Spring Alibaba

杨Yy_

456人浏览 · 2026-04-22 19:18:35

杨Yy_ · 2026-04-22 19:18:35 发布

阿里云百炼平台介绍

得益于阿⾥百炼强⼤的模型整合能⼒, logging前述所有智能场景都能在平台上获得⼀站式模型实现, 轻松构建AI应⽤. Spring AI Alibaba 开源项⽬基于 Spring AI 构建, 是阿⾥云通义系列模型及服务在 Java AI 应⽤开发领域的最佳实践, 提供⾼层次的 AI API 抽象与云原⽣基础设施集成⽅案, 帮助开发者快速构建 AI 应⽤.

申请阿⾥云百炼平台API-KEY

参考：https://help.aliyun.com/zh/model-studio/get-api-key?spm=7145af80.321ad293.0.0.72aa5e636mSOsh

获取APIKEY

Spring Alibab快速入门

创建项⽬

添加依赖

      <dependencies>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-web</artifactId>
        </dependency>
        <dependency>
            <groupId>org.springframework.boot</groupId>
            <artifactId>spring-boot-starter-test</artifactId>
            <scope>test</scope>
        </dependency>
        <dependency>
            <groupId>org.springframework</groupId>
            <artifactId>spring-webflux</artifactId>
        </dependency>
        <dependency>
            <groupId>com.alibaba.cloud.ai</groupId>
            <artifactId>spring-ai-alibaba-starter-dashscope</artifactId>
        </dependency>
    </dependencies>
    <dependencyManagement>
        <dependencies>
            <dependency>
                <groupId>com.alibaba.cloud.ai</groupId>
                <artifactId>spring-ai-alibaba-bom</artifactId>
                <version>1.0.0.2</version>
                <type>pom</type>
                <scope>import</scope>
            </dependency>
        </dependencies>
    </dependencyManagement>

添加配置

spring:
  ai:
    dashscope:
      api-key: sk-e5dc559f523f413a9c1985ebdffc05e5

创建一个测试用例

测试图片生成

代码讲解：

DashScopeImageModel 也是实现了 ImageModel接⼝, 图像模型的配置在 DashScopeImageAutoConfiguration 定义

@AutoConfiguration(
    after = {RestClientAutoConfiguration.class, WebClientAutoConfiguration.class, SpringAiRetryAutoConfiguration.class}
)
@ConditionalOnClass({DashScopeApi.class})
@ConditionalOnProperty(
    name = {"spring.ai.model.audio.speech"},
    havingValue = "openai",
    matchIfMissing = true
)
@EnableConfigurationProperties({DashScopeConnectionProperties.class, DashScopeImageProperties.class})
@ImportAutoConfiguration(
    classes = {SpringAiRetryAutoConfiguration.class, RestClientAutoConfiguration.class, WebClientAutoConfiguration.class}
)
public class DashScopeImageAutoConfiguration {
    public DashScopeImageAutoConfiguration() {
    }

    @Bean
    @ConditionalOnMissingBean
    public DashScopeImageModel dashScopeImageModel(DashScopeConnectionProperties commonProperties, DashScopeImageProperties imageProperties, RestClient.Builder restClientBuilder, WebClient.Builder webClientBuilder, RetryTemplate retryTemplate, ResponseErrorHandler responseErrorHandler, ObjectProvider<ObservationRegistry> observationRegistry, ObjectProvider<ImageModelObservationConvention> observationConvention) {
        ResolvedConnectionProperties resolved = DashScopeConnectionUtils.resolveConnectionProperties(commonProperties, imageProperties, "image");
        DashScopeImageApi dashScopeImageApi = new DashScopeImageApi(resolved.baseUrl(), resolved.apiKey(), resolved.workspaceId(), restClientBuilder, webClientBuilder, responseErrorHandler);
        DashScopeImageModel dashScopeImageModel = new DashScopeImageModel(dashScopeImageApi, imageProperties.getOptions(), retryTemplate, (ObservationRegistry)observationRegistry.getIfUnique(() -> {
            return ObservationRegistry.NOOP;
        }));
        Objects.requireNonNull(dashScopeImageModel);
        observationConvention.ifAvailable(dashScopeImageModel::setObservationConvention);
        return dashScopeImageModel;
    }
}

当vlue值是openai的时候就会把DashScopeImageModel这个对象注入到spring中去

matchIfMissing = true这个的意思是默认装配的，如果没有创建这个配置项，openai是自动装配的，如果值是别的，那么就不会进行注入

@ConfigurationProperties("spring.ai.dashscope.image")找到配置项是什么

配置的模型默认是万象v1

下面的options是可以在里面添加的配置：模型，张数，大小等，可以在配置中进行设置

他和各大模型厂商一样都是实现了imageoptions这个接口

可以在配置中更改模型等配置信息

一个是在代码中进行定义，一个在配置项中进行定义，这两个都是一样的效果

参数介绍:

参数取值与模型有关, 下⾯介绍⼏种通⽤的

通义千问:https://help.aliyun.com/zh/model-studio/qwen-image-api

通义万相V2版:https://help.aliyun.com/zh/model-studio/text-to-image-v2-api-reference

通义万相V1版：https://help.aliyun.com/zh/model-studio/text-to-image-api-reference

⽐较简单的区分⽅式: 通义模型前缀为 qwen , 万相模型名称前缀为 wan , 早期模型为 wanx

配置项前缀为: spring.ai.dashscope.image.options

配置属性：

配置项	数据类型	默认值	描述	取值范围
`spring.ai.dashscope.image.options.model`	String	`wanx-v1`	模型名称	`wanx-v1`、`wan2.2-t2i-flash`、`wan2.2-t2i-plus`、`qwen-image`
`spring.ai.dashscope.image.options.n`	Integer	`1`	生成图片数量	万相模型：1-4；通义图像模型：仅 1
`spring.ai.dashscope.image.options.width`	Integer	`1024`	图片宽度	512-1440（万相 v2）
`spring.ai.dashscope.image.options.height`	Integer	`1024`	图片高度	512-1440（万相 v2）
`spring.ai.dashscope.image.options.size`	String	-	图片尺寸（已废弃）	`1024x1024`、`720x1280`、`512x512`
`spring.ai.dashscope.image.options.style`	String	-	生成风格（仅`wanx-v1`）	`photography`、`portrait`、`3d cartoon`、`anime`、`oil painting`、`watercolor`、`sketch`、`chinese painting`、`flat illustration`
`spring.ai.dashscope.image.options.seed`	Long	`0`	随机种子	0 ~ 4294967290
`spring.ai.dashscope.image.options.response-format`	String	`url`	返回格式	`url`（图片链接）、`b64_json`（Base64）
`spring.ai.dashscope.image.options.ref-img`	String	-	参考图 URL	有效图片链接
`spring.ai.dashscope.image.options.ref-mode`	String	-	参考图模式	`style`（风格参考）、`content`（内容参考）、`style-content`
`spring.ai.dashscope.image.options.negative-prompt`	String	-	反向提示词（不希望出现的内容）	文本描述
`spring.ai.dashscope.image.options.cfg-scale`	Float	`7.0`	提示词引导系数	1.0 ~ 30.0

例子：

yml文件

spring:
  ai:
    dashscope:
      image:
        options:
          model: wan2.2-t2i-plus
          n: 2
          width: 1024
          height: 1024
          style: photography
          seed: 123456
          response-format: url
          negative-prompt: blurry, low resolution, ugly

Java中配置

DashScopeImageOptions options = DashScopeImageOptions.builder()
    .withModel("wan2.2-t2i-flash")
    .withN(2)
    .withWidth(1024)
    .withHeight(1024)
    .withStyle("anime")
    .withSeed(987654L)
    .withNegativePrompt("distorted, bad anatomy")
    .build();

生成三张图片

语音合成

@SpringBootTest
public class AudioModelTest {
    @Autowired
    private DashScopeSpeechSynthesisModel scopeSpeechSynthesisModel;

    private  static  final  String TEXT="白日依山尽，黄河入海流。这是测试";
    @Test
    void tts(){
        SpeechSynthesisPrompt prompt=new SpeechSynthesisPrompt(TEXT);
        SpeechSynthesisResponse response = scopeSpeechSynthesisModel.call(prompt);
        File file=new File(System.getProperty("user.dir")+"/out.mp3");
        try(FileOutputStream fos=new FileOutputStream(file)){
            ByteBuffer audio = response.getResult().getOutput().getAudio();
            fos.write(audio.array());
        }catch (IOException e){
            System.out.println("写入文件失败");
        }
    }
}

生成语音需要的是

DashScopeSpeechSynthesisModel这个对象，这个对象实现了

SpeechSynthesisModel接口

这个接口主要是定义了两个方法

这两个方法分别是call，传入一个提示词，进行响应，另一个strea是流式响应

其中提示词包括两部分

第一个是文本message，就是需要文本生成语音

还有就是模型配置的一些参数

List<SpeechSynthesisMessage>

options

SpeechSynthesisOptions 这里面有一个参数model，就是可以设置模型

SpeechSynthesisResponse调用模型之后给的响应

点进result

里面有一个output点进去

这里面的ByteBuffer就是语音合成的东西

关于配置实在

DashScopeAudioSpeechAutoConfiguration这个配置类中去进行定义的

@AutoConfiguration(
    after = {RestClientAutoConfiguration.class, WebClientAutoConfiguration.class, SpringAiRetryAutoConfiguration.class}
)
@ConditionalOnClass({DashScopeApi.class})
@ConditionalOnProperty(
    name = {"spring.ai.model.audio.speech"},
    havingValue = "openai",
    matchIfMissing = true
)
@EnableConfigurationProperties({DashScopeConnectionProperties.class, DashScopeAudioSpeechSynthesisProperties.class})
@ImportAutoConfiguration(
    classes = {SpringAiRetryAutoConfiguration.class, RestClientAutoConfiguration.class, WebClientAutoConfiguration.class}
)
public class DashScopeAudioSpeechAutoConfiguration {
    public DashScopeAudioSpeechAutoConfiguration() {
    }

    @Bean
    @ConditionalOnMissingBean
    public DashScopeSpeechSynthesisModel dashScopeSpeechSynthesisModel(RetryTemplate retryTemplate, DashScopeConnectionProperties commonProperties, DashScopeAudioSpeechSynthesisProperties speechProperties) {
        DashScopeSpeechSynthesisApi dashScopeSpeechSynthesisApi = this.dashScopeSpeechSynthesisApi(commonProperties, speechProperties);
        return new DashScopeSpeechSynthesisModel(dashScopeSpeechSynthesisApi, speechProperties.getOptions(), retryTemplate);
    }

    private DashScopeSpeechSynthesisApi dashScopeSpeechSynthesisApi(DashScopeConnectionProperties commonProperties, DashScopeAudioSpeechSynthesisProperties speechSynthesisProperties) {
        ResolvedConnectionProperties resolved = DashScopeConnectionUtils.resolveConnectionProperties(commonProperties, speechSynthesisProperties, "audio.synthesis");
        return new DashScopeSpeechSynthesisApi(resolved.apiKey(), resolved.workspaceId());
    }

@ConditionalOnProperty(
name = {"spring.ai.model.audio.speech"},
havingValue = "openai",
matchIfMissing = true
)当有这个配置信息的时候就会自动的去注入到spring中去，注入过程中需要使用到的配置信息在

DashScopeAudioSpeechSynthesisProperties这个类中

前缀，默认的模型是什么音色是longhua

DashScopeSpeechSynthesisOptions getOptions()配置信息可以在这里面进行配置

通过yml配置模型和音色

    @Test
    void tts2(){
        DashScopeSpeechSynthesisOptions options= DashScopeSpeechSynthesisOptions.builder()
                .model("cosyvoice-v2")
                .voice("longanqin")
                .build();
        SpeechSynthesisPrompt prompt=new SpeechSynthesisPrompt(TEXT,options);
        SpeechSynthesisResponse response = scopeSpeechSynthesisModel.call(prompt);
        File file=new File(System.getProperty("user.dir")+"/out.mp3");
        try(FileOutputStream fos=new FileOutputStream(file)){
            ByteBuffer audio = response.getResult().getOutput().getAudio();
            fos.write(audio.array());
        }catch (IOException e){
            System.out.println("写入文件失败");
        }
    }

通过代码配置模型和音色

使⽤DashScope SDK调⽤模型

Dashscope-sdk-java 是阿⾥云官⽅提供的, 不依赖任何特定框架的Java客⼾端库. 你可以在任何普通的 Java 应⽤程序中引⼊这个 SDK, 通过编程式地调⽤它的 API 来使⽤⼤模型服务.

我们现在使用的是spring alibaba是基于sdk进行的封装 spring alibaba只能是在spring boot项目中使用

sdk只要是java项目就可以

添加sdk的依赖

 <dependency>
            <groupId>com.alibaba</groupId>
            <artifactId>dashscope-sdk-java</artifactId>
            <version>2.21.2</version>
        </dependency>

语音识别

语音识别就是把语音转换成文本

@SpringBootTest
public class AudioTranscriptionModelTest {
    @Autowired
    private  AudioTranscriptionModel transcriptionModel;
    private final String DEFAULT_MODEL = "paraformer-v2";
    @Test
    void stt(){
        Resource resource = new DefaultResourceLoader()
                .getResource("https://dashscope.oss-cnbeijing.aliyuncs.com/samples/audio/paraformer/hello_world_female2.wav");
        AudioTranscriptionResponse response = transcriptionModel.call(
                new AudioTranscriptionPrompt(
                        resource, DashScopeAudioTranscriptionOptions.builder()
                        .withModel(DEFAULT_MODEL)
                        .build()
                )
        );
        System.out.println(response.getResult().getOutput());
    }
}

关于sdk的识别：官方文档sdk识别语音

 @Test
    void sttDashscope() {
        TranscriptionParam param =
                TranscriptionParam.builder()
// 若没有将API Key配置到环境变量中, 需将apiKey替换为⾃⼰的APIKey
//.apiKey("apikey")
                .model("paraformer-v2")
// "language_hints"只⽀持paraformer-v2模型
                .parameter("language_hints", new String[]{"zh", "en"})
                .fileUrls(
                        Arrays.asList(
                                "https://dashscope.oss-cnbeijing.aliyuncs.com/samples/audio/paraformer/hello_world_female2.wav",
                                "https://dashscope.oss-cnbeijing.aliyuncs.com/samples/audio/paraformer/hello_world_male2.wav"))
                .build();
        try {
            Transcription transcription = new Transcription();
// 提交转写请求
            TranscriptionResult result = transcription.asyncCall(param);
            System.out.println("RequestId: " + result.getRequestId());
// 阻塞等待任务完成并获取结果
            result = transcription.wait(TranscriptionQueryParam.FromTranscriptionParam(param, result.getTaskId()));
// 打印结果
            System.out.println(new GsonBuilder().setPrettyPrinting().create().toJson(result.getOutput()));
        } catch (Exception e) {
            System.out.println("error: " + e);
        }
    }