hhyykk/ipms-sjy
1
1
Fork 0
mirror of https://gitee.com/hhyykk/ipms-sjy.git synced 2025-07-19 21:45:06 +08:00
Code Issues Actions Packages Projects Releases Wiki Activity

【优化】兼容阿里云千问开源模型，和付费模型

Browse Source
This commit is contained in:
cherishsince
2024-04-27 16:51:41 +08:00
parent 2ef64a0a50
commit 63c5f90596
8 changed files with 121 additions and 49 deletions
Download Patch File Download Diff File Expand all files Collapse all files

67
yudao-module-ai/yudao-spring-boot-starter-ai/src/main/java/cn/iocoder/yudao/framework/ai/chatqianwen/QianWenChatClient.java
View File

@ -4,13 +4,14 @@ import cn.iocoder.yudao.framework.ai.chat.*;
import cn.iocoder.yudao.framework.ai.chat.prompt.ChatOptions;
import cn.iocoder.yudao.framework.ai.chat.prompt.Prompt;
import cn.iocoder.yudao.framework.ai.chatqianwen.api.QianWenApi;
import cn.iocoder.yudao.framework.ai.chatqianwen.api.QianWenChatCompletionRequest;
import cn.iocoder.yudao.framework.ai.chatyiyan.exception.YiYanApiException;
import com.alibaba.dashscope.aigc.generation.GenerationResult;
import com.alibaba.dashscope.aigc.generation.models.QwenParam;
import com.alibaba.dashscope.common.Message;
import com.google.common.collect.Lists;
import io.reactivex.Flowable;
import lombok.extern.slf4j.Slf4j;
import org.jetbrains.annotations.NotNull;
import org.springframework.http.ResponseEntity;
import org.springframework.retry.RetryCallback;
import org.springframework.retry.RetryContext;
@ -71,7 +72,7 @@ public class QianWenChatClient implements ChatClient, StreamingChatClient {
return this.retryTemplate.execute(ctx -> {
// ctx 会有重试的信息
// 创建 request 请求，stream模式需要供应商支持
QianWenChatCompletionRequest request = this.createRequest(prompt, false);
QwenParam request = this.createRequest(prompt, false);
// 调用 callWithFunctionSupport 发送请求
ResponseEntity<GenerationResult> responseEntity = qianWenApi.chatCompletionEntity(request);
// 获取结果封装 chatCompletion
@ -81,11 +82,41 @@ public class QianWenChatClient implements ChatClient, StreamingChatClient {
// response.getRequestId(), response.getCode(), response.getMessage()))));
// }
// 转换为 Generation 返回
return new ChatResponse(List.of(new Generation(response.getOutput().getText())));
return new ChatResponse(response.getOutput().getChoices().stream()
.map(choices -> new Generation(choices.getMessage().getContent()))
.collect(Collectors.toList()));
});
}
private QianWenChatCompletionRequest createRequest(Prompt prompt, boolean stream) {
private QwenParam createRequest(Prompt prompt, boolean stream) {
// 获取 ChatOptions
QianWenOptions chatOptions = getChatOptions(prompt);
//
List<Message> messageList = Lists.newArrayList();
prompt.getInstructions().stream().forEach(instruction -> {
Message message = new Message();
message.setRole(instruction.getMessageType().getValue());
message.setContent(instruction.getContent());
messageList.add(message);
});
return QwenParam.builder()
.model(qianWenApi.getQianWenChatModal().getValue())
.prompt(prompt.getContents())
.messages(messageList)
.maxTokens(chatOptions.getMaxTokens())
.resultFormat(QwenParam.ResultFormat.MESSAGE)
.topP(Double.valueOf(chatOptions.getTopP()))
.topK(chatOptions.getTopK())
.temperature(chatOptions.getTemperature())
// 控制流式输出模式，即后面的内容会包含已经输出的内容；设置为True，将开启增量输出模式，后面的输出不会包含已经输出的内容，您需要自行拼接整体输出
.incrementalOutput(true)
/* set the random seed, optional, default to 1234 if not set */
.seed(100)
.apiKey(qianWenApi.getApiKey())
.build();
}
private @NotNull QianWenOptions getChatOptions(Prompt prompt) {
// 两个都为null 则没有配置文件
if (qianWenOptions == null && prompt.getOptions() == null) {
throw new ChatException("ChatOptions 未配置参数!");
@ -96,37 +127,27 @@ public class QianWenChatClient implements ChatClient, StreamingChatClient {
options = (ChatOptions) prompt.getOptions();
}
// Prompt 里面是一个 ChatOptions，用户可以随意传入，这里做一下判断
if (!(options instanceof QianWenOptions qianWenOptions)) {
if (!(options instanceof QianWenOptions)) {
throw new ChatException("Prompt 传入的不是 QianWenOptions!");
}
return (QianWenChatCompletionRequest) QianWenChatCompletionRequest.builder()
.model(qianWenApi.getQianWenChatModal().getValue())
.apiKey(qianWenApi.getApiKey())
.messages(prompt.getInstructions().stream().map(m -> {
Message message = new Message();
message.setRole(m.getMessageType().getValue());
message.setContent(m.getContent());
return message;
}).collect(Collectors.toList()))
.resultFormat(QwenParam.ResultFormat.MESSAGE)
// 动态改变的三个参数
.topP(Double.valueOf(qianWenOptions.getTopP()))
.topK(qianWenOptions.getTopK())
.temperature(qianWenOptions.getTemperature())
.incrementalOutput(true)
.build();
return (QianWenOptions) options;
}
@Override
public Flux<ChatResponse> stream(Prompt prompt) {
// ctx 会有重试的信息
// 创建 request 请求，stream模式需要供应商支持
QianWenChatCompletionRequest request = this.createRequest(prompt, true);
QwenParam request = this.createRequest(prompt, true);
// 调用 callWithFunctionSupport 发送请求
Flowable<GenerationResult> responseResult = this.qianWenApi.chatCompletionStream(request);
return Flux.create(fluxSink ->
responseResult.subscribe(
value -> fluxSink.next(new ChatResponse(List.of(new Generation(value.getOutput().getText())))),
value -> fluxSink.next(
new ChatResponse(value.getOutput().getChoices().stream()
.map(choices -> new Generation(choices.getMessage().getContent()))
.collect(Collectors.toList()))
),
error -> fluxSink.error(error),
() -> fluxSink.complete()
)

6
yudao-module-ai/yudao-spring-boot-starter-ai/src/main/java/cn/iocoder/yudao/framework/ai/chatqianwen/QianWenOptions.java
View File

@ -15,7 +15,7 @@ import java.util.List;
* time: 2024/3/15 19:57
*/
@Data
@Accessors
@Accessors(chain = true)
public class QianWenOptions implements ChatOptions {
/**
@ -28,6 +28,10 @@ public class QianWenOptions implements ChatOptions {
* 默认值为0.8。注意，取值不要大于等于1
*/
private Float topP;
/**
* 用于限制模型生成token的数量，max_tokens设置的是生成上限，并不表示一定会生成这么多的token数量。其中qwen1.5-14b-chat、qwen1.5-7b-chat、qwen-14b-chat和qwen-7b-chat最大值和默认值均为1500，qwen-1.8b-chat、qwen-1.8b-longcontext-chat和qwen-72b-chat最大值和默认值均为2000
*/
private Integer maxTokens = 1500;
//
// 适配 ChatOptions

7
yudao-module-ai/yudao-spring-boot-starter-ai/src/main/java/cn/iocoder/yudao/framework/ai/chatqianwen/api/QianWenApi.java
View File

@ -4,6 +4,7 @@ import cn.iocoder.yudao.framework.ai.chatqianwen.QianWenChatModal;
import cn.iocoder.yudao.framework.ai.exception.AiException;
import com.alibaba.dashscope.aigc.generation.Generation;
import com.alibaba.dashscope.aigc.generation.GenerationResult;
import com.alibaba.dashscope.aigc.generation.models.QwenParam;
import com.alibaba.dashscope.common.Message;
import com.alibaba.dashscope.common.Role;
import com.alibaba.dashscope.exception.InputRequiredException;
@ -34,9 +35,7 @@ public class QianWenApi {
this.qianWenChatModal = qianWenChatModal;
}
public ResponseEntity<GenerationResult> chatCompletionEntity(QianWenChatCompletionRequest request) {
Message userMsg = Message.builder().role(Role.USER.getValue()).content("用萝卜、土豆、茄子做饭，给我个菜谱").build();
public ResponseEntity<GenerationResult> chatCompletionEntity(QwenParam request) {
GenerationResult call;
try {
call = gen.call(request);
@ -49,7 +48,7 @@ public class QianWenApi {
return new ResponseEntity<>(call, HttpStatusCode.valueOf(200));
}
public Flowable<GenerationResult> chatCompletionStream(QianWenChatCompletionRequest request) {
public Flowable<GenerationResult> chatCompletionStream(QwenParam request) {
Flowable<GenerationResult> resultFlowable;
try {
resultFlowable = gen.streamCall(request);

1
yudao-module-ai/yudao-spring-boot-starter-ai/src/main/java/cn/iocoder/yudao/framework/ai/config/YudaoAiAutoConfiguration.java
View File

@ -47,6 +47,7 @@ public class YudaoAiAutoConfiguration {
QianWenOptions qianWenOptions = new QianWenOptions();
qianWenOptions.setTopK(qianWenProperties.getTopK());
qianWenOptions.setTopP(qianWenProperties.getTopP());
qianWenOptions.setMaxTokens(qianWenProperties.getMaxTokens());
qianWenOptions.setTemperature(qianWenProperties.getTemperature());
return new QianWenChatClient(
new QianWenApi(

4
yudao-module-ai/yudao-spring-boot-starter-ai/src/main/java/cn/iocoder/yudao/framework/ai/config/YudaoAiProperties.java
View File

@ -47,6 +47,10 @@ public class YudaoAiProperties {
* api key
*/
private String apiKey;
/**
* 用于限制模型生成token的数量，max_tokens设置的是生成上限，并不表示一定会生成这么多的token数量
*/
private Integer maxTokens;
/**
* model
*/