原文链接
前言
前段时间由于种种原因我的AI BOT网站停运了数天,后来申请了百度的文心一言和阿里的通义千问开放接口,文心一言的接口很快就通过了,但是文心一言至今杳无音讯。文心一言通过审之后,很快将AI BOT的AI能力接入了文心一言,这里记录一下具体的接入方案。
文心一言应用创建
首先需要先申请文心千帆大模型,申请地址:文心一言 (baidu.com),点击加入体验,等通过审核之后就可以进入文心千帆大模型后台进行应用管理了。
在百度智能云首页即可看到文心千帆大模型平台
然后进入后台管理之后,点击应用接入,创建应用即可
创建完应应用之后,便可以调用文心一言的http开发接口进行交互了。
接口对接
接口文档
首先要看一下接口文档:API调用指南 - 文心千帆WENXINWORKSHOP | 百度智能云文档 (baidu.com)
这里我用的是ERNIE-Bot-turbo API,主要是由于它响应更快。
下面介绍一下具体接入的代码
代码示例
依赖
- 依赖安装
<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
<parent>
<artifactId>baidu</artifactId>
<groupId>com.walter</groupId>
<version>1.0</version>
</parent>
<modelVersion>4.0.0</modelVersion>
<version>1.0</version>
<artifactId>baidumodel</artifactId>
<description>百度大模型</description>
<repositories>
<repository>
<id>public</id>
<name>aliyun nexus</name>
<url>http://maven.aliyun.com/nexus/content/groups/public/</url>
<releases>
<enabled>true</enabled>
</releases>
<snapshots>
<enabled>false</enabled>
</snapshots>
</repository>
</repositories>
<dependencies>
<dependency>
<groupId>com.fasterxml.jackson.core</groupId>
<artifactId>jackson-databind</artifactId>
<version>2.11.1</version>
</dependency>
<dependency>
<groupId>cn.hutool</groupId>
<artifactId>hutool-http</artifactId>
<version>5.8.11</version>
</dependency>
<dependency>
<groupId>junit</groupId>
<artifactId>junit</artifactId>
<scope>test</scope>
</dependency>
<dependency>
<groupId>cn.hutool</groupId>
<artifactId>hutool-json</artifactId>
<version>5.8.11</version>
</dependency>
<dependency>
<groupId>org.projectlombok</groupId>
<artifactId>lombok</artifactId>
<version>1.18.12</version>
</dependency>
<dependency>
<groupId>com.squareup.okhttp3</groupId>
<artifactId>okhttp-sse</artifactId>
<version>3.14.9</version>
</dependency>
<dependency>
<groupId>org.slf4j</groupId>
<artifactId>slf4j-log4j12</artifactId>
<version>1.7.30</version>
</dependency>
<dependency>
<groupId>org.slf4j</groupId>
<artifactId>slf4j-api</artifactId>
<version>1.7.30</version>
</dependency>
</dependencies>
</project>
常量类
- ApiConstant.java
@Slf4j
public class ApiConstant {
/**
* ERNIE_BOT_TURBO 发起会话接口
*/
public static final String ERNIE_BOT_TURBO_INSTANT = "https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/chat/eb-instant?access_token=";
public static String getToken(String appKey, String secretKey) {
String url = "https://aip.baidubce.com/oauth/2.0/token?grant_type=client_credentials&client_id=" + appKey + "&client_secret=" + secretKey;
String s = HttpUtil.get(url);
Token bean = JSONUtil.toBean(s, Token.class);
return bean.getAccess_token();
}
}
实体类
- BaiduChatMessage.java
@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
public class BaiduChatMessage implements Serializable {
private String role;
private String content;
}
- ErnieBotTurboResponse.java
@Data
public class ErnieBotTurboResponse implements Serializable {
private String id;
private String object;
private Integer created;
private String sentence_id;
private Boolean is_end;
private Boolean is_truncated;
private String result;
private Boolean need_clear_history;
private Usage usage;
@Data
public static class Usage implements Serializable {
private Integer prompt_tokens;
private Integer completion_tokens;
private Integer total_tokens;
}
}
- ErnieBotTurboStreamParam.java
@Data
@Builder
@NoArgsConstructor
@AllArgsConstructor
public class ErnieBotTurboStreamParam implements Serializable {
private List<BaiduChatMessage> messages;
private Boolean stream;
private String user_id;
public boolean isStream() {
return Objects.equals(this.stream, true);
}
}
- Token.java
@Data
public class Token implements Serializable {
private String access_token;
private Integer expires_in;
private String error;
private String error_description;
}
- BaiduEventSourceListener.java
// 这个类主要是为了与文心一言API建立流式连接,实现数据的实时返回,而不是等完整的数据生成之后才将数据返回
// 可以减少用户等待时间,实现更好的交互体验
@Slf4j
public class BaiduEventSourceListener extends EventSourceListener {
@Override
public void onOpen(EventSource eventSource, Response response) {
log.info("baidu建立sse连接...");
}
@Override
public void onEvent(EventSource eventSource, String id, String type, String data) {
log.info("baidu返回数据:{}", data);
}
@Override
public void onClosed(EventSource eventSource) {
log.info("baidu关闭sse连接...");
}
@SneakyThrows
@Override
public void onFailure(EventSource eventSource, Throwable t, Response response) {
if(Objects.isNull(response)){
log.error("baidu sse连接异常:{}", t);
eventSource.cancel();
return;
}
ResponseBody body = response.body();
if (Objects.nonNull(body)) {
log.error("baidu sse连接异常data:{},异常:{}", body.string(), t);
} else {
log.error("baidu sse连接异常data:{},异常:{}", response, t);
}
eventSource.cancel();
}
}
- BaiduService.java
// 该类主要是处理接口请求,处理接口响应逻辑
@Slf4j
@Data
public class BaiduService {
private static final long TIME_OUT = 30;
private OkHttpClient okHttpClient;
private String appKey;
private String secretKey;
public BaiduService(@NonNull String appKey, @NonNull String secretKey) {
this.appKey = appKey;
this.secretKey = secretKey;
this.okHttpClient(30, 30, 30, null);
}
public BaiduService(@NonNull String appKey, @NonNull String secretKey, long connectTimeout, long writeTimeout, long readTimeout, Proxy proxy) {
this.appKey = appKey;
this.secretKey = secretKey;
this.okHttpClient(connectTimeout, writeTimeout, readTimeout, proxy);
}
private void okHttpClient(long connectTimeout, long writeTimeout, long readTimeout, Proxy proxy) {
OkHttpClient.Builder client = new OkHttpClient.Builder();
client.connectTimeout(connectTimeout, TimeUnit.SECONDS);
client.writeTimeout(writeTimeout, TimeUnit.SECONDS);
client.readTimeout(readTimeout, TimeUnit.SECONDS);
if (Objects.nonNull(proxy)) {
client.proxy(proxy);
}
this.okHttpClient = client.build();
}
// 该方法是同步请求API,会等大模型将数据完全生成之后,返回响应结果,可能需要等待较长时间,视生成文本长度而定
public ErnieBotTurboResponse ernieBotTurbo(ErnieBotTurboStreamParam param) {
if (param == null) {
log.error("参数异常:param不能为空");
throw new RuntimeException("参数异常:param不能为空");
}
if (param.isStream()) {
param.setStream(false);
}
String post = HttpUtil.post(ApiConstant.ERNIE_BOT_TURBO_INSTANT + ApiConstant.getToken(appKey, secretKey), JSONUtil.toJsonStr(param));
return JSONUtil.toBean(post, ErnieBotTurboResponse.class);
}
// 该方法是通过流的方式请求API,大模型每生成一些字符,就会通过流的方式相应给客户端,
// 我们是在 BaiduEventSourceListener.java 的 onEvent 方法中获取大模型响应的数据,其中data就是具体的数据,
// 我们获取到数据之后,就可以通过 SSE/webscocket 的方式实时相应给前端页面展示
public void ernieBotTurboStream(ErnieBotTurboStreamParam param, EventSourceListener eventSourceListener) {
if (Objects.isNull(eventSourceListener)) {
log.error("参数异常:EventSourceListener不能为空");
throw new RuntimeException("参数异常:EventSourceListener不能为空");
}
if (param == null) {
log.error("参数异常:param不能为空");
throw new RuntimeException("参数异常:param不能为空");
}
if (!param.isStream()) {
param.setStream(true);
}
try {
EventSource.Factory factory = EventSources.createFactory(this.okHttpClient);
ObjectMapper mapper = new ObjectMapper();
String requestBody = mapper.writeValueAsString(param);
Request request = new Request.Builder()
.url(ApiConstant.ERNIE_BOT_TURBO_INSTANT + ApiConstant.getToken(appKey, secretKey))
.post(RequestBody.create(MediaType.parse(ContentType.JSON.getValue()), requestBody))
.build();
//创建事件
EventSource eventSource = factory.newEventSource(request, eventSourceListener);
} catch (JsonProcessingException e) {
log.error("请求参数解析是失败!", e);
throw new RuntimeException("请求参数解析是失败!", e);
}
}
}
结束语
以上就是通过文心一言的OpenAPI与大模型交互的整体逻辑,等代码功能再做完善之后,改代码会以SDK的方式开源到Gitee,欢迎一起探讨