["reply_drafts"] = auto_replies
+    
+    prep = format_list("Prep list", auto_prep)
+    replies = format_replies(auto_replies)
     raw = json.dumps(data, indent=2, ensure_ascii=False)
     return orders_df, prep, replies, raw