1
<dependency>
2
    <groupId>dev.langchain4j</groupId>
3
    <artifactId>langchain4j</artifactId>
4
    <version>1.1.0</version>
5
</dependency>

1
public interface AiCodeHelperService {
2
    @SystemMessage("你是一位编程小助手")
3
    String chat(String userMessage);
4
}

1
public interface AiCodeHelperService {
2
    @SystemMessage(fromResource = "system-prompt.txt")
3
    String chat(String userMessage);
4
}

1
@Configuration
2
public class AiCodeHelperServiceFactory {
3
    @Resource
4
    private ChatModel qwenChatModel;
5

6
    @Bean
7
    public AiCodeHelperService aiCodeHelperService() {
8
        return AiServices.create(AiCodeHelperService.class, qwenChatModel);
9
    }
10
}

1
/**
2
 * 测试AI服务是否可用 * * @author ShineAcZ
3
 * @date 2026/4/25
4
 */@SpringBootTest
5
class AiCodeHelperServiceTest {
6

7
    @Resource
8
    public AiCodeHelperService aiCodeHelperService;
9

10
    @Test
11
    void testChat() {
12
        String result1 = aiCodeHelperService.chat("Java中的stream流如何使用?");
13
        System.out.println(result1);
14
    }
15
}

1
<dependency>
2
    <groupId>dev.langchain4j</groupId>
3
    <artifactId>langchain4j-spring-boot-starter</artifactId>
4
    <version>1.1.0-beta7</version>
5
</dependency>

1
@AiService
2
public interface AiCodeHelperService {
3

4
    @SystemMessage(fromResource = "system-prompt.txt")
5
    String chat(String userMessage);
6
}

1
@Configuration
2
public class AiCodeHelperServiceFactory {
3

4
    @Resource
5
    private ChatModel qwenChatModel;
6

7
    @Bean
8
    public AiCodeHelperService aiCodeHelperService() {
9
        // 会话记忆
10
        ChatMemory chatMemory = MessageWindowChatMemory.withMaxMessages(10);
11
        AiCodeHelperService aiCodeHelperService = AiServices.builder(AiCodeHelperService.class)
12
                .chatModel(qwenChatModel)
13
                .chatMemory(chatMemory)
14
                .build();
15
        return aiCodeHelperService;
16
    }
17
}

1
@Test
2
void chatWithMemory() {
3
    String result = aiCodeHelperService.chat("你好，我是程序员鱼皮");
4
    System.out.println(result);
5
    result = aiCodeHelperService.chat("你好，我是谁来着？");
6
    System.out.println(result);
7
}

1
String chat(@MemoryId int memoryId, @UserMessage String userMessage);

1
// 构造 AI Service
2
AiCodeHelperService aiCodeHelperService = AiServices.builder(AiCodeHelperService.class)
3
        .chatModel(qwenChatModel)
4
        .chatMemoryProvider(memoryId -> MessageWindowChatMemory.withMaxMessages(10))
5
        .build();

1
<dependency>
2
    <groupId>dev.langchain4j</groupId>
3
    <artifactId>langchain4j-easy-rag</artifactId>
4
    <version>1.1.0-beta7</version>
5
</dependency>

1
// RAG
2
// 1. 加载文档
3
List<Document> documents = FileSystemDocumentLoader.loadDocuments("src/main/resources/docs");
4
// 2. 使用内置的 EmbeddingModel 转换文本为向量，然后存储到自动注入的内存 embeddingStore 中
5
EmbeddingStoreIngestor.ingest(documents, embeddingStore);
6
// 构造 AI Service
7
AiCodeHelperService aiCodeHelperService = AiServices.builder(AiCodeHelperService.class)
8
        .chatModel(qwenChatModel)
9
        .chatMemory(chatMemory)
10
        // RAG：从内存 embeddingStore 中检索匹配的文本片段
11
        .contentRetriever(EmbeddingStoreContentRetriever.from(embeddingStore))
12
        .build();

1
langchain4j:
2
  community:
3
    dashscope:
4
      chat-model:
5
        model-name: qwen-max
6
        api-key: <You API Key here>
7
      embedding-model:
8
        model-name: text-embedding-v4
9
        api-key: <You API Key here>

1
/**
2
 * Rag标准版配置类
3
 * @author ShineAcZ
4
 * @date 2026/4/30
5
 */
6
@Configuration
7
public class RagConfig {
8

9
    @Resource
10
    private EmbeddingModel qwenEmbeddingModel;
11

12
    // 内存向量数据库(存文档切片)
13
    @Resource
14
    private EmbeddingStore<TextSegment> embeddingStore;
15

16
    // 内容检索器
17
    @Bean
18
    public ContentRetriever contentRetriever(){
19
        // ---RAG---
20
        // 1. 加载文档        List<Document> documents = FileSystemDocumentLoader.loadDocuments("src/main/resources/docs");
21
        // 2. 切割文档:按照段落切割,最大1000字符,最大重叠200字符(有多种不同的切割策略,但是这里用的是按段落切割)
22
        DocumentByParagraphSplitter documentByParagraphSplitter = new DocumentByParagraphSplitter(1000, 200);
23
        // 3. 定义文档加载器,把文档转化为向量并存到向量数据库中
24
        EmbeddingStoreIngestor embeddingStoreIngestor = EmbeddingStoreIngestor.builder()
25
                .documentSplitter(documentByParagraphSplitter)
26
                // 给文档切片加上元数据信息(这里是加上了文件名在前面,可以给ai提供更多可用信息)
27
                .textSegmentTransformer(textSegment -> TextSegment.from(textSegment.metadata().getString("file_name")+"\n"+textSegment.text(),textSegment.metadata()))
28
                // 使用的向量模型
29
                .embeddingModel(qwenEmbeddingModel)
30
                .embeddingStore(embeddingStore)
31
                .build();
32
        // 加载文档
33
        embeddingStoreIngestor.ingest(documents);
34
        // 4. 自定义内容检索器
35
        return EmbeddingStoreContentRetriever.builder()
36
                .embeddingStore(embeddingStore)
37
                .embeddingModel(qwenEmbeddingModel)
38
                // 最多五条
39
                .maxResults(5)
40
                // 最低相似度(低于这个值将查不出来,需要结果测试才能得知合适的值)
41
                .minScore(0.75)
42
                .build();
43
    }
44
}

1
@Resource
2
private ContentRetriever contentRetriever;
3

4
@Bean
5
public AiCodeHelperService aiCodeHelperService() {
6
    // 会话记忆
7
    ChatMemory chatMemory = MessageWindowChatMemory.withMaxMessages(10);
8
    // 构造 AI Service
9
    AiCodeHelperService aiCodeHelperService = AiServices.builder(AiCodeHelperService.class)
10
            .chatModel(qwenChatModel)
11
            .chatMemory(chatMemory)
12
            .contentRetriever(contentRetriever) // RAG 检索增强生成
13
            .build();
14
    return aiCodeHelperService;
15
}

1
@Test
2
void testChatRAG() {
3
    String chat = aiCodeHelperService.chat("怎么学习Java?有哪些常见的面试题?");
4
    System.out.println(chat);
5
}

1
@SystemMessage(fromResource = "system-prompt.txt")
2
Result<String> chatWithRag(String userMessage);

1
@Test
2
void chatWithRag() {
3
    Result<String> result = aiCodeHelperService.chatWithRag("怎么学习 Java？有哪些常见面试题？");
4
    String content = result.content();
5
    List<Content> sources = result.sources();
6
    System.out.println(content);
7
    System.out.println(sources);
8
}

1
AiServices.builder(xxx.class)
2
    ...
3
    .retrievalAugmentor(retrievalAugmentor)
4
    .build();

1
OpenAiChatModel.builder()
2
    ...
3
    .logRequests(true)
4
    .logResponses(true)
5
    .build();
6
langchain4j.open-ai.chat-model.log-requests = true
7
langchain4j.open-ai.chat-model.log-responses = true
8
logging.level.dev.langchain4j = DEBUG

1
@Configuration
2
@Slf4j
3
public class ChatModelListenerConfig {
4

5
    @Bean
6
    ChatModelListener chatModelListener() {
7
        return new ChatModelListener() {
8
            @Override
9
            public void onRequest(ChatModelRequestContext requestContext) {
10
                log.info("onRequest(): {}", requestContext.chatRequest());
11
            }
12

13
            @Override
14
            public void onResponse(ChatModelResponseContext responseContext) {
15
                log.info("onResponse(): {}", responseContext.chatResponse());
16
            }
17

18
            @Override
19
            public void onError(ChatModelErrorContext errorContext) {
20
                log.info("onError(): {}", errorContext.error().getMessage());
21
            }
22
        };
23
    }
24
}

1
@Configuration
2
@ConfigurationProperties(prefix = "langchain4j.community.dashscope.chat-model")
3
@Data
4
public class QwenChatModelConfig {
5

6
    private String modelName;
7

8
    private String apiKey;
9

10
    @Resource
11
    private ChatModelListener chatModelListener;
12

13
    @Bean
14
    public ChatModel myQwenChatModel() {
15
        return QwenChatModel.builder()
16
                .apiKey(apiKey)
17
                .modelName(modelName)
18
                .listeners(List.of(chatModelListener))
19
                .build();
20
    }
21
}

1
interface Assistant {
2

3
    TokenStream chat(String message);
4
}
5

6
StreamingChatModel model = OpenAiStreamingChatModel.builder()
7
    .apiKey(System.getenv("OPENAI_API_KEY"))
8
    .modelName(GPT_4_O_MINI)
9
    .build();
10

11
Assistant assistant = AiServices.create(Assistant.class, model);
12

13
TokenStream tokenStream = assistant.chat("Tell me a joke");
14

15
tokenStream.onPartialResponse((String partialResponse) -> System.out.println(partialResponse))
16
    .onRetrieved((List<Content> contents) -> System.out.println(contents))
17
    .onToolExecuted((ToolExecution toolExecution) -> System.out.println(toolExecution))
18
    .onCompleteResponse((ChatResponse response) -> System.out.println(response))
19
    .onError((Throwable error) -> error.printStackTrace())
20
    .start();

1
interface Assistant {
2

3
  Flux<String> chat(String message);
4
}

1
<dependency>
2
    <groupId>dev.langchain4j</groupId>
3
    <artifactId>langchain4j-reactor</artifactId>
4
    <version>1.1.0-beta7</version>
5
</dependency>

1
// 流式对话
2
Flux<String> chatStream(@MemoryId int memoryId, @UserMessage String userMessage);

1
langchain4j:
2
  community:
3
    dashscope:
4
      streaming-chat-model:
5
        model-name: qwen-max
6
        api-key: <Your Api Key>

1
@Resource
2
private StreamingChatModel qwenStreamingChatModel;
3

4
AiCodeHelperService aiCodeHelperService = AiServices.builder(AiCodeHelperService.class)
5
        .chatModel(myQwenChatModel)
6
        .streamingChatModel(qwenStreamingChatModel)
7
        .chatMemory(chatMemory)
8
        .chatMemoryProvider(memoryId ->
9
                MessageWindowChatMemory.withMaxMessages(10)) // 每个会话独立存储
10
        .contentRetriever(contentRetriever) // RAG 检索增强生成
11
        .tools(new InterviewQuestionTool()) // 工具调用
12
        .toolProvider(mcpToolProvider) // MCP 工具调用
13
        .build();

1
@RestController
2
@RequestMapping("/ai")
3
public class AiController {
4

5
    @Resource
6
    private AiCodeHelperService aiCodeHelperService;
7

8
    @GetMapping("/chat")
9
    public Flux<ServerSentEvent<String>> chat(int memoryId, String message) {
10
        return aiCodeHelperService.chatStream(memoryId, message)
11
                .map(chunk -> ServerSentEvent.<String>builder()
12
                        .data(chunk)
13
                        .build());
14
    }
15
}

1
server:
2
  port: 8081
3
  servlet:
4
    context-path: /api

1
curl -G 'http://localhost:8081/api/ai/chat' \
2
  --data-urlencode 'message=我是小明' \
3
  --data-urlencode 'memoryId=1'

1
/**
2
 * 全局跨域配置
3
 */
4
@Configuration
5
public class CorsConfig implements WebMvcConfigurer {
6

7
    @Override
8
    public void addCorsMappings(CorsRegistry registry) {
9
        // 覆盖所有请求
10
        registry.addMapping("/**")
11
                // 允许发送 Cookie
12
                .allowCredentials(true)
13
                // 放行哪些域名（必须用 patterns，否则 * 会和 allowCredentials 冲突）
14
                .allowedOriginPatterns("*")
15
                .allowedMethods("GET", "POST", "PUT", "DELETE", "OPTIONS")
16
                .allowedHeaders("*")
17
                .exposedHeaders("*");
18
    }
19
}