Dianping Information Query Skill

Overview

Core Function Scope

Supported Query Types

• Merchant Search: Search merchants by keyword, location, category
• Business District Query: Support overall information query of popular business districts (such as Muslim Quarter, Nanjing Road, etc.)
• Merchant Details: Obtain basic merchant information (name, address, phone number, business hours)
• Rating Information: Obtain overall rating, sub-item ratings for taste/environment/service/ingredients (newly added ingredient rating support)
• Extended Information: Number of reviews, average consumption, cuisine category, business district location, group booking/reservation mark
• City Switching: Support flexible switching between all Dianping-covered cities across the country
• Deep Information Query: Obtain recommended dish lists, user review content, review tags and high-frequency mention counts, business hours and other deep information
• Multi-Tab Management: Automatically handle new tabs generated during the search process to ensure operational continuity

Unsupported Functions

• Personalized content requiring in-depth login verification (such as personal favorites, history records)
• Batch automated queries (mainly single queries)

Login State Detection and Processing

Dual-State Detection Mechanism

1. Not Logged-in State Detection:
   • Check if the text "Please Login/Register" exists in the upper right corner of the page
   • Check if the URL contains

     account.dianping.com/pclogin

   • Find login-related QR code images or form elements
2. Logged-in State Detection:
   • Check if the username element (

     .nick-name

     or

     .username

     ) exists
   • Verify if the user avatar element (

     .user-face

     ) exists
   • Confirm if the personal center link is accessible

User Guidance Process

1. Immediately pause automated operations
2. Inform the user that manual login is required
3. Provide specific login guidance (mainly QR code scan login)
4. Wait for the user to confirm login completion before continuing the query

1. Directly execute query operations
2. Use the logged-in state to obtain more complete merchant information
3. Re-detect the login state if an exception occurs

Standardized Operation SOP Process

Basic Query Process

Step 1: Navigate to Dianping homepage (https://www.dianping.com)
Step 2: Use backbone to analyze page structure and detect login state (dual-state detection)
Step 3: 
   - Not logged in: Pause and guide the user to complete login manually
   - Logged in: Continue to execute the query
Step 4: Dynamically locate the search box and enter query keywords
   - Use the search tool to dynamically find multiple selectors: #myInput, #J-search-input, input.j-search-input
   - Prioritize using the placeholder attribute for fuzzy matching: input[placeholder*='搜索']
   - Avoid hard-coded ref references to improve cross-page stability
Step 5: Submit the search and wait for the result page to load
Step 6: Automatically handle tab switching (if new tabs are generated)
Step 7: Use readability to extract structured information (including complete sub-item ratings and user tags)
Step 8: Return the organized merchant information

City Switching Process

Step 1: Navigate to the current city page
Step 2: Click the city selection icon (img.city-select-icon)
Step 3: Search for the target city on the city list page
Step 4: Click the target city link to complete the switch
Step 5: Verify successful city switching (check URL and current city display)
Step 6: Execute the query operation for the target city

Deep Information Query Process

Step 1: Complete the basic query and enter the merchant details page
Step 2: Use readability to extract complete page content (automatically includes ingredient ratings and user tag counts)
Step 3: Locate the recommended dish area (#shop-dish) to obtain the dish list
Step 4: Extract user review information (ratings, number of reviews, review tags and mention counts, detailed review content)
Step 5: Organize and return complete deep information

Multi-Tab Management Process

Step 1: Perform operations that may generate new tabs (such as search submission)
Step 2: Use list_tabs to get all open tabs
Step 3: Identify the newly opened tab (usually the search result page)
Step 4: Automatically focus on the latest tab
Step 5: Verify the tab URL and title to ensure correctness
Step 6: Continue subsequent operations

Browser Tool Action Support List

Navigation and Page Analysis

• navigate

  : Access the specified URL (homepage or specific merchant page)

• backbone

  : Analyze the overall page structure and identify key interactive elements

• search

  : Prioritized for dynamically locating specific elements (search boxes, merchant cards, rating elements, etc.)

• readability

  : Extract structured text content from merchant details pages (Recommended first choice)

• list_tabs

  : Get information of all open tabs (used for multi-tab management)

User Interaction

• type

  : Enter query keywords in the search box

• click

  : Click the search button, merchant link or city selection icon

• press

  : Simulate keyboard operations (such as Enter to submit)

• focus_tab

  : Switch to the specified tab (used for multi-tab scenarios)

Content Extraction

• readability

  : Preferred solution to extract complete structured content from merchant details pages (including ingredient ratings, user tag counts, etc.)

• snapshot

  : Obtain the complete DOM for complex content analysis (Alternative solution)

Specific Operation Examples

Example 1: Query Chengdu Xiaolongkan Hot Pot (Logged-in State)

// Step 1: Navigate to homepage
use_browser(action="navigate", url="https://www.dianping.com")

// Step 2: Detect login state
use_browser(action="backbone")
// Check if the username element exists (logged-in state)

// Step 3: Dynamically locate the search box
use_browser(action="search", query="#myInput")

// Step 4: Execute search (logged in, execute directly)
use_browser(action="type", ref="found-ref", text="成都小龙坎老火锅", submit=true)

// Step 5: Extract complete results (including ingredient rating 4.7, user tag counts, etc.)
use_browser(action="readability")

Example 2: Handle Not Logged-in State

// Step 1: Navigate to homepage
use_browser(action="navigate", url="https://www.dianping.com")

// Step 2: Detect not logged-in state
use_browser(action="backbone")
// Found the text "Please Login/Register"

// Step 3: Guide user to login manually
// [Pause operation, prompt user to scan QR code to login]

// Step 4: Continue after user confirms login
use_browser(action="search", query="#myInput")
use_browser(action="type", ref="found-ref", text="北京中关村 咖啡店", submit=true)

Example 3: City Switching Operation

// Step 1: Navigate to current city
use_browser(action="navigate", url="https://www.dianping.com")

// Step 2: Click city selection icon
use_browser(action="search", query=".city-select-icon")
use_browser(action="click", ref="found-ref")

// Step 3: Search for target city
use_browser(action="search", query="成都")

// Step 4: Click Chengdu city link
use_browser(action="click", ref="found-beijing-ref")

// Step 5: Execute query after verifying successful switch
use_browser(action="search", query="#myInput")
use_browser(action="type", ref="found-ref", text="小龙坎火锅", submit=true)

Example 4: Deep Information Query (Including Ingredient Ratings and User Tags)

// Step 1: After entering the merchant details page
use_browser(action="readability") // Extract complete page content

// Results will include:
// - Overall Rating: 4.4 points
// - Sub-item Ratings: Taste 4.4, Environment 4.4, Service 4.4, Ingredients 4.7
// - User Tags: ["Queue situation(2269 times)", "Cashier attitude good(647 times)", "Environment clean and tidy(646 times)", ...]
// - Recommended Dishes: ["Mandarin Duck Pot", "Rose Meatball", "Milk Duck Blood", ...]
// - Business Hours: 11:00-02:00 next day

Example 5: Dynamic Search Box Positioning (Enhanced Stability)

// Use search tool to dynamically locate search box, avoid hard coding
use_browser(action="search", query="#myInput")
// Or on search result page
use_browser(action="search", query="#J-search-input")

// Perform operations based on returned ref
use_browser(action="type", ref="found-ref", text="Query Keyword", submit=true)

Example 6: Cross-Page Search Box Dynamic Positioning (Latest Optimization)

// Unified search box positioning strategy, compatible with different pages
use_browser(action="search", query="input[placeholder*='搜索'],#myInput,#J-search-input,input.j-search-input")
// Select the first found search box
use_browser(action="type", ref="first-found-ref", text="西安回民街", submit=true)

Example 7: Automatic Tab Management (Latest Optimization)

// Automatically handle tabs after executing search
use_browser(action="type", ref="search-input-ref", text="回民街", submit=true)

// Get all tabs
use_browser(action="list_tabs")

// Automatically switch to the latest tab (search result page)
use_browser(action="focus_tab", targetId="latest-search-tab-id")

// Continue subsequent operations
use_browser(action="readability")

Example 8: Business District Query Support (Latest Optimization)

// Query popular business districts (such as Muslim Quarter, Nanjing Road, etc.)
use_browser(action="search", query="input[placeholder*='搜索']")
use_browser(action="type", ref="found-ref", text="西安回民街", submit=true)

// Extract overall business district information
use_browser(action="readability")
// Results include business district rating, total number of reviews, popular tags, geographic location, etc.

Error Handling and Best Practices

Common Issue Handling

• Abnormal Login State: Re-detect login state, re-guide if necessary
• Network Timeout: Retry 1-2 times, report error if still failed
• Element Not Found: Use search for dynamic positioning, avoid hard-coded refs (key optimization)
• Incomplete Content Loading: Add appropriate waiting time or use wait_for
• Empty Search Results: Provide friendly empty result prompts
• City Switching Failed: Verify city list page structure, re-locate city link
• Verification Code/Slider Verification: Identify security mechanisms, guide user to handle manually
• Abnormal Tab Management: Use list_tabs to re-obtain tab list, ensure operation correctness

Performance Optimization Suggestions

• Prioritize

  readability

  over

  snapshot

  to improve efficiency and information completeness
• Use

  search

  tool for dynamic element positioning (core optimization point)
• Implement automatic multi-tab management to avoid manual switching errors
• Avoid unnecessary page refreshes and repeated navigation
• Set reasonable timeout (30-60 seconds recommended)
• Utilize login state persistence to reduce repeated logins
• Verify URL changes after city switching to ensure success
• Extract complete sub-item ratings, including the newly added ingredient rating dimension
• Extract user tags and mention counts to provide richer user feedback analysis
• Support business district queries to expand application scenario coverage

Reference Resources

Browser Tool Documentation

• references/browser-actions.md

  : Detailed browser tool Action usage guide (updated with dynamic positioning strategy)

• references/dianping-selectors.md

  : Common CSS selector reference for Dianping (updated with ingredient ratings and user tags)

• references/tab-management.md

  : Best practice guide for multi-tab processing (newly added)

• references/city-switching.md

  : City switching operation guide

• references/deep-info-extraction.md

  : Deep information extraction methods (updated with business district query support)

Operation Process Templates

• references/query-workflow.md

  : Standard query workflow template

• references/login-handling.md

  : Best practices for dual-state login handling

Usage Scenario Trigger Conditions

• "Help me check [location][type] stores on Dianping"
• "Search [keyword] on Dianping"
• "Get detailed information of a store on Dianping"
• "Query merchant ratings and addresses on Dianping"
• "Check the number of reviews and average consumption of a store on Dianping"
• "Switch to [city] to query [keyword]"
• "Get recommended dishes and user reviews of a store"
• "Query ingredient ratings and user tags of a merchant"
• "Query overall information of popular business districts (such as Muslim Quarter, Nanjing Road)"
• Any request involving extracting structured merchant information from the Dianping website

Verification Status

• ✅ Query of coffee shops in Zhongguancun, Beijing
• ✅ Query of restaurants in the Bund, Shanghai
• ✅ Query of Cantonese restaurants in Tianhe, Guangzhou
• ✅ Query of Xiaolongkan Hot Pot in Chengdu (latest verification, confirmed extraction of ingredient rating 4.7 and user tag counts)
• ✅ Query of Laoshan Noodle House in Hangzhou
• ✅ Query of Muslim Quarter business district in Xi'an (latest verification, confirmed cross-region city switching, business district information extraction, multi-tab management functions)