Alibaba Cloud Model Studio:Tongyi Xiaomi Conversation Analysis

Core Features

• Conversation Information Extraction: Automatically extracts key information from unstructured conversations, such as time, location, product, issue description, and customer requests.

  Typical use cases: Auto-fill for ticketing systems, CRM customer profile updates, issue categorization and archiving.

• Scenario Classification: Identifies the business scenario of a conversation and supports custom scenario categories.

  Typical use cases: Automatic routing of conversations to specialized teams, tracking consultation volume by scenario, targeted service process optimization.

• Satisfaction Mining: Determines customer satisfaction levels based on custom rules and supports multi-dimensional evaluation.

  Typical use cases: Agent performance evaluation, service quality monitoring, customer experience improvement.

• Complex Quality Inspection: Supports complex inspection rules involving nested conditions and logical reasoning, accurately enforcing organization-specific quality standards.

  Typical use cases: Quality inspection in highly regulated industries such as finance and healthcare, multi-step business process validation, complex script compliance checks.

• Multi-Turn Conversation Understanding: Automatically links different parts of a dialogue to accurately interpret references and contextual dependencies, capturing the user's true intent across extended interactions.

  Typical use cases: Conversation analysis for complex workflows, intent recognition in long dialogues, context-aware information extraction.

• Deep Semantic Understanding: Accurately identifies implicit meaning, emotional tone, and implied intent in conversations.

  Typical use cases: Customer sentiment monitoring, early warning for potential complaints, sales opportunity identification.

You can implement these features using custom prompts. For more information, see Feature Examples.

Applicability

• Supported regions: Only the Beijing region is supported. Use an API key from this region.

• Supported languages: The conversation analysis model supports only Chinese conversations.

• Supported models:

  • tongyi-xiaomi-analysis-flash: Designed for structured analysis tasks such as information extraction and scenario classification. Suitable for online analysis with low response latency requirements and cost-sensitive batch analytics.

  • tongyi-xiaomi-analysis-pro: Built for complex logical reasoning and deep semantic understanding. Ideal for scenarios involving nested quality inspection rules and context-dependent multi-turn conversation analysis. Compared to the Flash model, it offers stronger reasoning and comprehension capabilities but incurs higher cost and latency.

  For more information, see Model List.

Model Selection Guide

Getting Started

from openai import OpenAI
import os

dialogue = """
[1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today?
[2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running.
[3] Agent: I’m sorry for the inconvenience. Does the noise occur immediately after startup or after it’s been running for a while?
[4] Customer: It starts a few seconds after turning on, and the sound is very sharp.
[5] Agent: Understood. Could you please record a short video of it running and send it to our technical team for verification?
[6] Customer: Sure, I’ll send it shortly.
[7] Agent: Great. We’ll reply with a resolution within two hours of receiving the video.
[8] Customer: If it’s confirmed as a quality issue, can I get a direct replacement?
[9] Agent: Yes, if it’s verified as a quality issue, we’ll replace it free of charge and cover shipping.
[10] Customer: Okay, I’ll wait for your feedback. Thank you for your patient assistance.
"""

analysis_prompt = f"""
## Customer Service Quality Inspection: Satisfaction Assessment Rules

### Objective
Based on the agent-customer conversation, accurately analyze the customer’s emotion and feedback, determine their **satisfaction level** (Satisfied / Dissatisfied / Neutral), and provide reasoning for quality inspection and service improvement.

### Assessment Workflow (Match rules top-down; stop immediately when a condition is met)

1. **No Customer Input**
   - Condition: No customer utterances in the entire conversation.
   - Result: **Neutral**

2. **Complaint or Strong Dissatisfaction**
   - Condition: Customer directly or indirectly mentions filing a complaint (against riders, couriers, merchants, agents, etc.), or shows clear anxiety, questioning, or complaining while following up on issue progress.
   - Examples:
     - "I want to complain about your delivery person"
     - "Why hasn’t this been handled yet? When will it be fixed?"
   - Result: **Dissatisfied** (terminate assessment immediately)

3. **Negative Emotion in Final Utterance**
   - Condition: Customer’s last statement expresses negative emotion.
   - Result: **Dissatisfied**

4. **No Expression of Gratitude**
   - Condition: Customer never says “thank you” or similar explicit thanks.
   - Result: **Neutral**

5. **Explicit Satisfaction or Praise**
   - Condition: Customer clearly states satisfaction with the agent or gives positive feedback.
   - Result: **Satisfied** (overrides Rule 4)

6. **Gratitude After Solution Offered**
   - Condition: Before closing, after the agent offers compensation, refund, or solution, the customer explicitly and positively expresses thanks.
   - Notes:
     - “Okay” or “Alright” does not count as thanks
     - No reply does not count as acknowledgment
   - Result: **Satisfied**

7. **All Other Cases**
   - Condition: None of the above apply, and no complaints or clear dissatisfaction exist.
   - Result: **Neutral**

### Conversation Content
```text
{dialogue}
```

### Output Format
Include two parts:
1. **Satisfaction Label**: Satisfied / Dissatisfied / Neutral
2. **Reasoning**: Brief explanation based on the rules above.

Format:
```
Satisfaction Label#Reasoning
```
Examples:
```
Neutral#The customer expressed no thanks and showed no complaint or dissatisfaction, so the result is neutral.
Dissatisfied#The customer showed clear frustration and questioning while following up, matching Rule 2.
```
"""

# Get API Key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# If DASHSCOPE_API_KEY environment variable is not set, replace the next line with: api_key="sk-xxx"
client = OpenAI(
    api_key=os.getenv("DASHSCOPE_API_KEY"),
    base_url="https://dashscope.aliyuncs.com/compatible-mode/v1",
)

completion = client.chat.completions.create(
    model="tongyi-xiaomi-analysis-flash",
    messages=[
        {
            'role': 'user',
            'content': analysis_prompt
        }
    ],
    temperature=0,
    extra_body={
        "top_k": 1
    }
)

print(completion.choices[0].message.content)

import com.openai.client.OpenAIClient;
import com.openai.client.okhttp.OpenAIOkHttpClient;
import com.openai.core.JsonValue;
import com.openai.models.chat.completions.ChatCompletion;
import com.openai.models.chat.completions.ChatCompletionCreateParams;

public class Main {
    public static void main(String[] args) {
        try {
            OpenAIClient client = OpenAIOkHttpClient.builder()
                    // Get API Key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
                    // If DASHSCOPE_API_KEY environment variable is not set, replace the next line with .apiKey("sk-xxx")
                    .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                    .baseUrl("https://dashscope.aliyuncs.com/compatible-mode/v1")
                    .build();
            // Create ChatCompletion parameters
            ChatCompletionCreateParams params = ChatCompletionCreateParams.builder()
                    .model("tongyi-xiaomi-analysis-flash")
                    .addUserMessage(getUserMessage())
                    .temperature(0.0)
                    .putAdditionalBodyProperty("top_k", JsonValue.from(1))
                    .build();
            // Send request and get response
            ChatCompletion chatCompletion = client.chat().completions().create(params);
            String content = chatCompletion.choices().get(0).message().content().orElse("No valid content returned");
            System.out.println(content);
            // To view full response, uncomment the next line
            // System.out.println(chatCompletion);

        } catch (Exception e) {
            System.err.println("Error message: " + e.getMessage());
        }
    }

    private static String getUserMessage() {
        String dialogue ="[1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today?\n" +
                "                [2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running.\n" +
                "                [3] Agent: I’m sorry for the inconvenience. Does the noise occur immediately after startup or after it’s been running for a while?\n" +
                "                [4] Customer: It starts a few seconds after turning on, and the sound is very sharp.\n" +
                "                [5] Agent: Understood. Could you please record a short video of it running and send it to our technical team for verification?\n" +
                "                [6] Customer: Sure, I’ll send it shortly.\n" +
                "                [7] Agent: Great. We’ll reply with a resolution within two hours of receiving the video.\n" +
                "                [8] Customer: If it’s confirmed as a quality issue, can I get a direct replacement?\n" +
                "                [9] Agent: Yes, if it’s verified as a quality issue, we’ll replace it free of charge and cover shipping.\n" +
                "                [10] Customer: Okay, I’ll wait for your feedback. Thank you for your patient assistance.";

        String analysisPrompt = "## Customer Service Quality Inspection: Satisfaction Assessment Rules\n" +
                "\n" +
                "### Objective\n" +
                "Based on the agent-customer conversation, accurately analyze the customer’s emotion and feedback, determine their **satisfaction level** (Satisfied / Dissatisfied / Neutral), and provide reasoning for quality inspection and service improvement.\n" +
                "\n" +
                "### Assessment Workflow (Match rules top-down; stop immediately when a condition is met)\n" +
                "\n" +
                "1. **No Customer Input**\n" +
                "   - Condition: No customer utterances in the entire conversation.\n" +
                "   - Result: **Neutral**\n" +
                "\n" +
                "2. **Complaint or Strong Dissatisfaction**\n" +
                "   - Condition: Customer directly or indirectly mentions filing a complaint (against riders, couriers, merchants, agents, etc.), or shows clear anxiety, questioning, or complaining while following up on issue progress.\n" +
                "   - Examples:\n" +
                "     - \"I want to complain about your delivery person\"\n" +
                "     - \"Why hasn’t this been handled yet? When will it be fixed?\"\n" +
                "   - Result: **Dissatisfied** (terminate assessment immediately)\n" +
                "\n" +
                "3. **Negative Emotion in Final Utterance**\n" +
                "   - Condition: Customer’s last statement expresses negative emotion.\n" +
                "   - Result: **Dissatisfied**\n" +
                "\n" +
                "4. **No Expression of Gratitude**\n" +
                "   - Condition: Customer never says \"thank you\" or similar explicit thanks.\n" +
                "   - Result: **Neutral**\n" +
                "\n" +
                "5. **Explicit Satisfaction or Praise**\n" +
                "   - Condition: Customer clearly states satisfaction with the agent or gives positive feedback.\n" +
                "   - Result: **Satisfied** (overrides Rule 4)\n" +
                "\n" +
                "6. **Gratitude After Solution Offered**\n" +
                "   - Condition: Before closing, after the agent offers compensation, refund, or solution, the customer explicitly and positively expresses thanks.\n" +
                "   - Notes:\n" +
                "     - \"Okay\" or \"Alright\" does not count as thanks\n" +
                "     - No reply does not count as acknowledgment\n" +
                "   - Result: **Satisfied**\n" +
                "\n" +
                "7. **All Other Cases**\n" +
                "   - Condition: None of the above apply, and no complaints or clear dissatisfaction exist.\n" +
                "   - Result: **Neutral**\n" +
                "\n" +
                "### Conversation Content\n" +
                "```text\n" +
                dialogue + "\n" +
                "```\n" +
                "\n" +
                "### Output Format\n" +
                "Include two parts:\n" +
                "1. **Satisfaction Label**: Satisfied / Dissatisfied / Neutral\n" +
                "2. **Reasoning**: Brief explanation based on the rules above.\n" +
                "\n" +
                "Format:\n" +
                "```\n" +
                "Satisfaction Label#Reasoning\n" +
                "```\n" +
                "Examples:\n" +
                "```\n" +
                "Neutral#The customer expressed no thanks and showed no complaint or dissatisfaction, so the result is neutral.\n" +
                "Dissatisfied#The customer showed clear frustration and questioning while following up, matching Rule 2.\n" +
                "```";

        return analysisPrompt;
    }
}

// Requires Node.js v18+ and must run in an ES Module environment
import OpenAI from "openai";
const openai = new OpenAI(
    {
        // Get API Key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
        // If DASHSCOPE_API_KEY environment variable is not set, replace the next line with: apiKey: "sk-xxx",
        apiKey: process.env.DASHSCOPE_API_KEY,
        baseURL: "https://dashscope.aliyuncs.com/compatible-mode/v1"
    }
);
const completion = await openai.chat.completions.create({
    model: "tongyi-xiaomi-analysis-flash",
    messages: [
        { 
            role: "user", 
            content: "Please extract key information from the following conversation: [1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today? [2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running." 
        }
    ],
    temperature: 0,
    top_k: 1
});
console.log(completion.choices[0].message.content);
// To view full response, uncomment the next line
// console.log(JSON.stringify(completion, null, 4));

// OpenAI Go SDK version must be v2.4.0 or higher
package main

import (
	"context"
	// To view full response, uncomment the lines below and at the end
	// "encoding/json"
	"fmt"
	"os"

	"github.com/openai/openai-go/v2"
	"github.com/openai/openai-go/v2/option"
)

func main() {
	// Get API Key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
	// If DASHSCOPE_API_KEY environment variable is not set, replace the next line with: apiKey := "sk-xxx"
	apiKey := os.Getenv("DASHSCOPE_API_KEY")
	client := openai.NewClient(
		option.WithAPIKey(apiKey),
		option.WithBaseURL("https://dashscope.aliyuncs.com/compatible-mode/v1"),
	)

	chatCompletion, err := client.Chat.Completions.New(
		context.TODO(), openai.ChatCompletionNewParams{
			Messages: []openai.ChatCompletionMessageParamUnion{
				openai.UserMessage("Please extract key information from the following conversation: [1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today? [2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running."),
			},
			Model:       "tongyi-xiaomi-analysis-flash",
			Temperature: openai.Float(0.0),
		},
		option.WithJSONSet("top_k", 1),
	)
	if err != nil {
		fmt.Fprintf(os.Stderr, "Request failed: %v\n", err)
		os.Exit(1)
	}
	if len(chatCompletion.Choices) > 0 {
		fmt.Println(chatCompletion.Choices[0].Message.Content)
	}
	// To view full response, uncomment the next lines
	// jsonData, _ := json.MarshalIndent(chatCompletion, "", "  ")
	// fmt.Println(string(jsonData))
}

curl --location "https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions" \
-H "Authorization: Bearer $DASHSCOPE_API_KEY" \
-H "Content-Type: application/json" \
-d '{
    "model": "tongyi-xiaomi-analysis-flash",
    "messages": [
        {
            "role": "user", 
            "content": "Please extract key information from the following conversation: [1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today? [2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running."
        }
    ],
    "temperature": 0.0,
    "top_k": 1
}'

from http import HTTPStatus
import dashscope
import os

dashscope.base_http_api_url = "https://dashscope.aliyuncs.com/api/v1"

dialogue = """
[1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today?
[2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running.
[3] Agent: I’m sorry for the inconvenience. Does the noise occur immediately after startup or after it’s been running for a while?
[4] Customer: It starts a few seconds after turning on, and the sound is very sharp.
[5] Agent: Understood. Could you please record a short video of it running and send it to our technical team for verification?
[6] Customer: Sure, I’ll send it shortly.
[7] Agent: Great. We’ll reply with a resolution within two hours of receiving the video.
[8] Customer: If it’s confirmed as a quality issue, can I get a direct replacement?
[9] Agent: Yes, if it’s verified as a quality issue, we’ll replace it free of charge and cover shipping.
[10] Customer: Okay, I’ll wait for your feedback. Thank you for your patient assistance.
"""

analysis_prompt = f"""
## Customer Service Quality Inspection: Satisfaction Assessment Rules

### Objective
Based on the agent-customer conversation, accurately analyze the customer’s emotion and feedback, determine their **satisfaction level** (Satisfied / Dissatisfied / Neutral), and provide reasoning for quality inspection and service improvement.

### Assessment Workflow (Match rules top-down; stop immediately when a condition is met)

1. **No Customer Input**
   - Condition: No customer utterances in the entire conversation.
   - Result: **Neutral**

2. **Complaint or Strong Dissatisfaction**
   - Condition: Customer directly or indirectly mentions filing a complaint (against riders, couriers, merchants, agents, etc.), or shows clear anxiety, questioning, or complaining while following up on issue progress.
   - Examples:
     - "I want to complain about your delivery person"
     - "Why hasn’t this been handled yet? When will it be fixed?"
   - Result: **Dissatisfied** (terminate assessment immediately)

3. **Negative Emotion in Final Utterance**
   - Condition: Customer’s last statement expresses negative emotion.
   - Result: **Dissatisfied**

4. **No Expression of Gratitude**
   - Condition: Customer never says “thank you” or similar explicit thanks.
   - Result: **Neutral**

5. **Explicit Satisfaction or Praise**
   - Condition: Customer clearly states satisfaction with the agent or gives positive feedback.
   - Result: **Satisfied** (overrides Rule 4)

6. **Gratitude After Solution Offered**
   - Condition: Before closing, after the agent offers compensation, refund, or solution, the customer explicitly and positively expresses thanks.
   - Notes:
     - “Okay” or “Alright” does not count as thanks
     - No reply does not count as acknowledgment
   - Result: **Satisfied**

7. **All Other Cases**
   - Condition: None of the above apply, and no complaints or clear dissatisfaction exist.
   - Result: **Neutral**

### Conversation Content
```text
{dialogue}
```

### Output Format
Include two parts:
1. **Satisfaction Label**: Satisfied / Dissatisfied / Neutral
2. **Reasoning**: Brief explanation based on the rules above.

Format:
```
Satisfaction Label#Reasoning
```
Examples:
```
Neutral#The customer expressed no thanks and showed no complaint or dissatisfaction, so the result is neutral.
Dissatisfied#The customer showed clear frustration and questioning while following up, matching Rule 2.
```
"""

# Get API Key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
# If DASHSCOPE_API_KEY environment variable is not set, replace the next line with: api_key="sk-xxx"
response = dashscope.Generation.call(
    api_key=os.getenv('DASHSCOPE_API_KEY'),
    model="tongyi-xiaomi-analysis-flash",
    messages=[
        {
            'role': 'user',
            'content': analysis_prompt
        }
    ],
    temperature=0.0,
    top_k=1,
    result_format="message"
)

if response.status_code == HTTPStatus.OK:
    print(response.output.choices[0].message.content)
else:
    print('Request id: %s, Status code: %s, error code: %s, error message: %s' % (
        response.request_id, response.status_code,
        response.code, response.message
    ))

import java.util.Arrays;
import java.lang.System;
import com.alibaba.dashscope.aigc.generation.Generation;
import com.alibaba.dashscope.aigc.generation.GenerationParam;
import com.alibaba.dashscope.aigc.generation.GenerationResult;
import com.alibaba.dashscope.common.Message;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.ApiException;
import com.alibaba.dashscope.exception.InputRequiredException;
import com.alibaba.dashscope.exception.NoApiKeyException;
import com.alibaba.dashscope.protocol.Protocol;

public class Main {
    public static GenerationResult callWithMessage() throws ApiException, NoApiKeyException, InputRequiredException {
        Generation gen = new Generation(Protocol.HTTP.getValue(), "https://dashscope.aliyuncs.com/api/v1");

        Message userMsg = Message.builder()
                .role(Role.USER.getValue())
                .content(getContent())
                .build();

        GenerationParam param = GenerationParam.builder()
                // Get API Key: https://www.alibabacloud.com/help/zh/model-studio/get-api-key
                // If DASHSCOPE_API_KEY environment variable is not set, replace the next line with: .apiKey("sk-xxx")
                .apiKey(System.getenv("DASHSCOPE_API_KEY"))
                .model("tongyi-xiaomi-analysis-flash")
                .messages(Arrays.asList(userMsg))
                .temperature(0.0f)
                .topK(1)
                .resultFormat(GenerationParam.ResultFormat.MESSAGE)
                .build();
        return gen.call(param);
    }

    public static void main(String[] args) {
        try {
            GenerationResult result = callWithMessage();
            System.out.println(result.getOutput().getChoices().get(0).getMessage().getContent());
            // To view full response, uncomment the next line
            // System.out.println(JsonUtils.toJson(result));
        } catch (ApiException | NoApiKeyException | InputRequiredException e) {
            System.err.println("Error message: "+e.getMessage());
        }
    }

    private static String getContent() {
        String dialogue ="[1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today?\n" +
                "                [2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running.\n" +
                "                [3] Agent: I’m sorry for the inconvenience. Does the noise occur immediately after startup or after it’s been running for a while?\n" +
                "                [4] Customer: It starts a few seconds after turning on, and the sound is very sharp.\n" +
                "                [5] Agent: Understood. Could you please record a short video of it running and send it to our technical team for verification?\n" +
                "                [6] Customer: Sure, I’ll send it shortly.\n" +
                "                [7] Agent: Great. We’ll reply with a resolution within two hours of receiving the video.\n" +
                "                [8] Customer: If it’s confirmed as a quality issue, can I get a direct replacement?\n" +
                "                [9] Agent: Yes, if it’s verified as a quality issue, we’ll replace it free of charge and cover shipping.\n" +
                "                [10] Customer: Okay, I’ll wait for your feedback. Thank you for your patient assistance.";

        String analysisPrompt = "## Customer Service Quality Inspection: Satisfaction Assessment Rules\n" +
                "\n" +
                "### Objective\n" +
                "Based on the agent-customer conversation, accurately analyze the customer’s emotion and feedback, determine their **satisfaction level** (Satisfied / Dissatisfied / Neutral), and provide reasoning for quality inspection and service improvement.\n" +
                "\n" +
                "### Assessment Workflow (Match rules top-down; stop immediately when a condition is met)\n" +
                "\n" +
                "1. **No Customer Input**\n" +
                "   - Condition: No customer utterances in the entire conversation.\n" +
                "   - Result: **Neutral**\n" +
                "\n" +
                "2. **Complaint or Strong Dissatisfaction**\n" +
                "   - Condition: Customer directly or indirectly mentions filing a complaint (against riders, couriers, merchants, agents, etc.), or shows clear anxiety, questioning, or complaining while following up on issue progress.\n" +
                "   - Examples:\n" +
                "     - \"I want to complain about your delivery person\"\n" +
                "     - \"Why hasn’t this been handled yet? When will it be fixed?\"\n" +
                "   - Result: **Dissatisfied** (terminate assessment immediately)\n" +
                "\n" +
                "3. **Negative Emotion in Final Utterance**\n" +
                "   - Condition: Customer’s last statement expresses negative emotion.\n" +
                "   - Result: **Dissatisfied**\n" +
                "\n" +
                "4. **No Expression of Gratitude**\n" +
                "   - Condition: Customer never says \"thank you\" or similar explicit thanks.\n" +
                "   - Result: **Neutral**\n" +
                "\n" +
                "5. **Explicit Satisfaction or Praise**\n" +
                "   - Condition: Customer clearly states satisfaction with the agent or gives positive feedback.\n" +
                "   - Result: **Satisfied** (overrides Rule 4)\n" +
                "\n" +
                "6. **Gratitude After Solution Offered**\n" +
                "   - Condition: Before closing, after the agent offers compensation, refund, or solution, the customer explicitly and positively expresses thanks.\n" +
                "   - Notes:\n" +
                "     - \"Okay\" or \"Alright\" does not count as thanks\n" +
                "     - No reply does not count as acknowledgment\n" +
                "   - Result: **Satisfied**\n" +
                "\n" +
                "7. **All Other Cases**\n" +
                "   - Condition: None of the above apply, and no complaints or clear dissatisfaction exist.\n" +
                "   - Result: **Neutral**\n" +
                "\n" +
                "### Conversation Content\n" +
                "```text\n" +
                dialogue + "\n" +
                "```\n" +
                "\n" +
                "### Output Format\n" +
                "Include two parts:\n" +
                "1. **Satisfaction Label**: Satisfied / Dissatisfied / Neutral\n" +
                "2. **Reasoning**: Brief explanation based on the rules above.\n" +
                "\n" +
                "Format:\n" +
                "```\n" +
                "Satisfaction Label#Reasoning\n" +
                "```\n" +
                "Examples:\n" +
                "```\n" +
                "Neutral#The customer expressed no thanks and showed no complaint or dissatisfaction, so the result is neutral.\n" +
                "Dissatisfied#The customer showed clear frustration and questioning while following up, matching Rule 2.\n" +
                "```";

        return analysisPrompt;
    }
}

curl --location "https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation" \
--header "Authorization: Bearer $DASHSCOPE_API_KEY" \
--header "Content-Type: application/json" \
--data '{
    "model": "tongyi-xiaomi-analysis-flash",
    "input":{
        "messages":[
            {
                "role": "user",
                "content": "Please extract key information from the following conversation: [1] Agent: Hello, welcome to AB E-commerce platform. How can I help you today? [2] Customer: Hi, I bought a food processor from your store last week, and it’s making a strange noise when running."
            }
        ],
        "temperature": 0.0,
        "top_k": 1
    },
    "parameters": {
        "result_format": "message"
    }
}'

Feature Examples

Each feature example below provides a ready-to-run content block. To test a feature, replace the content in the code examples with the corresponding content block from this section.

Extract key information from the following conversation and output in JSON format.

Conversation:
```
[1] Agent: Hello, I’m Xiao Wang. How can I help you?
[2] Customer: Hi, I bought an iPhone 15 last Wednesday at your Xinjiekou store in Nanjing. My order number is 123456789, and now the screen looks yellowish.
[3] Agent: Got it. May I have your mobile phone number?
[4] Customer: 138****5678
[5] Agent: Received. I’ve found your order. Regarding the yellowish screen, when did you first notice it?
[6] Customer: During normal use, especially when viewing white backgrounds—it’s very obvious.
```

Information to extract:
1. Purchase time
2. Store location
3. Product name
4. Order number
5. Customer contact information
6. Issue description
7. Context where issue was noticed

Output format (JSON):
```json
{{
  "Purchase Time": "...",
  "Store Location": "...",
  "Product Name": "...",
  "Order Number": "...",
  "Customer Contact Information": "...",
  "Issue Description": "...",
  "Context Where Issue Was Noticed": "..."
}}
```

Note: If any field is not mentioned in the conversation, fill in "Not mentioned".

Determine which business scenario the following conversation belongs to.

Conversation:
```
[1] Agent: Hello, welcome to our customer service hotline. How can I help you?
[2] Customer: I’d like to know what promotions you have for members.
[3] Agent: Currently, new members get 20% off their first month, and existing members earn points for referring friends.
[4] Customer: How do I sign up for membership?
[5] Agent: You can register through our app or mini program by filling in basic information.
[6] Customer: Got it, thanks.
```

Available scenario categories:
1. Pre-sales Inquiry – Customer asks about product features, pricing, or purchase methods
2. Post-sales Complaint – Customer reports product issues, service dissatisfaction, or requests returns/exchanges
3. Logistics Inquiry – Customer asks about order status, delivery progress, or shipping details
4. Membership Services – Customer inquires about membership benefits, points, or promotions
5. Technical Support – Customer seeks help with product usage or troubleshooting
6. Account Issues – Customer reports login, password, or account security problems

Output format:
```
Scenario Category: [Select one from the list above]
Confidence Level: [High/Medium/Low]
Reasoning: [Brief justification in 1–2 sentences]
```

Assess customer satisfaction based on the following conversation.

Conversation:
```
[1] Agent: Hello, I’m customer service agent Xiao Li. How can I help you?
[2] Customer: Why hasn’t my order shipped yet? It’s been three days!
[3] Agent: I sincerely apologize for the inconvenience. Let me check right away. May I have your order number?
[4] Customer: 1234567890
[5] Agent: Found it. Due to recent logistics peak, shipping is delayed. I’ll escalate this to the warehouse for urgent processing—guaranteed to ship today.
[6] Customer: When will it arrive?
[7] Agent: Estimated delivery in 3–5 business days. As an apology, I’ve added a 20 CNY coupon to your account for your next purchase.
[8] Customer: Alright, just make sure it ships soon.
[9] Agent: Absolutely! I’ll monitor your order closely and notify you immediately once shipped. Anything else I can assist with?
[10] Customer: No.
```

Satisfaction criteria:
1. **Dissatisfied**: Customer explicitly complains, expresses dissatisfaction, or demands compensation/refund
2. **Neutral**: Issue resolved but no thanks or acknowledgment, or customer shows reservation about the solution
3. **Satisfied**: Customer explicitly thanks, acknowledges service, or accepts the solution positively

Evaluation dimensions:
- Customer emotion (initial vs. ending state)
- Issue resolution status (was a solution provided?)
- Customer feedback (thanks, acknowledgment, or continued dissatisfaction?)
- Compensation offered (any extra incentives or discounts?)

Output format:
```
Satisfaction Level: [Satisfied/Neutral/Dissatisfied]
Initial Emotion: [Describe customer’s initial emotional state]
Ending Emotion: [Describe customer’s emotional state at conversation end]
Issue Resolved: [Yes/No/Partially]
Reasoning: [Detailed justification based on evaluation dimensions]
```

Perform quality inspection on the following financial industry customer service conversation based on the rules below.

Conversation:
```
[1] Agent: Hello, welcome to XX Bank Credit Card Center. I’m agent Xiao Zhang, ID 8888. How can I help you?
[2] Customer: I’d like to increase my credit card limit.
[3] Agent: Sure. May I have your name and ID number?
[4] Customer: My name is Alice, ID 110101199001011234.
[5] Agent: Thank you, Mr. Alice. Your current limit is 20,000 CNY. Based on your usage history and credit score, we can raise it to 50,000 CNY. Would you like to proceed?
[6] Customer: Yes, please.
[7] Agent: Confirming: your credit limit will be increased from 20,000 CNY to 50,000 CNY, effective immediately. Do you confirm?
[8] Customer: Confirmed.
[9] Agent: Submitted. Processing will complete within 2 business days. Thank you for calling. Have a great day!
```

Quality inspection rules (must be strictly enforced):

**1. Opening Script (Mandatory)**
- Rule 1: Must include “Hello” or “Welcome”
- Rule 2: Must state agent ID
- Rule 3: Must specify department (e.g., “XX Bank Credit Card Center”)
- **Verdict**: All three must be satisfied to pass

**2. Identity Verification (Mandatory)**
- Rule 4: For account-related actions, must verify customer identity (name + ID number OR name + card number)
- **Verdict**: Failure to verify during account action = critical violation

**3. Critical Information Disclosure (Mandatory)**
- Rule 5: For limit adjustments, rate changes, or product modifications, must:
  - Clearly state pre- and post-change details
  - Disclose effective time
  - Obtain explicit customer confirmation (“yes”, “confirmed”, “agree”, etc.)
- **Verdict**: Missing any item = critical violation

**4. Closing Script (Recommended)**
- Rule 6: Should include expression of thanks
- Rule 7: Should include well-wishes
- **Verdict**: Failure affects scoring but not overall pass/fail

**Output format**:
```
Inspection Result: [Pass/Fail]
Number of Violations: [Number]

Detailed Check:
1. Opening Script: [Pass/Fail] - [Explanation]
2. Identity Verification: [Pass/Fail] - [Explanation]
3. Critical Information Disclosure: [Pass/Fail] - [Explanation]
4. Closing Script: [Pass/Fail (Recommended)] - [Explanation]

Summary: [1–2 sentence evaluation]
```

Extract complete ticket information from the following conversation by integrating scattered details across multiple turns.

Conversation:
```
[1] Agent: Hello, how can I help you?
[2] Customer: The router I bought last week keeps disconnecting.
[3] Agent: Which model did you purchase?
[4] Customer: AX3000, the white one.
[5] Agent: Noted. How often does it disconnect?
[6] Customer: Every 2–3 hours, which really disrupts my work.
[7] Agent: When did this start?
[8] Customer: Just these past two days—it worked fine before.
[9] Agent: Have you moved the router or changed any settings recently?
[10] Customer: No, it’s still in the living room, and settings haven’t changed.
[11] Agent: This might be due to an outdated firmware version. I suggest upgrading the firmware first.
[12] Customer: How do I upgrade?
[13] Agent: Open our app on your phone, find this device, and tap “Firmware Upgrade.”
[14] Customer: Okay, I’ll try now.
[15] Agent: The upgrade takes about 5 minutes—network will disconnect during this time. Don’t power off the router.
[16] Customer: Will it reconnect automatically after the upgrade?
[17] Agent: Yes, it will restart automatically, then just reconnect to Wi-Fi.
[18] Customer: What if it still disconnects after the upgrade?
[19] Agent: If the issue persists, it may be a hardware fault. We’ll arrange a replacement or on-site repair.
[20] Customer: Got it. Thanks.
```

Extract the following information (note: details are scattered across turns—use context to link them):

1. **Product Details** (model, appearance, purchase time)
2. **Issue Description** (symptoms, frequency)
3. **Timeline** (when it started, prior status)
4. **Customer Actions** (any special operations or adjustments)
5. **Agent Diagnosis** (initial suspected cause)
6. **Proposed Solution** (recommended steps)
7. **Contingency Plan** (alternative if solution fails)

Output in structured format and indicate which turns each piece of information comes from.

Perform deep semantic analysis on the following conversation.

Conversation:
```
[1] Agent: Hello, how can I help you?
[2] Customer: Why did my membership expire suddenly? It was working last month.
[3] Agent: Let me check. May I have your phone number?
[4] Customer: 138****5678
[5] Agent: Your annual membership expired on October 15 last year.
[6] Customer: Oh, I see.
[7] Agent: Member-exclusive coupons become invalid after expiration. But you can renew your membership to restore them.
[8] Customer: Alright, I understand.
[9] Agent: Anything else I can assist with?
[10] Customer: No, thanks.
```

Analysis dimensions:

1. **Customer Emotion Analysis**
   - Turn 2: “Why did my membership expire suddenly? It was working last month.” — What is the emotion and implied meaning?
   - Turn 6: “Oh, I see.” — How has the emotion shifted?
   - Turn 8: “Alright, I understand.” — What is the true sentiment?

2. **Potential Issue Identification**
   - Is the customer dissatisfied?
   - Is there risk of a complaint?
   - Does the customer misunderstand service rules?

3. **True Customer Needs**
   - What matters most to the customer?
   - Was the issue resolved satisfactorily?
   - Are there unspoken needs?

4. **Improvement Suggestions**
   - Was the agent’s response appropriate?
   - How could the interaction be improved to boost satisfaction?

Output format: Provide clear analysis results for each of the four dimensions above.

API Reference

For information about model input and output parameters, see Qwen.