基于 Java GenKit 构建生成式 AI 应用指南（完整版）

<!-- Java GenKit 核心依赖 -->
<dependency>
    <groupId>com.google.genkit</groupId>
    <artifactId>genkit-core</artifactId>
    <version>1.0.0</version>
</dependency>
<!-- 通用模型对接适配器 -->
<dependency>
    <groupId>com.google.genkit</groupId>
    <artifactId>genkit-model</artifactId>
    <version>1.0.0</version>
</dependency>
<!-- 对接国内大模型：百度文心一言适配器（示例） -->
<dependency>
    <groupId>com.baidu.ai</groupId>
    <artifactId>genkit-wenxin</artifactId>
    <version>1.0.0</version>
</dependency>
<!-- Spring Boot 集成依赖（企业级开发推荐） -->
<dependency>
    <groupId>com.google.genkit</groupId>
    <artifactId>genkit-spring-boot-starter</artifactId>
    <version>1.0.0</version>
</dependency>

<mirror>
    <id>aliyunmaven</id>
    <mirrorOf>central</mirrorOf>
    <url>https://maven.aliyun.com/repository/public</url>
</mirror>

genkit:
  model:
    # 百度文心一言配置
    wenxin:
      api-key: 你的文心一言API_KEY
      secret-key: 你的文心一言SECRET_KEY
      base-url: https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/ernie-3.5-8k
    # 阿里通义千问配置（可选）
    qianwen:
      api-key: 你的通义千问API_KEY
      base-url: https://dashscope.aliyuncs.com/api/v1/services/aigc/text-generation/generation

import com.google.genkit.core.chat.ChatCompletion;
import com.google.genkit.core.chat.ChatMessage;
import com.google.genkit.core.chat.ChatModel;
import com.google.genkit.wenxin.WenxinChatModel;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.web.bind.annotation.GetMapping;
import org.springframework.web.bind.annotation.RequestParam;
import org.springframework.web.bind.annotation.RestController;

import java.util.List;

@RestController
public class SimpleChatController {
    // 注入文心一言聊天模型（由GenKit Starter自动装配）
    @Autowired
    private WenxinChatModel wenxinChatModel;

    // 简单对话接口：接收用户问题，返回AI回答
    @GetMapping("/chat")
    public String chat(@RequestParam String question) {
        // 构造用户消息
        ChatMessage userMsg = ChatMessage.user(question);
        // 调用模型生成回答
        ChatCompletion completion = wenxinChatModel.generate(List.of(userMsg));
        // 获取AI回复内容
        return completion.getChoices().get(0).getMessage().getContent();
    }
}

import com.google.genkit.core.prompt.PromptTemplate;
import org.springframework.stereotype.Component;

import java.util.Map;

@Component
public class PromptTemplateManager {
    // 定义“Java代码生成”提示词模板
    private final PromptTemplate javaCodeGenTemplate = PromptTemplate.of(
            "请作为资深Java开发工程师，根据需求生成可运行的{framework}代码，要求：1. 符合{codeStyle}编码规范；2. 增加必要的注释；3. 给出测试示例。需求：{requirement}"
    );

    // 填充模板参数，生成最终提示词
    public String buildJavaCodeGenPrompt(Map<String, String> params) {
        return javaCodeGenTemplate.render(params);
    }
}

@Autowired
private PromptTemplateManager promptManager;

public void genJavaCode() {
    Map<String, String> params = Map.of(
            "framework", "Spring Boot",
            "codeStyle", "阿里巴巴Java开发手册",
            "requirement", "实现一个用户信息查询的Rest接口，支持根据用户ID查询"
    );
    String prompt = promptManager.buildJavaCodeGenPrompt(params);
    // 将prompt传入聊天模型，即可生成符合要求的代码
    ChatMessage userMsg = ChatMessage.user(prompt);
    String code = wenxinChatModel.generate(List.of(userMsg)).getChoices().get(0).getMessage().getContent();
    System.out.println(code);
}

import com.google.genkit.core.tool.Tool;
import com.google.genkit.core.tool.ToolInput;
import com.google.genkit.core.tool.ToolOutput;
import org.springframework.http.HttpEntity;
import org.springframework.http.HttpHeaders;
import org.springframework.http.HttpMethod;
import org.springframework.web.client.RestTemplate;

import java.util.Map;

// 天气查询工具
public class WeatherQueryTool implements Tool {
    // 高德地图天气API密钥（自行申请）
    private static final String AMAP_KEY = "你的高德地图API_KEY";
    private static final String AMAP_WEATHER_URL = "https://restapi.amap.com/v3/weather/weatherInfo?city={city}&key={key}&extensions=base";

    @Override
    public String name() {
        // 工具名称，供AI识别
        return "weather_query";
    }

    @Override
    public String description() {
        // 工具描述，告诉AI该工具的用途和入参，越详细越准确
        return "用于查询中国国内城市的实时天气信息，入参为city（城市名，如北京、上海），返回该城市的温度、天气状况、风力等基础信息";
    }

    @Override
    public ToolOutput execute(ToolInput input) {
        // 获取AI传入的城市参数
        String city = input.get("city").toString();
        // 调用高德地图天气API
        RestTemplate restTemplate = new RestTemplate();
        Map<String, String> urlParams = Map.of("city", city, "key", AMAP_KEY);
        String weatherResult = restTemplate.exchange(
                AMAP_WEATHER_URL,
                HttpMethod.GET,
                new HttpEntity<>(new HttpHeaders()),
                String.class,
                urlParams
        ).getBody();
        // 返回工具执行结果
        return ToolOutput.of(weatherResult);
    }
}

import com.google.genkit.core.tool.ToolRegistry;
import com.google.genkit.wenxin.WenxinChatModel;
import org.springframework.beans.factory.InitializingBean;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Service;

@Service
public class AiToolService implements InitializingBean {
    @Autowired
    private WenxinChatModel wenxinChatModel;

    // 初始化：将天气查询工具注册到GenKit工具注册表
    @Override
    public void afterPropertiesSet() throws Exception {
        ToolRegistry.register(new WeatherQueryTool());
        // 为文心一言模型开启工具调用能力
        wenxinChatModel.enableToolCalling(ToolRegistry.allTools());
    }

    // AI工具调用入口
    public String aiWithTool(String question) {
        ChatMessage userMsg = ChatMessage.user(question);
        // 模型会自动判断是否需要调用工具，无需手动干预
        ChatCompletion completion = wenxinChatModel.generate(List.of(userMsg));
        return completion.getChoices().get(0).getMessage().getContent();
    }
}

@Service
public class MultiModelService {
    @Autowired
    private WenxinChatModel wenxinChatModel;
    @Autowired
    private QianwenChatModel qianwenChatModel;

    // 多模型调用方法，实现兜底
    public String multiModelGenerate(String prompt) {
        ChatMessage userMsg = ChatMessage.user(prompt);
        try {
            // 优先调用文心一言
            return wenxinChatModel.generate(List.of(userMsg)).getChoices().get(0).getMessage().getContent();
        } catch (Exception e) {
            // 主模型失败，调用通义千问兜底
            System.err.println("文心一言调用失败，切换为通义千问：" + e.getMessage());
            return qianwenChatModel.generate(List.of(userMsg)).getChoices().get(0).getMessage().getContent();
        }
    }
}

@RestController
@RequestMapping("/java-qa")
public class JavaQaController {
    @Autowired
    private PromptTemplateManager promptManager;
    @Autowired
    private MultiModelService multiModelService;
    @Autowired
    private AiToolService aiToolService;

    // 通用问答接口
    @PostMapping("/ask")
    public Map<String, String> ask(@RequestBody Map<String, String> request) {
        String question = request.get("question");
        String type = request.get("type"); // question：基础问答，code：代码生成，trouble：问题排查
        // 根据类型选择对应的提示词模板
        Map<String, String> promptParams = Map.of("requirement", question);
        String prompt = switch (type) {
            case "code" -> promptManager.buildJavaCodeGenPrompt(promptParams);
            case "trouble" -> promptManager.buildJavaTroubleShootPrompt(promptParams);
            default -> promptManager.buildJavaBasicQaPrompt(promptParams);
        };
        // 调用多模型服务获取回答
        String answer = multiModelService.multiModelGenerate(prompt);
        return Map.of("code", "200", "msg", "success", "answer", answer);
    }

    // 工具调用专属接口（如查询技术文档、天气等）
    @GetMapping("/ask-with-tool")
    public Map<String, String> askWithTool(@RequestParam String question) {
        String answer = aiToolService.aiWithTool(question);
        return Map.of("code", "200", "msg", "success", "answer", answer);
    }
}

# 基于阿里云的OpenJDK 17镜像
FROM registry.cn-hangzhou.aliyuncs.com/openjdk/openjdk:17-jdk-slim
# 设定工作目录
WORKDIR /app
# 复制打包后的JAR包到容器
COPY target/xxx-0.0.1-SNAPSHOT.jar app.jar
# 配置国内镜像源，加速容器内依赖（若有）
RUN sed -i 's/deb.debian.org/mirrors.aliyun.com/g' /etc/apt/sources.list
# 暴露端口
EXPOSE 8080
# 启动命令
ENTRYPOINT ["java", "-jar", "app.jar"]

// 定义代码生成工作流
Workflow codeGenWorkflow = Workflow.builder()
        .step("analysis", req -> promptManager.buildRequirementAnalysisPrompt(req)) // 需求分析
        .step("genCode", analysisResult -> multiModelService.multiModelGenerate(analysisResult)) // 代码生成
        .step("formatCode", code -> codeFormatTool.execute(ToolInput.of(Map.of("code", code))).getContent()) // 代码格式化
        .step("genTest", code -> promptManager.buildTestGenPrompt(Map.of("code", code))) // 测试用例生成
        .build();
// 执行工作流
String finalResult = codeGenWorkflow.execute(userRequirement);