2025 最新技术：手把手教你 AI 与传统 Java 应用集成的实操指南

<!-- DJL核心依赖 -->
<dependency>
    <groupId>ai.djl</groupId>
    <artifactId>djl-spring-boot-starter</artifactId>
    <version>0.27.0</version>
</dependency>
<!-- Ollama客户端 -->
<dependency>
    <groupId>com.github.ollama-java</groupId>
    <artifactId>ollama-java</artifactId>
    <version>1.1.0</version>
</dependency>
<!-- Milvus客户端 -->
<dependency>
    <groupId>io.milvus</groupId>
    <artifactId>milvus-sdk-java</artifactId>
    <version>2.4.0</version>
</dependency>

spring:
  application:
    name: crm-ai-integration

# Ollama配置
ollama:
  base-url: http://localhost:11434
  default-model: qwen2:7b

# Milvus配置
milvus:
  host: localhost
  port: 19530
  database: default

public interface SentimentAnalysisService {
    /**
     * 分析文本情感
     * @param text 待分析文本
     * @return 情感分析结果
     */
    SentimentResult analyze(String text);
}

@Service
public class ConversationService {
    private final SentimentAnalysisService sentimentAnalysisService;
    private final ConversationRepository conversationRepository;
    
    // 构造函数注入依赖（略）
    
    /**
     * 保存对话记录并进行情感分析
     */
    @Transactional
    public Conversation saveConversation(ConversationDTO dto) {
        // 1. 保存原始对话记录
        Conversation conversation = new Conversation();
        conversation.setContent(dto.getContent());
        conversation.setCustomerId(dto.getCustomerId());
        conversation.setSalesId(dto.getSalesId());
        conversation.setCreateTime(LocalDateTime.now());
        
        // 2. 调用AI进行情感分析
        SentimentResult result = sentimentAnalysisService.analyze(dto.getContent());
        
        // 3. 保存分析结果
        conversation.setSentiment(result.getSentiment());
        conversation.setSentimentScore(result.getScore());
        
        return conversationRepository.save(conversation);
    }
}

@Service
public class EmbeddingService {
    private final Model embeddingModel;
    private final Predictor<String[], float[][]> embeddingPredictor;

    public EmbeddingService() throws ModelException, IOException {
        // 加载嵌入模型（使用Sentence-BERT）
        this.embeddingModel = ModelZoo.loadModel(new Criteria.Builder<String[], float[][]>()
                .setTypes(String[].class, float[][].class)
                .optModelUrls("djl://ai.djl.huggingface.pytorch/sentence-transformers/all-MiniLM-L6-v2")
                .optTranslator(new TextEmbeddingTranslator())
                .build());
        
        this.embeddingPredictor = embeddingModel.newPredictor();
    }
    
    /**
     * 将文本转换为向量
     */
    public float[] embed(String text) {
        try {
            float[][] embeddings = embeddingPredictor.predict(new String[]{text});
            return embeddings[0];
        } catch (PredictException e) {
            throw new RuntimeException("文本嵌入失败", e);
        }
    }
}

@SpringBootTest
public class AiIntegrationTests {
    @Autowired
    private SentimentAnalysisService sentimentAnalysisService;
    
    @Autowired
    private EmailGenerationService emailGenerationService;
    
    @Test
    void testSentimentAnalysis() {
        String positiveText = "这款产品非常棒，解决了我们长期以来的难题，非常满意！";
        SentimentResult result = sentimentAnalysisService.analyze(positiveText);
        assertEquals("positive", result.getSentiment());
        assertTrue(result.getScore() > 0.7);
    }
    
    @Test
    void testEmailGeneration() {
        String email = emailGenerationService.generateSalesEmail(1L, 1001L);
        assertNotNull(email);
        assertTrue(email.contains("主题:"));
        assertTrue(email.length() > 200);
    }
}

  rateLimiter.acquire(); // 阻塞直到获取令牌

  return analyze(text);

FROM eclipse-temurin:17-jre-alpine

WORKDIR /app

COPY target/crm-ai-integration.jar app.jar

# 设置JVM参数，优化AI模型运行性能
ENV JAVA_OPTS="-XX:+UseContainerSupport -XX:MaxRAMPercentage=75.0"

EXPOSE 8080

ENTRYPOINT ["sh", "-c", "java $JAVA_OPTS -jar app.jar"]

<dependency>
    <groupId>org.springframework.boot</groupId>
    <artifactId>spring-boot-starter-actuator</artifactId>
</dependency>

management:
  endpoints:
    web:
      exposure:
        include: health,info,metrics,prometheus
  metrics:
    export:
      prometheus:
        enabled: true
  endpoint:
    health:
      show-details: always

@Component
public class AiMetrics {
    private final MeterRegistry meterRegistry;
    private final Timer sentimentAnalysisTimer;
    private final Counter sentimentAnalysisCounter;
    
    public AiMetrics(MeterRegistry meterRegistry) {
        this.meterRegistry = meterRegistry;
        this.sentimentAnalysisTimer = Timer.builder("ai.sentiment.analysis.time")
                .description("情感分析耗时")
                .register(meterRegistry);
        this.sentimentAnalysisCounter = Counter.builder("ai.sentiment.analysis.count")
                .description("情感分析调用次数")
                .register(meterRegistry);
    }
    
    // 用于计时的包装方法
    public <T> T timeSentimentAnalysis(Supplier<T> task) {
        sentimentAnalysisCounter.increment();
        return sentimentAnalysisTimer.record(task);
    }
}