From b51e14a5709a7433bc3ca4a2dab06b0e2a64156f Mon Sep 17 00:00:00 2001
From: Administrator <15274802129@163.com>
Date: Tue, 02 Sep 2025 13:54:39 +0800
Subject: [PATCH] refactor(ai): 优化 ApiMemberTalkStreamServiceImpl 中的代码逻辑- 移除了无用的代码行 - 调整了变量赋值的顺序，提高了代码可读性
---
src/main/java/cc/mrbird/febs/ai/service/impl/AiServiceImpl.java | 434 ++++++++++++++++++++++++++++++++++++++++++++++++------
1 files changed, 385 insertions(+), 49 deletions(-)
diff --git a/src/main/java/cc/mrbird/febs/ai/service/impl/AiServiceImpl.java b/src/main/java/cc/mrbird/febs/ai/service/impl/AiServiceImpl.java
index 3ab9f2c..11c6794 100644
--- a/src/main/java/cc/mrbird/febs/ai/service/impl/AiServiceImpl.java
+++ b/src/main/java/cc/mrbird/febs/ai/service/impl/AiServiceImpl.java
@@ -1,11 +1,25 @@
package cc.mrbird.febs.ai.service.impl;
+import cc.mrbird.febs.ai.entity.AiTalkItem;
+import cc.mrbird.febs.ai.enumerates.AiTypeEnum;
import cc.mrbird.febs.ai.entity.AiProductRole;
+import cc.mrbird.febs.ai.req.ai.AiMessage;
import cc.mrbird.febs.ai.req.ai.AiRequest;
+import cc.mrbird.febs.ai.req.talk.AiTalkAnswerStream;
import cc.mrbird.febs.ai.res.ai.AiResponse;
+import cc.mrbird.febs.ai.res.ai.RadarDataItem;
import cc.mrbird.febs.ai.res.ai.Report;
+import cc.mrbird.febs.ai.res.memberTalk.ApiMemberTalkStreamVo;
import cc.mrbird.febs.ai.service.AiProductRoleService;
import cc.mrbird.febs.ai.service.AiService;
+import cc.mrbird.febs.ai.service.AiTalkItemService;
+import cc.mrbird.febs.ai.strategy.enumerates.LlmStrategyContextEnum;
+import cc.mrbird.febs.common.entity.FebsResponse;
+import cc.mrbird.febs.mall.entity.DataDictionaryCustom;
+import cc.mrbird.febs.mall.mapper.DataDictionaryCustomMapper;
+import cn.hutool.core.collection.CollUtil;
+import cn.hutool.core.util.StrUtil;
+import cn.hutool.json.JSONUtil;
import com.fasterxml.jackson.core.JsonProcessingException;
import com.fasterxml.jackson.databind.JsonNode;
import com.fasterxml.jackson.databind.ObjectMapper;
@@ -18,12 +32,15 @@
import org.springframework.beans.factory.annotation.Value;
import org.springframework.stereotype.Service;
import org.springframework.util.StringUtils;
+import reactor.core.publisher.Flux;
import javax.annotation.PostConstruct;
import javax.annotation.PreDestroy;
import java.util.ArrayList;
+import java.util.Arrays;
import java.util.List;
import java.util.concurrent.TimeUnit;
+import java.util.function.Consumer;
import java.util.regex.Matcher;
import java.util.regex.Pattern;
import java.util.stream.Collectors;
@@ -37,11 +54,30 @@
public class AiServiceImpl implements AiService {
private static final String CODE_SUCCESS = "200";
+ private static final String CODE_GOING_ON = "199";
private static final String CODE_NOT_FOUND = "201";
private static final String CODE_ERROR = "500";
+ private static final String SCHEMA_JSON = "{\n" +
+ " \"radar_data\": {\n" +
+ " \"problem_understanding\": \"object\",\n" +
+ " \"fluency\": \"object\",\n" +
+ " \"principle_adherence\": \"object\",\n" +
+ " \"logicality\": \"object\",\n" +
+ " \"knowledge_mastery\": \"object\"\n" +
+ " },\n" +
+ " \"evaluation\": {\n" +
+ " \"highlight\": \"object\",\n" +
+ " \"suggestion\": \"object\",\n" +
+ " \"reference_answer\": \"object\",\n" +
+ " \"key_knowledge\": \"object\"\n" +
+ " }\n" +
+ " }";
+
private final AiProductRoleService aiProductRoleService;
private final ObjectMapper objectMapper;
+ private final AiTalkItemService aiTalkItemService;
+ private final DataDictionaryCustomMapper dataDictionaryCustomMapper;
@Value("${ai.service.ak}")
private String ak;
@@ -56,8 +92,13 @@
@PostConstruct
public void init() {
- ConnectionPool connectionPool = new ConnectionPool(10, 30, TimeUnit.SECONDS);
+ // 增加连接池大小和存活时间
+ ConnectionPool connectionPool = new ConnectionPool(32, 60, TimeUnit.SECONDS);
Dispatcher dispatcher = new Dispatcher();
+ // 增加并发请求数量
+ dispatcher.setMaxRequests(128);
+ dispatcher.setMaxRequestsPerHost(32);
+
this.service = ArkService.builder()
.dispatcher(dispatcher)
.connectionPool(connectionPool)
@@ -75,7 +116,20 @@
}
@Override
- public AiResponse start(String productRoleId, String content) {
+ public Integer getSystemSetAiType() {
+ Integer type = 2;
+ DataDictionaryCustom dataDictionaryCustom = dataDictionaryCustomMapper.selectDicDataByTypeAndCode(
+ LlmStrategyContextEnum.LLM_STRATEGY.getCode(),
+ LlmStrategyContextEnum.LLM_STRATEGY.getCode()
+ );
+ if (dataDictionaryCustom != null) {
+ type = Integer.parseInt(dataDictionaryCustom.getValue());
+ }
+ return type;
+ }
+
+ @Override
+ public AiResponse start(List<AiMessage> aiMessageDtoList,Integer type,String productRoleId, String content, String question) {
if (!StringUtils.hasText(productRoleId)) {
log.warn("productRoleId 不能为空");
return buildErrorResponse(CODE_NOT_FOUND, "AI陪练不存在");
@@ -87,20 +141,36 @@
return buildErrorResponse(CODE_NOT_FOUND, "AI陪练不存在");
}
- String promptTemplate = aiProductRole.getPromptTemplate();
- String linkId = aiProductRole.getLinkId();
- if (!StringUtils.hasText(promptTemplate) || !StringUtils.hasText(linkId)) {
- log.warn("角色配置不完整，promptTemplate 或 linkId 为空，productRoleId: {}", productRoleId);
+ String promptTemplate = "作为一个智能助手，请回答我提出的问题。";
+ if (AiTypeEnum.QUESTION.getCode() == type){
+ promptTemplate = aiProductRole.getPromptHead();
+ }
+ if (AiTypeEnum.ANSWER.getCode() == type){
+ promptTemplate = aiProductRole.getPromptTemplate()+question;
+ }
+ log.info("promptTemplate: {}", promptTemplate);
+ String linkId = aiProductRole.getLinkId();
+ String jsonTemplate = aiProductRole.getJsonTemplate();
+
+ if (
+ !StringUtils.hasText(promptTemplate)
+ || !StringUtils.hasText(linkId)
+ || !StringUtils.hasText(jsonTemplate)
+ ) {
+ log.warn("角色配置不完整，promptTemplate 或 linkId 或 jsonTemplate为空，productRoleId: {}", productRoleId);
return buildErrorResponse(CODE_ERROR, "角色配置不完整");
}
AiRequest aiRequest = new AiRequest();
aiRequest.setPromptTemplate(promptTemplate);
+ aiRequest.setJsonTemplate(jsonTemplate);
aiRequest.setLinkId(linkId);
aiRequest.setContent(content);
-
- return question(aiRequest);
+ if (CollUtil.isNotEmpty(aiMessageDtoList)){
+ aiRequest.setAiMessageDtoList(aiMessageDtoList);
+ }
+ return this.question(aiRequest);
}
@Override
@@ -108,36 +178,34 @@
String promptTemplate = aiRequest.getPromptTemplate();
String linkId = aiRequest.getLinkId();
String content = aiRequest.getContent();
- if (!StringUtils.hasText(promptTemplate) || !StringUtils.hasText(linkId) || !StringUtils.hasText(content)) {
- log.warn("请求参数不完整，promptTemplate: {}, linkId: {}, content: {}", promptTemplate, linkId, content);
+ String jsonTemplate = aiRequest.getJsonTemplate();
+ if (
+ !StringUtils.hasText(promptTemplate)
+ || !StringUtils.hasText(linkId)
+ || !StringUtils.hasText(content)
+ || !StringUtils.hasText(jsonTemplate)
+ ) {
+ log.warn("请求参数不完整，promptTemplate: {}, linkId: {}, content: {}, jsonTemplate: {}", promptTemplate, linkId, content, jsonTemplate);
return buildErrorResponse(CODE_ERROR, "请求参数不完整");
}
- final List<ChatMessage> messages = new ArrayList<>();
- final ChatMessage systemMessage = ChatMessage.builder().role(ChatMessageRole.SYSTEM).content(promptTemplate).build();
- final ChatMessage userMessage = ChatMessage.builder().role(ChatMessageRole.USER).content(content).build();
+ List<ChatMessage> messages = new ArrayList<>();
+ ChatMessage systemMessage = ChatMessage.builder().role(ChatMessageRole.SYSTEM).content(promptTemplate).build();
+ ChatMessage userMessage = ChatMessage.builder().role(ChatMessageRole.USER).content(content).build();
messages.add(systemMessage);
+ if (CollUtil.isNotEmpty(aiRequest.getAiMessageDtoList())){
+ aiRequest.getAiMessageDtoList().forEach(aiMessageDto -> {
+ ChatMessage message = ChatMessage.builder()
+ .role(aiMessageDto.getRole())
+ .content(aiMessageDto.getContent())
+ .build();
+ messages.add(message);
+ });
+ }
messages.add(userMessage);
- // 生成 JSON Schema
- String schemaJson = "{\n" +
- " \"radar_data\": {\n" +
- " \"problem_understanding\": \"object\",\n" +
- " \"fluency\": \"object\",\n" +
- " \"principle_adherence\": \"object\",\n" +
- " \"logicality\": \"object\",\n" +
- " \"knowledge_mastery\": \"object\"\n" +
- " },\n" +
- " \"evaluation\": {\n" +
- " \"highlight\": \"object\",\n" +
- " \"suggestion\": \"object\",\n" +
- " \"reference_answer\": \"object\",\n" +
- " \"key_knowledge\": \"object\"\n" +
- " }\n" +
- " }";
try {
- JsonNode schemaNode = objectMapper.readTree(schemaJson);
- // 配置响应格式
+ JsonNode schemaNode = objectMapper.readTree(jsonTemplate);
ChatCompletionRequest.ChatCompletionRequestResponseFormat responseFormat = new ChatCompletionRequest.ChatCompletionRequestResponseFormat(
"json_schema",
new ResponseFormatJSONSchemaJSONSchemaParam(
@@ -152,28 +220,136 @@
.messages(messages)
.stream(false)
.responseFormat(responseFormat)
- .temperature(1.0)
- .topP(0.7)
- .maxTokens(4096)
+ .temperature(0.7) // 降低温度参数，提高确定性，可能提升速度
+ .topP(0.9) // 调整topP参数
+ .maxTokens(2048) // 减少最大token数
.frequencyPenalty(0.0)
.build();
+
List<ChatCompletionChoice> choices = service.createChatCompletion(chatCompletionRequest).getChoices();
String result = choices.stream()
.map(choice -> choice.getMessage().getContent())
.filter(contentObj -> contentObj != null)
.map(Object::toString)
.collect(Collectors.joining());
-
Report report = this.extractReportData(result);
return buildSuccessResponse(report, result);
} catch (JsonProcessingException e) {
log.error("初始化AI服务失败，JSON格式化输出初始化失败", e);
return buildErrorResponse(CODE_ERROR, "AI服务调用失败");
} catch (Exception e) {
- log.error("调用AI服务失败，modelId: {}, content: {}", linkId, content, e);
+ log.error("调用AI服务失败，modelId: {}", linkId, e);
return buildErrorResponse(CODE_ERROR, "AI服务调用失败");
}
}
+
+ public static void main(String[] args) {
+ Report report = new Report();
+ List<RadarDataItem> radarDataItems = new ArrayList<>();
+
+ RadarDataItem item1 = new RadarDataItem();
+ item1.setName("A");
+ item1.setCode("A");
+ item1.setScore("80");
+ radarDataItems.add(item1);
+
+ RadarDataItem item2 = new RadarDataItem();
+ item2.setName("A");
+ item2.setCode("A");
+ item2.setScore("80");
+ radarDataItems.add(item2);
+ report.setRadarDataItems(radarDataItems);
+
+ System.out.println(JSONUtil.parse( report));
+
+ }
+
+ @Override
+ public void streamQuestion(AiRequest aiRequest, Consumer<AiResponse> callback) {
+
+ String promptTemplate = aiRequest.getPromptTemplate();
+ String linkId = aiRequest.getLinkId();
+ String content = aiRequest.getContent();
+ if (!StringUtils.hasText(promptTemplate) || !StringUtils.hasText(linkId) || !StringUtils.hasText(content)) {
+ log.warn("请求参数不完整，promptTemplate: {}, linkId: {}, content: {}", promptTemplate, linkId, content);
+ }
+
+ final List<ChatMessage> messages = new ArrayList<>();
+ final ChatMessage systemMessage = ChatMessage.builder().role(ChatMessageRole.SYSTEM).content(promptTemplate).build();
+ final ChatMessage userMessage = ChatMessage.builder().role(ChatMessageRole.USER).content(content).build();
+ messages.add(systemMessage);
+ messages.add(userMessage);
+
+ try {
+ JsonNode schemaNode = objectMapper.readTree(SCHEMA_JSON);
+ ChatCompletionRequest.ChatCompletionRequestResponseFormat responseFormat = new ChatCompletionRequest.ChatCompletionRequestResponseFormat(
+ "json_schema",
+ new ResponseFormatJSONSchemaJSONSchemaParam(
+ "ai_response",
+ "json数据响应",
+ schemaNode,
+ true
+ )
+ );
+ ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder()
+ .model(linkId)
+ .messages(messages)
+ .stream(true) // 启用流式响应
+ .responseFormat(responseFormat)
+ .temperature(0.7)
+ .topP(0.9)
+ .maxTokens(2048)
+ .build();
+
+ service.streamChatCompletion(chatCompletionRequest)
+ .doOnError(Throwable::printStackTrace) // 处理错误
+ .blockingForEach(response -> {
+ AiResponse partialResponse = new AiResponse();
+ if (response.getChoices() != null && !response.getChoices().isEmpty()) {
+ String responseStr = String.valueOf(response.getChoices().get(0).getMessage().getContent());
+ if (responseStr != null) {
+ // 构造部分响应并回调
+ partialResponse = buildGOINGONResponse(responseStr);
+ }
+ }else{
+ partialResponse = buildPartialResponse("成功");
+ }
+ callback.accept(partialResponse);
+ });
+// service.streamChatCompletion(chatCompletionRequest)
+// .doOnError(throwable -> {
+// log.error("流式调用AI服务失败", throwable);
+// callback.accept(buildErrorResponse(CODE_ERROR, "AI服务调用失败"));
+// })
+// .subscribe(chatCompletionChunk -> {
+// // 处理每个数据块
+// Object chunkContent = chatCompletionChunk.getChoices().get(0).getMessage().getContent();
+// // 构造部分响应并回调
+// AiResponse partialResponse = buildGOINGONResponse(chunkContent);
+// callback.accept(partialResponse);
+// });
+ } catch (Exception e) {
+ log.error("调用AI服务失败", e);
+ callback.accept(buildErrorResponse(CODE_ERROR, "AI服务调用失败"));
+ }
+ }
+
+ private AiResponse buildGOINGONResponse(Object chunkContent) {
+ AiResponse response = new AiResponse();
+ response.setCode(CODE_GOING_ON);
+ response.setDescription("成功");
+ response.setResContext(chunkContent.toString());
+ return response;
+ }
+
+ private AiResponse buildPartialResponse(Object chunkContent) {
+ AiResponse response = new AiResponse();
+ response.setCode(CODE_SUCCESS);
+ response.setDescription("成功");
+ response.setResContext(chunkContent.toString());
+ return response;
+ }
+
private static final Pattern JSON_PATTERN = Pattern.compile(
"<\\|FunctionCallBegin\\|>(.*?)<\\|FunctionCallEnd\\|>",
@@ -182,22 +358,19 @@
@Override
public Report extractReportData(String modelOutput) {
- // 提取JSON部分
Matcher matcher = JSON_PATTERN.matcher(modelOutput);
if (!matcher.find()) {
- log.warn("未匹配到FunctionCall内容，原始输出: {}", modelOutput);
+ log.warn("未匹配到FunctionCall内容，原始输出长度: {}", modelOutput);
return null;
}
String jsonContent = matcher.group(1);
- log.debug("提取到的JSON内容: {}", jsonContent);
+ log.debug("提取到的JSON内容长度: {}", jsonContent.length());
- // 解析JSON到Report对象
try {
return objectMapper.readValue(jsonContent, Report.class);
} catch (JsonProcessingException e) {
- log.error("JSON解析失败，原始内容: {}", jsonContent, e);
- // 尝试修复截断的JSON（可选）
+ log.error("JSON解析失败，原始内容长度: {}", jsonContent.length(), e);
Report repairedReport = tryRepairTruncatedJson(jsonContent);
if (repairedReport != null) {
log.info("成功修复截断的JSON");
@@ -207,13 +380,176 @@
}
}
- /**
- * 尝试修复截断的JSON字符串
- * @param truncatedJson 可能被截断的JSON字符串
- * @return 修复后的Report对象，如果无法修复则返回null
- */
+ // 修改服务实现
+ @Override
+ public Flux<FebsResponse> answerStream(String question) {
+ log.info("----- standard request -----");
+
+ final ChatMessage systemMessage = ChatMessage.builder()
+ .role(ChatMessageRole.SYSTEM)
+ .content("你是豆包，是由字节跳动开发的 AI 人工智能助手")
+ .build();
+
+ final ChatMessage userMessage = ChatMessage.builder()
+ .role(ChatMessageRole.USER)
+ .content(question)
+ .build();
+
+ List<ChatMessage> messages = Arrays.asList(systemMessage, userMessage);
+
+ ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder()
+ .model("ep-20250805124033-lhxbf")
+ .messages(messages)
+ .stream(true)
+ .thinking(new ChatCompletionRequest.ChatCompletionRequestThinking("enabled"))
+ .temperature(0.7)
+ .topP(0.9)
+ .maxTokens(2048)
+ .frequencyPenalty(0.0)
+ .build();
+
+ return Flux.from(service.streamChatCompletion(chatCompletionRequest))
+ .map(response -> {
+ if (response == null || response.getChoices() == null || response.getChoices().isEmpty()) {
+ return new FebsResponse().success().data("END");
+ }
+
+ ChatCompletionChoice choice = response.getChoices().get(0);
+ if (choice == null || choice.getMessage() == null) {
+ return new FebsResponse().success().data("END");
+ }
+
+
+ ApiMemberTalkStreamVo apiMemberTalkStreamVo = new ApiMemberTalkStreamVo();
+ // 判断是否触发深度思考，触发则打印模型输出的思维链内容
+ ChatMessage message = choice.getMessage();
+ if (message.getReasoningContent()!= null &&!message.getReasoningContent().isEmpty()) {
+ apiMemberTalkStreamVo.setReasoningContent(message.getReasoningContent());
+// System.out.print(message.getReasoningContent());
+ }
+
+ String content = message.getContent() == null ? "" : message.getContent().toString();
+ apiMemberTalkStreamVo.setContent(content);
+ System.out.print(content);
+ return new FebsResponse().success().data(apiMemberTalkStreamVo);
+ })
+ .onErrorResume(throwable -> {
+ log.error("流式调用AI服务失败，问题输入: {}", question, throwable);
+ FebsResponse errorResponse = new FebsResponse().fail().message("AI服务调用失败");
+ return Flux.just(errorResponse);
+ });
+ }
+
+ @Override
+ public Flux<FebsResponse> answerStreamV2(AiTalkAnswerStream dto) {
+ String question = dto.getQuestion();
+ log.info("----- standard request -----");
+
+ // 参数校验
+ if (StrUtil.isBlank(question)) {
+ return Flux.just(new FebsResponse().fail().message("问题不能为空"));
+ }
+
+ List<ChatMessage> messages = new ArrayList<>();
+
+ final ChatMessage systemMessage = ChatMessage.builder()
+ .role(ChatMessageRole.SYSTEM)
+ .content("你是豆包，是由字节跳动开发的 AI 人工智能助手")
+ .build();
+ messages.add(systemMessage);
+
+ // 获取历史消息记录
+ if (StrUtil.isNotEmpty(dto.getTalkId())) {
+ List<AiTalkItem> aiTalkItems = aiTalkItemService.getListByTalkId(dto.getTalkId());
+ if (CollUtil.isNotEmpty(aiTalkItems)) {
+ for (AiTalkItem aiTalkItem : aiTalkItems) {
+ ChatMessage chatMessage = buildChatMessageFromItem(aiTalkItem);
+ if (chatMessage != null) {
+ messages.add(chatMessage);
+ }
+ }
+ }
+ }
+
+ final ChatMessage userMessage = ChatMessage.builder()
+ .role(ChatMessageRole.USER)
+ .content(question)
+ .build();
+ messages.add(userMessage);
+
+ ChatCompletionRequest chatCompletionRequest = ChatCompletionRequest.builder()
+ .model("ep-20250805124033-lhxbf")
+ .messages(messages)
+ .stream(true)
+ .thinking(new ChatCompletionRequest.ChatCompletionRequestThinking("enabled"))
+ .temperature(0.7)
+ .topP(0.9)
+ .maxTokens(2048)
+ .frequencyPenalty(0.0)
+ .build();
+
+ return Flux.from(service.streamChatCompletion(chatCompletionRequest))
+ .map(response -> {
+ if (response == null || response.getChoices() == null || response.getChoices().isEmpty()) {
+ return new FebsResponse().success().data("END");
+ }
+
+ ChatCompletionChoice choice = response.getChoices().get(0);
+ if (choice == null || choice.getMessage() == null) {
+ return new FebsResponse().success().data("END");
+ }
+
+ ChatMessage message = choice.getMessage();
+ ApiMemberTalkStreamVo apiMemberTalkStreamVo = new ApiMemberTalkStreamVo();
+
+ // 处理 reasoning content
+ String reasoningContent = message.getReasoningContent();
+ if (StrUtil.isNotEmpty(reasoningContent)) {
+ apiMemberTalkStreamVo.setReasoningContent(reasoningContent);
+ log.debug("Reasoning Content: {}", reasoningContent);
+ }
+
+ // 安全处理 content
+ String content = "";
+ if (message.getContent() != null) {
+ content = message.getContent().toString();
+ }
+ apiMemberTalkStreamVo.setContent(content);
+ System.out.print(content);
+ log.debug("Content: {}", content);
+
+ return new FebsResponse().success().data(apiMemberTalkStreamVo);
+ })
+ .onErrorResume(throwable -> {
+ log.error("流式调用AI服务失败，问题输入: {}", question, throwable);
+ FebsResponse errorResponse = new FebsResponse().fail().message("AI服务调用失败");
+ return Flux.just(errorResponse);
+ });
+ }
+
+ // 提取为私有方法，提高可读性和复用性
+ private ChatMessage buildChatMessageFromItem(AiTalkItem item) {
+ if (item == null) return null;
+
+ switch (item.getType()) {
+ case 1:
+ return ChatMessage.builder()
+ .role(ChatMessageRole.USER)
+ .content(item.getContext())
+ .build();
+ case 2:
+ return ChatMessage.builder()
+ .role(ChatMessageRole.ASSISTANT)
+ .content(item.getContext())
+ .build();
+ default:
+ return null;
+ }
+ }
+
+
+
private Report tryRepairTruncatedJson(String truncatedJson) {
- // 简单的修复策略：尝试添加缺失的结束括号
String[] repairAttempts = {
truncatedJson + "\"}}}",
truncatedJson + "}}}",
@@ -229,7 +565,7 @@
}
}
- log.warn("无法修复截断的JSON: {}", truncatedJson);
+ log.warn("无法修复截断的JSON，原始内容长度: {}", truncatedJson.length());
return null;
}
--
Gitblit v1.9.1