DeepSeek V4 + Spring Boot 3 实战：从零搭一个生产级AI接口，代码全公开

本文介绍了如何将DeepSeek V4大模型API集成到SpringBoot 3项目中。主要内容包括：1) 环境准备与项目创建；2) 核心依赖配置；3) 封装DeepSeek客户端；4) 实现AI服务类，支持流式和非流式调用；5) 构建REST接口；6) 测试验证方法。文章提供了完整的代码示例，重点解决了Java开发者集成AI服务的痛点，包括流式输出、异常重试等功能，并给出了生产环境部署建议。作者

若谷6

255人浏览 · 2026-05-10 19:45:00

若谷6 · 2026-05-10 19:45:00 发布

为什么是DeepSeek V4 + Spring Boot 3？

最近DeepSeek V4全面开放API，MoE架构升级后推理速度提升了将近40%，价格反而降了。对Java开发者来说，最大的痛从来不是"AI能不能用"，而是"怎么把它干净地集成到现有项目里"。

本文手把手带你从零搭一个Spring Boot 3 + DeepSeek V4的生产级AI接口，包括流式输出、上下文管理、异常重试——全部代码可复现。

一、准备工作

1.1 环境要求

JDK 17+（Spring Boot 3 要求）
Maven 3.8+ 或 Gradle 8+
DeepSeek API Key（platform.deepseek.com 注册获取）

1.2 创建项目

用 Spring Initializr 生成项目骨架，依赖选择：

Spring Web
Spring Boot Actuator（可选，生产必备）
Lombok（减少样板代码）
WebFlux（流式输出需要）

二、核心依赖与配置

2.1 pom.xml 关键依赖

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

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

<dependency>
    <groupId>org.projectlombok</groupId>
    <artifactId>lombok</artifactId>
    <optional>true</optional>
</dependency>

<dependency>
    <groupId>com.squareup.okhttp3</groupId>
    <artifactId>okhttp-sse</artifactId>
    <version>4.12.0</version>
</dependency>

2.2 application.yml 配置

deepseek:
  api-key: ${DEEPSEEK_API_KEY}
  base-url: https://api.deepseek.com/v1
  model: deepseek-chat
  timeout: 60s

spring:
  jackson:
    property-naming-strategy: SNAKE_CASE

三、封装DeepSeek客户端

3.1 配置类

@Configuration
@ConfigurationProperties(prefix = "deepseek")
@Data
public class DeepSeekConfig {
    private String apiKey;
    private String baseUrl = "https://api.deepseek.com/v1";
    private String model = "deepseek-chat";
    private Duration timeout = Duration.ofSeconds(60);
}

3.2 WebClient 配置

@Configuration
public class WebClientConfig {
    
    @Bean
    public WebClient deepSeekClient(DeepSeekConfig config) {
        return WebClient.builder()
            .baseUrl(config.getBaseUrl())
            .defaultHeader(HttpHeaders.AUTHORIZATION, "Bearer " + config.getApiKey())
            .defaultHeader(HttpHeaders.CONTENT_TYPE, MediaType.APPLICATION_JSON_VALUE)
            .build();
    }
}

四、核心服务实现

4.1 消息结构体

@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
public class ChatMessage {
    private String role;
    private String content;
}

@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
public class ChatRequest {
    private String model;
    private List<ChatMessage> messages;
    private Boolean stream;
    @JsonProperty("max_tokens")
    private Integer maxTokens;
}

@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
public class ChatResponse {
    private String id;
    private String object;
    private Long created;
    private String model;
    private List<Choice> choices;
    private Usage usage;
}

@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
public class Choice {
    private Integer index;
    private ChatMessage message;
    @JsonProperty("finish_reason")
    private String finishReason;
}

@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
public class Usage {
    @JsonProperty("prompt_tokens")
    private Integer promptTokens;
    @JsonProperty("completion_tokens")
    private Integer completionTokens;
    @JsonProperty("total_tokens")
    private Integer totalTokens;
}

4.2 AI服务类（核心逻辑）

@Service
@RequiredArgsConstructor
@Slf4j
public class DeepSeekService {

    private final WebClient deepSeekClient;
    private final DeepSeekConfig config;

    // 非流式调用
    public ChatResponse chat(List<ChatMessage> messages) {
        ChatRequest request = ChatRequest.builder()
            .model(config.getModel())
            .messages(messages)
            .stream(false)
            .maxTokens(4096)
            .build();

        return deepSeekClient.post()
            .uri("/chat/completions")
            .bodyValue(request)
            .retrieve()
            .bodyToMono(ChatResponse.class)
            .block(config.getTimeout());
    }

    // 流式调用（SSE）
    public Flux<String> chatStream(List<ChatMessage> messages) {
        ChatRequest request = ChatRequest.builder()
            .model(config.getModel())
            .messages(messages)
            .stream(true)
            .maxTokens(4096)
            .build();

        return deepSeekClient.post()
            .uri("/chat/completions")
            .bodyValue(request)
            .retrieve()
            .bodyToFlux(String.class)
            .filter(content -> content != null && !content.isEmpty());
    }

    // 带重试的调用
    @Retryable(
        retryFor = {TimeoutException.class, WebClientResponseException.class},
        maxAttempts = 3,
        backoff = @Backoff(delay = 1000, multiplier = 2)
    )
    public ChatResponse chatWithRetry(List<ChatMessage> messages) {
        return chat(messages);
    }
}

五、Controller 层

5.1 REST接口

@RestController
@RequestMapping("/api/ai")
@RequiredArgsConstructor
public class AIController {

    private final DeepSeekService deepSeekService;

    @PostMapping("/chat")
    public ResponseEntity<ChatResponse> chat(@RequestBody ChatRequestDTO request) {
        List<ChatMessage> messages = new ArrayList<>();
        messages.add(ChatMessage.builder()
            .role("system")
            .content("你是一个专业的Java开发助手，擅长Spring Boot和微服务架构。")
            .build());
        messages.add(ChatMessage.builder()
            .role("user")
            .content(request.getPrompt())
            .build());

        ChatResponse response = deepSeekService.chat(messages);
        return ResponseEntity.ok(response);
    }

    // 流式输出接口（SSE）
    @PostMapping(value = "/chat/stream", produces = MediaType.TEXT_EVENT_STREAM_VALUE)
    public Flux<ServerSentEvent<String>> chatStream(@RequestBody ChatRequestDTO request) {
        List<ChatMessage> messages = new ArrayList<>();
        messages.add(ChatMessage.builder()
            .role("system")
            .content("你是一个专业的Java开发助手。")
            .build());
        messages.add(ChatMessage.builder()
            .role("user")
            .content(request.getPrompt())
            .build());

        return deepSeekService.chatStream(messages)
            .map(content -> ServerSentEvent.<String>builder()
                .data(content)
                .build())
            .concatWith(Flux.just(
                ServerSentEvent.<String>builder()
                    .event("done")
                    .data("[DONE]")
                    .build()
            ));
    }
}

5.2 请求体DTO

@Data
public class ChatRequestDTO {
    @NotBlank(message = "prompt不能为空")
    private String prompt;
    
    @Size(max = 4096)
    private String systemPrompt;
    
    private Boolean stream = false;
}

六、测试验证

# 非流式调用
curl -X POST http://localhost:8080/api/ai/chat \
  -H "Content-Type: application/json" \
  -d '{"prompt": "用Java写一个单例模式，要求线程安全"}'
  
# 流式调用
curl -X POST http://localhost:8080/api/ai/chat/stream \
  -H "Content-Type: application/json" \
  -d '{"prompt": "给我讲一个程序员的笑话", "stream": true}'