┌─────────────────────┐
│ Customer Message Input │
└──────────┬──────────┘
           │
           ▼
┌─────────────────────┐
│   AI Intent Recognition       │
│   Analyze Customer Needs       │
└──────────┬──────────┘
           │
     ┌─────┼─────┬──────────┐
     │     │     │          │
     ▼     ▼     ▼          ▼
┌───────┐┌───────┐┌───────┐┌───────┐
│Order  ││Product││Complaint││FAQ   │
│Query  ││Recommendation││Ticket││      │
└───┬───┘└───┬───┘└───┬───┘└───┬───┘
    │        │        │        │
    ▼        ▼        ▼        ▼
┌───────┐┌───────┐┌───────┐┌───────┐
│Playwright│ AI   ││Create││Knowledge│
│Query Backend│Analysis││Ticket││Base Matching│
└───┬───┘└───┬───┘└───┬───┘└───┬───┘
    │        │        │        │
    └────────┴────────┴────────┘
                  │
                  ▼
         ┌─────────────────┐
         │ Generate Customer Service Reply │
         │ Save Conversation History │
         └─────────────────┘

User: Help me reply to this customer:
"Hello, why hasn't my order 123456 been shipped yet? It's been 3 days!"

User: Process these customer service messages [message list/file]

Analyze the intent of the following customer message:

【Customer Message】
{message}

【Output Format】
{
  "intent": "order_query|refund|product_inquiry|complaint|promotion|after_sale|chat",
  "confidence": 0.95,
  "entities": {
    "order_id": "123456",
    "product_name": "",
    "emotion": "negative|neutral|positive"
  },
  "urgency": "high|medium|low"
}

// Playwright MCP operation steps

1. Open e-commerce backend
   browser_navigate({ url: "https://seller.taobao.com" })
   // Or other e-commerce platform backends

2. Check login status
   - If not logged in, prompt the user to log in first
   - Save login status for subsequent use

3. Enter order management
   browser_click({ element: "Order Management" })

4. Search for order number
   browser_type({
     element: "Order search box",
     text: "{order_id}"
   })
   browser_click({ element: "Search" })

5. Get order status
   browser_snapshot() // Screenshot to get order information

6. Parse order status
   - Order status (pending shipping/shipped/delivered)
   - Logistics information
   - Estimated arrival time

{
  "order_id": "123456",
  "status": "shipped",
  "status_text": "Shipped, in transit",
  "logistics": {
    "company": "SF Express",
    "tracking_no": "SF1234567890",
    "last_update": "2025-12-28 15:30",
    "location": "Shenzhen Transit Center"
  },
  "estimated_delivery": "2025-12-30"
}

1. Parse customer needs (category, price range, preference)
2. Match product library/search for products
3. Generate recommendation script

【Recommendation Script Template】
Dear customer, according to your needs, we recommend the following products for you:

1️⃣ **{Product Name 1}** - ¥{Price}
   {Product Highlights}

2️⃣ **{Product Name 2}** - ¥{Price}
   {Product Highlights}

Which one do you think is more suitable? Feel free to ask me any questions～

// Get product details from e-commerce platform

1. Open product page
   browser_navigate({ url: "{product_url}" })

2. Get product information
   browser_snapshot()

3. Parse information
   - Product name
   - Price
   - Stock status
   - Specification parameters

{
  "ticket_id": "TK20251229001",
  "type": "complaint",
  "customer_id": "C12345",
  "order_id": "123456",
  "description": "Customer complains about slow logistics",
  "priority": "high",
  "status": "open",
  "created_at": "2025-12-29T10:30:00Z"
}

You are a professional e-commerce customer service representative, please generate a reply based on the following information:

【Customer Message】
{customer_message}

【Intent Analysis】
Intent: {intent}
Emotion: {emotion}
Urgency: {urgency}

【Query Result】
{query_result}

【Reply Requirements】
1. Friendly address (Dear/Hello)
2. Empathize first, then solve the problem
3. Accurate and complete information
4. Gentle and professional tone
5. Apologize actively if there is a problem
6. Ask if there are other needs at the end

【Output Format】
Output the reply content directly, you can use emojis appropriately

{
  "conversation_id": "conv_20251229_001",
  "customer_id": "C12345",
  "messages": [
    {
      "role": "customer",
      "content": "Why hasn't my order been shipped yet?",
      "timestamp": "2025-12-29T10:30:00Z"
    },
    {
      "role": "assistant",
      "content": "Dear customer, I'm very sorry to keep you waiting...",
      "timestamp": "2025-12-29T10:30:05Z",
      "intent": "order_query"
    }
  ]
}

Dear customer, your order has been shipped! 🚚

Logistics information:
📦 Courier company: {company}
📝 Tracking number: {tracking_no}
📍 Current location: {location}
⏰ Estimated delivery: {estimated_delivery}

You can click the order details to view real-time logistics～
Do you have any other questions?

Dear customer, I'm very sorry to keep you waiting! 🙏

Your order is currently being processed urgently and is expected to be shipped before {ship_date}.
We are deeply sorry for the inconvenience caused to you, and we will arrange it for you as soon as possible～

If you are in a hurry, I can help you urge the warehouse!

Dear customer, we have received your refund application～

I have submitted it for processing here:
📋 Refund number: {refund_id}
💰 Refund amount: ¥{amount}
⏰ Expected arrival: 1-3 working days

The refund will be returned to the original account, please pay attention to the account changes then.
Feel free to contact me if you have any questions!

Dear customer, I'm really sorry for the bad experience we brought you! 🙏

I fully understand your mood, this is indeed our problem.
I have already reported your situation to the supervisor, and we will give you a satisfactory solution as soon as possible.

To express our apology, we have applied for a {coupon} coupon for you,
Hope it can make up for a little bit of this unpleasant experience.

Would you mind leaving your contact number? Our supervisor will call you back personally to handle it.

Dear customer, the detailed information of this product is as follows:

📦 {product_name}
💰 Price: ¥{price}
📏 Specification: {specs}
🎁 Gift: {gifts}
📦 Stock: {stock_status}

{product_highlights}

There is also a {promotion} event when you place an order now～
Do you need me to check the size for you?

User: Help me reply: "Where is order 123456?"

Claude:
1. Identify intent: Order query
2. [Optional] Use Playwright to query backend order status
3. Generate reply:
   "Dear customer, your order 123456 has been shipped!
    Courier: SF Express SF1234567890
    Current location: Shenzhen Transit Center
    Expected to arrive tomorrow～ Do you have any other questions?"

User: Batch reply to these 10 customer messages

Claude:
1. Analyze intent one by one
2. Batch query relevant information
3. Generate 10 replies
4. Output results for copy and use

User: Connect to Taobao backend, query the real status of order 123456

Claude:
1. Use Playwright to open Taobao Seller Center
2. Check login status (prompt if login is required)
3. Search for order number
4. Screenshot to get order status
5. Parse and generate reply

platform: taobao
login_url: https://login.taobao.com
seller_url: https://myseller.taobao.com
order_path: /home.htm#/order-manage
search_selector: "#keyword"

platform: jd
login_url: https://passport.jd.com
seller_url: https://shop.jd.com
order_path: /order/list
search_selector: ".search-input"

platform: shopify
login_url: https://{store}.myshopify.com/admin
order_path: /admin/orders
search_selector: "#search-query"