使用Spring AI框架结合Sentinel限流的简单案例

一只爱撸猫的程序猿
126 阅读6分钟

我将实现一个简单的使用Spring AI框架结合Sentinel限流的案例，场景是一个智能客服系统。

场景描述

构建一个AI智能客服系统，需要对AI接口调用进行限流保护，防止恶意请求或突发流量导致系统崩溃。

项目结构和依赖

1. Maven依赖配置

<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
         xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
         xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 
         http://maven.apache.org/xsd/maven-4.0.0.xsd">
    <modelVersion>4.0.0</modelVersion>
    
    <parent>
        <groupId>org.springframework.boot</groupId>
        <artifactId>spring-boot-starter-parent</artifactId>
        <version>3.2.0</version>
        <relativePath/>
    </parent>
    
    <groupId>com.example</groupId>
    <artifactId>ai-customer-service</artifactId>
    <version>1.0.0</version>
    
    <dependencies>
        <!-- Spring Boot Starter -->
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-web</artifactId>
        </dependency>
        
        <!-- Spring AI -->
        <dependency>
            <groupId>org.springframework.ai</groupId>
            <artifactId>spring-ai-openai-spring-boot-starter</artifactId>
            <version>0.8.1</version>
        </dependency>
        
        <!-- Sentinel核心库 -->
        <dependency>
            <groupId>com.alibaba.csp</groupId>
            <artifactId>sentinel-core</artifactId>
            <version>1.8.6</version>
        </dependency>
        
        <!-- Sentinel Spring Boot Starter -->
        <dependency>
            <groupId>com.alibaba.cloud</groupId>
            <artifactId>spring-cloud-starter-alibaba-sentinel</artifactId>
            <version>2022.0.0.0</version>
        </dependency>
        
        <!-- Sentinel Dashboard -->
        <dependency>
            <groupId>com.alibaba.csp</groupId>
            <artifactId>sentinel-transport-simple-http</artifactId>
            <version>1.8.6</version>
        </dependency>
        
        <!-- Redis for rate limiting -->
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-data-redis</artifactId>
        </dependency>
        
        <!-- JSON处理 -->
        <dependency>
            <groupId>com.fasterxml.jackson.core</groupId>
            <artifactId>jackson-databind</artifactId>
        </dependency>
    </dependencies>
</project>

2. 应用配置

# application.yml
server:
  port: 8080

spring:
  application:
    name: ai-customer-service
  
  # Spring AI OpenAI配置
  ai:
    openai:
      api-key: ${OPENAI_API_KEY:your-api-key-here}
      base-url: https://api.openai.com
      chat:
        options:
          model: gpt-3.5-turbo
          temperature: 0.7
          max-tokens: 1000
  
  # Redis配置
  redis:
    host: localhost
    port: 6379
    database: 0
    timeout: 2000ms
    jedis:
      pool:
        max-active: 8
        max-idle: 8
        min-idle: 0

# Sentinel配置
management:
  endpoints:
    web:
      exposure:
        include: "*"

logging:
  level:
    com.alibaba.csp.sentinel: DEBUG
    com.example: DEBUG

3. Sentinel配置类

package com.example.config;

import com.alibaba.csp.sentinel.annotation.aspectj.SentinelResourceAspect;
import com.alibaba.csp.sentinel.slots.block.RuleConstant;
import com.alibaba.csp.sentinel.slots.block.flow.FlowRule;
import com.alibaba.csp.sentinel.slots.block.flow.FlowRuleManager;
import com.alibaba.csp.sentinel.slots.block.flow.param.ParamFlowRule;
import com.alibaba.csp.sentinel.slots.block.flow.param.ParamFlowRuleManager;
import org.springframework.context.annotation.Bean;
import org.springframework.context.annotation.Configuration;

import javax.annotation.PostConstruct;
import java.util.ArrayList;
import java.util.List;

@Configuration
public class SentinelConfig {

    @Bean
    public SentinelResourceAspect sentinelResourceAspect() {
        return new SentinelResourceAspect();
    }

    @PostConstruct
    public void initFlowRules() {
        initBasicFlowRules();
        initParamFlowRules();
    }

    /**
     * 初始化基础限流规则
     */
    private void initBasicFlowRules() {
        List<FlowRule> rules = new ArrayList<>();

        // AI聊天接口限流规则 - 令牌桶算法
        FlowRule chatRule = new FlowRule();
        chatRule.setResource("ai-chat");
        chatRule.setGrade(RuleConstant.FLOW_GRADE_QPS);
        chatRule.setCount(10); // 每秒10个请求
        chatRule.setControlBehavior(RuleConstant.CONTROL_BEHAVIOR_RATE_LIMITER);
        rules.add(chatRule);

        // AI问答接口限流规则 - 预热算法
        FlowRule qaRule = new FlowRule();
        qaRule.setResource("ai-qa");
        qaRule.setGrade(RuleConstant.FLOW_GRADE_QPS);
        qaRule.setCount(20);
        qaRule.setControlBehavior(RuleConstant.CONTROL_BEHAVIOR_WARM_UP);
        qaRule.setWarmUpPeriodSec(30); // 30秒预热时间
        rules.add(qaRule);

        // 知识库查询接口 - 排队等待
        FlowRule knowledgeRule = new FlowRule();
        knowledgeRule.setResource("knowledge-search");
        knowledgeRule.setGrade(RuleConstant.FLOW_GRADE_QPS);
        knowledgeRule.setCount(15);
        knowledgeRule.setControlBehavior(RuleConstant.CONTROL_BEHAVIOR_RATE_LIMITER);
        knowledgeRule.setMaxQueueingTimeMs(500); // 最大等待500ms
        rules.add(knowledgeRule);

        FlowRuleManager.loadRules(rules);
    }

    /**
     * 初始化热点参数限流规则
     */
    private void initParamFlowRules() {
        List<ParamFlowRule> rules = new ArrayList<>();

        // 基于用户ID的限流
        ParamFlowRule userRule = new ParamFlowRule();
        userRule.setResource("ai-chat");
        userRule.setParamIdx(0); // 第一个参数是用户ID
        userRule.setGrade(RuleConstant.FLOW_GRADE_QPS);
        userRule.setCount(3); // 单个用户每秒最多3次请求
        userRule.setControlBehavior(RuleConstant.CONTROL_BEHAVIOR_RATE_LIMITER);
        rules.add(userRule);

        ParamFlowRuleManager.loadRules(rules);
    }
}

4. 自定义限流异常处理

package com.example.handler;

import com.alibaba.csp.sentinel.slots.block.BlockException;
import com.alibaba.csp.sentinel.slots.block.flow.FlowException;
import com.alibaba.csp.sentinel.slots.block.flow.param.ParamFlowException;
import lombok.extern.slf4j.Slf4j;
import org.springframework.stereotype.Component;

@Slf4j
@Component
public class SentinelBlockHandler {

    /**
     * AI聊天接口限流降级处理
     */
    public static String handleAiChatBlock(String userId, String message, BlockException ex) {
        log.warn("AI聊天接口被限流，用户ID: {}, 异常类型: {}", userId, ex.getClass().getSimpleName());
        
        if (ex instanceof FlowException) {
            return "系统繁忙，请稍后再试。我们正在为您排队处理...";
        } else if (ex instanceof ParamFlowException) {
            return "您的请求过于频繁，请稍后再试。";
        }
        
        return "系统暂时无法处理您的请求，请稍后重试。";
    }

    /**
     * AI问答接口限流降级处理
     */
    public static String handleAiQaBlock(String question, BlockException ex) {
        log.warn("AI问答接口被限流，问题: {}, 异常类型: {}", question, ex.getClass().getSimpleName());
        return "当前咨询人数较多，系统正在预热中，请稍后再试。";
    }

    /**
     * 知识库搜索限流降级处理
     */
    public static String handleKnowledgeSearchBlock(String keyword, BlockException ex) {
        log.warn("知识库搜索被限流，关键词: {}", keyword);
        return "知识库查询繁忙，请稍后再试或联系人工客服。";
    }
}

5. AI服务层

package com.example.service;

import com.alibaba.csp.sentinel.annotation.SentinelResource;
import com.alibaba.csp.sentinel.slots.block.BlockException;
import com.example.handler.SentinelBlockHandler;
import lombok.extern.slf4j.Slf4j;
import org.springframework.ai.chat.ChatClient;
import org.springframework.ai.chat.ChatResponse;
import org.springframework.ai.chat.messages.UserMessage;
import org.springframework.ai.chat.prompt.Prompt;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Service;

import java.time.LocalDateTime;
import java.util.concurrent.CompletableFuture;

@Service
@Slf4j
public class AiCustomerService {

    @Autowired
    private ChatClient chatClient;

    /**
     * AI聊天服务 - 使用热点参数限流
     */
    @SentinelResource(
        value = "ai-chat",
        blockHandler = "handleAiChatBlock",
        blockHandlerClass = SentinelBlockHandler.class
    )
    public String chatWithAi(String userId, String message) {
        log.info("用户 {} 发起聊天请求: {}", userId, message);
        
        try {
            // 构建聊天上下文
            String systemPrompt = "你是一个专业的客服助手，请礼貌、准确地回答用户问题。";
            String fullPrompt = systemPrompt + "\n用户问题: " + message;
            
            Prompt prompt = new Prompt(new UserMessage(fullPrompt));
            ChatResponse response = chatClient.call(prompt);
            
            String aiResponse = response.getResult().getOutput().getContent();
            log.info("AI回复用户 {}: {}", userId, aiResponse);
            
            return aiResponse;
        } catch (Exception e) {
            log.error("AI聊天服务异常，用户ID: {}", userId, e);
            return "抱歉，AI服务暂时不可用，请联系人工客服。";
        }
    }

    /**
     * AI问答服务 - 使用预热限流
     */
    @SentinelResource(
        value = "ai-qa",
        blockHandler = "handleAiQaBlock",
        blockHandlerClass = SentinelBlockHandler.class
    )
    public String answerQuestion(String question) {
        log.info("收到问答请求: {}", question);
        
        try {
            String prompt = String.format(
                "作为客服专家，请简洁准确地回答以下问题：%s\n" +
                "要求：1. 回答要专业且易懂 2. 控制在200字以内 3. 如果不确定请说明", 
                question
            );
            
            ChatResponse response = chatClient.call(new Prompt(prompt));
            return response.getResult().getOutput().getContent();
        } catch (Exception e) {
            log.error("AI问答服务异常", e);
            return "抱歉，暂时无法回答您的问题，请稍后再试。";
        }
    }

    /**
     * 知识库搜索 - 使用排队等待限流
     */
    @SentinelResource(
        value = "knowledge-search",
        blockHandler = "handleKnowledgeSearchBlock",
        blockHandlerClass = SentinelBlockHandler.class
    )
    public String searchKnowledge(String keyword) {
        log.info("知识库搜索: {}", keyword);
        
        try {
            // 模拟知识库搜索
            Thread.sleep(100); // 模拟搜索耗时
            
            String searchPrompt = String.format(
                "基于关键词 '%s' 搜索相关知识，提供简洁的信息摘要。", 
                keyword
            );
            
            ChatResponse response = chatClient.call(new Prompt(searchPrompt));
            return response.getResult().getOutput().getContent();
        } catch (InterruptedException e) {
            Thread.currentThread().interrupt();
            return "搜索被中断，请重试。";
        } catch (Exception e) {
            log.error("知识库搜索异常", e);
            return "知识库搜索失败，请联系技术支持。";
        }
    }

    /**
     * 异步AI处理 - 用于处理复杂请求
     */
    public CompletableFuture<String> processComplexRequest(String userId, String request) {
        return CompletableFuture.supplyAsync(() -> {
            try {
                return chatWithAi(userId, request);
            } catch (BlockException e) {
                return "请求过于频繁，已加入处理队列，请稍后查看结果。";
            }
        });
    }
}

6. 控制器层

package com.example.controller;

import com.example.service.AiCustomerService;
import lombok.extern.slf4j.Slf4j;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.http.ResponseEntity;
import org.springframework.web.bind.annotation.*;

import java.util.HashMap;
import java.util.Map;
import java.util.concurrent.CompletableFuture;

@RestController
@RequestMapping("/api/ai")
@Slf4j
public class AiCustomerController {

    @Autowired
    private AiCustomerService aiCustomerService;

    /**
     * AI聊天接口
     */
    @PostMapping("/chat")
    public ResponseEntity<Map<String, Object>> chat(
            @RequestParam String userId,
            @RequestBody Map<String, String> request) {
        
        String message = request.get("message");
        log.info("接收到聊天请求 - 用户: {}, 消息: {}", userId, message);
        
        Map<String, Object> response = new HashMap<>();
        
        try {
            String aiResponse = aiCustomerService.chatWithAi(userId, message);
            response.put("success", true);
            response.put("data", aiResponse);
            response.put("timestamp", System.currentTimeMillis());
            return ResponseEntity.ok(response);
        } catch (Exception e) {
            log.error("聊天接口异常", e);
            response.put("success", false);
            response.put("error", "服务暂时不可用");
            return ResponseEntity.internalServerError().body(response);
        }
    }

    /**
     * AI问答接口
     */
    @PostMapping("/qa")
    public ResponseEntity<Map<String, Object>> qa(@RequestBody Map<String, String> request) {
        String question = request.get("question");
        log.info("接收到问答请求: {}", question);
        
        Map<String, Object> response = new HashMap<>();
        
        try {
            String answer = aiCustomerService.answerQuestion(question);
            response.put("success", true);
            response.put("question", question);
            response.put("answer", answer);
            response.put("timestamp", System.currentTimeMillis());
            return ResponseEntity.ok(response);
        } catch (Exception e) {
            log.error("问答接口异常", e);
            response.put("success", false);
            response.put("error", "问答服务暂时不可用");
            return ResponseEntity.internalServerError().body(response);
        }
    }

    /**
     * 知识库搜索接口
     */
    @GetMapping("/knowledge/search")
    public ResponseEntity<Map<String, Object>> searchKnowledge(@RequestParam String keyword) {
        log.info("接收到知识库搜索请求: {}", keyword);
        
        Map<String, Object> response = new HashMap<>();
        
        try {
            String result = aiCustomerService.searchKnowledge(keyword);
            response.put("success", true);
            response.put("keyword", keyword);
            response.put("result", result);
            response.put("timestamp", System.currentTimeMillis());
            return ResponseEntity.ok(response);
        } catch (Exception e) {
            log.error("知识库搜索异常", e);
            response.put("success", false);
            response.put("error", "知识库搜索服务暂时不可用");
            return ResponseEntity.internalServerError().body(response);
        }
    }

    /**
     * 异步处理复杂请求
     */
    @PostMapping("/chat/async")
    public ResponseEntity<Map<String, Object>> chatAsync(
            @RequestParam String userId,
            @RequestBody Map<String, String> request) {
        
        String message = request.get("message");
        Map<String, Object> response = new HashMap<>();
        
        CompletableFuture<String> future = aiCustomerService.processComplexRequest(userId, message);
        
        response.put("success", true);
        response.put("message", "请求已提交，正在处理中...");
        response.put("userId", userId);
        response.put("taskId", System.currentTimeMillis());
        
        // 实际应用中可以返回任务ID，客户端轮询结果
        return ResponseEntity.ok(response);
    }
}

7. 启动类

package com.example;

import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.scheduling.annotation.EnableAsync;

@SpringBootApplication
@EnableAsync
public class AiCustomerServiceApplication {
    public static void main(String[] args) {
        // 设置Sentinel Dashboard地址
        System.setProperty("csp.sentinel.dashboard.server", "localhost:8080");
        System.setProperty("project.name", "ai-customer-service");
        
        SpringApplication.run(AiCustomerServiceApplication.class, args);
    }
}

8. 测试用例

package com.example.test;

import org.junit.jupiter.api.Test;
import org.springframework.boot.test.context.SpringBootTest;
import org.springframework.test.context.junit.jupiter.SpringJUnitConfig;

import java.util.concurrent.CountDownLatch;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;

@SpringJUnitConfig
@SpringBootTest
public class RateLimitTest {

    @Test
    public void testConcurrentRequests() throws InterruptedException {
        ExecutorService executor = Executors.newFixedThreadPool(20);
        CountDownLatch latch = new CountDownLatch(50);
        
        for (int i = 0; i < 50; i++) {
            final int requestId = i;
            executor.submit(() -> {
                try {
                    // 模拟并发请求
                    System.out.println("请求 " + requestId + " 开始");
                    Thread.sleep(100);
                    System.out.println("请求 " + requestId + " 完成");
                } catch (InterruptedException e) {
                    Thread.currentThread().interrupt();
                } finally {
                    latch.countDown();
                }
            });
        }
        
        latch.await();
        executor.shutdown();
    }
}

使用示例

1. 启动应用

mvn spring-boot:run

2. 测试接口

聊天接口测试：

curl -X POST "http://localhost:8080/api/ai/chat?userId=user123" \
     -H "Content-Type: application/json" \
     -d '{"message": "你好，我想了解产品信息"}'

问答接口测试：

curl -X POST "http://localhost:8080/api/ai/qa" \
     -H "Content-Type: application/json" \
     -d '{"question": "如何退换货？"}'

知识库搜索测试：

curl "http://localhost:8080/api/ai/knowledge/search?keyword=退款政策"

关键特性说明

  1. 多种限流算法: 演示了令牌桶、预热、排队等待等不同算法的使用场景
  2. 热点参数限流: 基于用户ID进行个性化限流
  3. 优雅降级: 提供友好的限流提示而非直接拒绝
  4. 异步处理: 对于被限流的复杂请求提供异步处理选项
  5. 监控友好: 集成Sentinel Dashboard进行实时监控

这个案例展示了如何在实际的AI应用中合理使用Sentinel的各种限流算法，既保护了系统稳定性，也提供了良好的用户体验。

avatar
文章
阅读
粉丝