mirror of
https://gitcode.com/ageerle/ruoyi-ai.git
synced 2026-03-13 20:53:42 +08:00
Merge pull request #196 from zhangyue-mars/add-deepseek-java-files
feat: update ChatRequest and DeepSeekChatImpl for DeepSeek integration
This commit is contained in:
ageerle
GitHub
@@ -72,6 +72,11 @@ public class ChatRequest {
|
||||
*/
|
||||
private Boolean hasAttachment;
|
||||
|
||||
/**
|
||||
* 是否启用深度思考
|
||||
*/
|
||||
private Boolean enableThinking;
|
||||
|
||||
/**
|
||||
* 是否自动切换模型
|
||||
*/
|
||||
@@ -82,9 +87,4 @@ public class ChatRequest {
|
||||
*/
|
||||
private String token;
|
||||
|
||||
/**
|
||||
* 消息ID（保存消息成功后设置，用于后续扣费更新）
|
||||
*/
|
||||
private Long messageId;
|
||||
|
||||
}
|
||||
|
||||
@@ -1,21 +1,39 @@
|
||||
package org.ruoyi.chat.service.chat.impl;
|
||||
|
||||
|
||||
import dev.langchain4j.data.message.ChatMessage;
|
||||
import dev.langchain4j.data.message.SystemMessage;
|
||||
import dev.langchain4j.data.message.UserMessage;
|
||||
import dev.langchain4j.model.chat.StreamingChatModel;
|
||||
import dev.langchain4j.model.chat.response.ChatResponse;
|
||||
import dev.langchain4j.model.chat.response.StreamingChatResponseHandler;
|
||||
import dev.langchain4j.model.openai.OpenAiStreamingChatModel;
|
||||
import lombok.SneakyThrows;
|
||||
import lombok.extern.slf4j.Slf4j;
|
||||
import okhttp3.*;
|
||||
import okhttp3.Response;
|
||||
import org.ruoyi.chat.enums.ChatModeType;
|
||||
import org.ruoyi.chat.service.chat.IChatService;
|
||||
import org.ruoyi.chat.support.ChatServiceHelper;
|
||||
import org.ruoyi.common.chat.entity.chat.Message;
|
||||
import org.ruoyi.common.chat.request.ChatRequest;
|
||||
import org.ruoyi.domain.vo.ChatModelVo;
|
||||
import org.ruoyi.service.IChatModelService;
|
||||
import org.springframework.beans.factory.annotation.Autowired;
|
||||
import org.springframework.stereotype.Service;
|
||||
import org.springframework.web.servlet.mvc.method.annotation.SseEmitter;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.ArrayList;
|
||||
import java.util.HashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
import java.util.concurrent.TimeUnit;
|
||||
|
||||
import java.io.BufferedReader;
|
||||
import java.io.InputStreamReader;
|
||||
import com.fasterxml.jackson.databind.ObjectMapper;
|
||||
|
||||
/**
|
||||
* deepseek
|
||||
*/
|
||||
@@ -26,9 +44,24 @@ public class DeepSeekChatImpl implements IChatService {
|
||||
@Autowired
|
||||
private IChatModelService chatModelService;
|
||||
|
||||
private static final MediaType JSON = MediaType.get("application/json; charset=utf-8");
|
||||
|
||||
// 创建一个用于直接API调用的OkHttpClient
|
||||
private final OkHttpClient client = new OkHttpClient.Builder()
|
||||
.connectTimeout(30, TimeUnit.SECONDS)
|
||||
.readTimeout(60, TimeUnit.SECONDS)
|
||||
.writeTimeout(30, TimeUnit.SECONDS)
|
||||
.build();
|
||||
|
||||
@Override
|
||||
public SseEmitter chat(ChatRequest chatRequest, SseEmitter emitter) {
|
||||
ChatModelVo chatModelVo = chatModelService.selectModelByName(chatRequest.getModel());
|
||||
|
||||
// 检查是否启用深度思考且是deepseek模型
|
||||
if (Boolean.TRUE.equals(chatRequest.getEnableThinking())) {
|
||||
return handleDeepSeekWithThinking(chatRequest, emitter, chatModelVo);
|
||||
}
|
||||
|
||||
StreamingChatModel chatModel = OpenAiStreamingChatModel.builder()
|
||||
.baseUrl(chatModelVo.getApiHost())
|
||||
.apiKey(chatModelVo.getApiKey())
|
||||
@@ -39,7 +72,31 @@ public class DeepSeekChatImpl implements IChatService {
|
||||
.build();
|
||||
// 发送流式消息
|
||||
try {
|
||||
chatModel.chat(chatRequest.getPrompt(), new StreamingChatResponseHandler() {
|
||||
// 构建消息列表，包含历史对话消息和当前用户消息
|
||||
List<ChatMessage> messages = new ArrayList<>();
|
||||
|
||||
// 添加历史对话消息
|
||||
if (chatRequest.getMessages() != null) {
|
||||
for (Message message : chatRequest.getMessages()) {
|
||||
// 检查消息内容是否有效
|
||||
if (message.getContent() == null || String.valueOf(message.getContent()).trim().isEmpty()) {
|
||||
continue; // 跳过空消息
|
||||
}
|
||||
|
||||
if (Message.Role.SYSTEM.getName().equals(message.getRole())) {
|
||||
messages.add(new SystemMessage(String.valueOf(message.getContent())));
|
||||
} else if (Message.Role.USER.getName().equals(message.getRole())) {
|
||||
messages.add(new UserMessage(String.valueOf(message.getContent())));
|
||||
} else if (Message.Role.ASSISTANT.getName().equals(message.getRole())) {
|
||||
messages.add(new dev.langchain4j.data.message.AiMessage(String.valueOf(message.getContent())));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// 添加当前用户消息
|
||||
messages.add(new UserMessage(chatRequest.getPrompt()));
|
||||
|
||||
chatModel.chat(messages, new StreamingChatResponseHandler() {
|
||||
@SneakyThrows
|
||||
@Override
|
||||
public void onPartialResponse(String partialResponse) {
|
||||
@@ -70,8 +127,208 @@ public class DeepSeekChatImpl implements IChatService {
|
||||
return emitter;
|
||||
}
|
||||
|
||||
/**
|
||||
* 处理启用深度思考的deepseek模型请求
|
||||
*/
|
||||
private SseEmitter handleDeepSeekWithThinking(ChatRequest chatRequest, SseEmitter emitter, ChatModelVo chatModelVo) {
|
||||
try {
|
||||
// 构建请求到外部API
|
||||
String url = chatModelVo.getApiHost() + "/v1/chat/completions";
|
||||
String apiKey = chatModelVo.getApiKey();
|
||||
|
||||
// 构建请求体
|
||||
Map<String, Object> requestBody = new HashMap<>();
|
||||
requestBody.put("model", chatModelVo.getModelName());
|
||||
requestBody.put("response_format", Map.of("type", "text"));
|
||||
requestBody.put("max_tokens", 4000); // 修复：将max_tokens从81920改为4000，符合API要求
|
||||
requestBody.put("temperature", 1);
|
||||
requestBody.put("top_p", 1);
|
||||
requestBody.put("top_k", 50);
|
||||
requestBody.put("enable_thinking", chatRequest.getEnableThinking());
|
||||
requestBody.put("stream", chatRequest.getStream());
|
||||
|
||||
// 构建消息 - DeepSeek模型不需要系统提示词
|
||||
List<Map<String, Object>> messages = new ArrayList<>();
|
||||
|
||||
// 添加历史对话消息 (只添加用户和助手消息)
|
||||
if (chatRequest.getMessages() != null) {
|
||||
for (Message message : chatRequest.getMessages()) {
|
||||
// 检查消息内容是否有效
|
||||
if (message.getContent() == null || String.valueOf(message.getContent()).trim().isEmpty()) {
|
||||
continue; // 跳过空消息
|
||||
}
|
||||
|
||||
// DeepSeek模型在深度思考模式下只接受user和assistant角色的消息
|
||||
if (Message.Role.SYSTEM.getName().equals(message.getRole())) {
|
||||
// 跳过系统消息
|
||||
continue;
|
||||
}
|
||||
|
||||
Map<String, Object> historyMessage = new HashMap<>();
|
||||
historyMessage.put("role", message.getRole());
|
||||
historyMessage.put("content", String.valueOf(message.getContent()));
|
||||
messages.add(historyMessage);
|
||||
}
|
||||
}
|
||||
|
||||
// 添加当前用户消息
|
||||
Map<String, Object> userMessage = new HashMap<>();
|
||||
userMessage.put("role", "user");
|
||||
userMessage.put("content", String.valueOf(chatRequest.getPrompt()));
|
||||
messages.add(userMessage);
|
||||
|
||||
requestBody.put("messages", messages);
|
||||
|
||||
// 创建ObjectMapper实例
|
||||
com.fasterxml.jackson.databind.ObjectMapper objectMapper = new com.fasterxml.jackson.databind.ObjectMapper();
|
||||
String requestBodyStr = objectMapper.writeValueAsString(requestBody);
|
||||
|
||||
// 打印请求体用于调试
|
||||
log.info("打印请求体: {}", requestBodyStr);
|
||||
|
||||
// 创建请求
|
||||
Request request = new Request.Builder()
|
||||
.url(url)
|
||||
.header("Authorization", "Bearer " + apiKey)
|
||||
.header("Content-Type", "application/json")
|
||||
.post(RequestBody.create(requestBodyStr, JSON))
|
||||
.build();
|
||||
|
||||
// 执行异步请求
|
||||
this.client.newCall(request).enqueue(new Callback() {
|
||||
@Override
|
||||
public void onFailure(Call call, IOException e) {
|
||||
try {
|
||||
log.error("深度思考请求失败: {}", e.getMessage(), e);
|
||||
emitter.send("深度思考请求失败: " + e.getMessage());
|
||||
emitter.complete();
|
||||
} catch (IOException ioException) {
|
||||
log.error("发送错误消息失败: {}", ioException.getMessage(), ioException);
|
||||
}
|
||||
}
|
||||
|
||||
@Override
|
||||
public void onResponse(Call call, Response response) throws IOException {
|
||||
if (!response.isSuccessful()) {
|
||||
// 打印完整的错误响应体
|
||||
String errorBody = "";
|
||||
if (response.body() != null) {
|
||||
errorBody = response.body().string();
|
||||
}
|
||||
log.error("深度思考请求失败，状态码: {}，响应体: {}", response.code(), errorBody);
|
||||
try {
|
||||
emitter.send("深度思考请求失败，状态码: " + response.code() + "，响应体: " + errorBody);
|
||||
emitter.complete();
|
||||
return;
|
||||
} catch (IOException e) {
|
||||
log.error("发送错误消息失败: {}", e.getMessage(), e);
|
||||
return;
|
||||
}
|
||||
}
|
||||
|
||||
try (ResponseBody responseBody = response.body()) {
|
||||
if (responseBody == null) {
|
||||
log.error("响应体为空");
|
||||
emitter.send("响应体为空");
|
||||
emitter.complete();
|
||||
return;
|
||||
}
|
||||
|
||||
// 流式读取响应
|
||||
processThinkingResponse(responseBody, emitter);
|
||||
} catch (Exception e) {
|
||||
log.error("处理响应时出错: {}", e.getMessage(), e);
|
||||
try {
|
||||
emitter.send("处理响应时出错: " + e.getMessage());
|
||||
emitter.complete();
|
||||
} catch (IOException ioException) {
|
||||
log.error("发送错误消息失败: {}", ioException.getMessage(), ioException);
|
||||
}
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
} catch (Exception e) {
|
||||
log.error("处理深度思考请求时出错: {}", e.getMessage(), e);
|
||||
ChatServiceHelper.onStreamError(emitter, e.getMessage());
|
||||
}
|
||||
return emitter;
|
||||
}
|
||||
|
||||
/**
|
||||
* 处理深度思考的流式响应（边解析边推送）
|
||||
*/
|
||||
private void processThinkingResponse(ResponseBody responseBody, SseEmitter emitter) throws IOException {
|
||||
// 标记是否进入正式回答阶段
|
||||
boolean thinkingComplete = false;
|
||||
|
||||
try (BufferedReader reader = new BufferedReader(responseBody.charStream())) {
|
||||
String line;
|
||||
while ((line = reader.readLine()) != null) {
|
||||
if (!line.startsWith("data: ")) {
|
||||
continue;
|
||||
}
|
||||
|
||||
String jsonData = line.substring(6).trim();
|
||||
if ("[DONE]".equals(jsonData)) {
|
||||
break;
|
||||
}
|
||||
|
||||
try {
|
||||
ObjectMapper mapper = new ObjectMapper();
|
||||
Map<String, Object> chunk = mapper.readValue(jsonData, Map.class);
|
||||
|
||||
if (chunk.containsKey("choices") && chunk.get("choices") instanceof List) {
|
||||
List<Map<String, Object>> choices = (List<Map<String, Object>>) chunk.get("choices");
|
||||
if (!choices.isEmpty()) {
|
||||
Map<String, Object> choice = choices.get(0);
|
||||
if (choice.containsKey("delta") && choice.get("delta") instanceof Map) {
|
||||
Map<String, Object> delta = (Map<String, Object>) choice.get("delta");
|
||||
|
||||
// 推送思考过程
|
||||
if (delta.containsKey("reasoning_content") && delta.get("reasoning_content") != null) {
|
||||
String reasoningChunk = delta.get("reasoning_content").toString();
|
||||
emitter.send(SseEmitter.event().data(reasoningChunk).name("thinking"));
|
||||
log.debug("Reasoning Chunk: {}", reasoningChunk);
|
||||
}
|
||||
|
||||
// 推送正式回答
|
||||
if (delta.containsKey("content") && delta.get("content") != null) {
|
||||
String content = delta.get("content").toString();
|
||||
|
||||
// 第一次进入回答阶段时，加个提示头
|
||||
if (!thinkingComplete) {
|
||||
emitter.send(SseEmitter.event().data("\n\n回答内容：\n").name("answer-header"));
|
||||
thinkingComplete = true;
|
||||
}
|
||||
|
||||
emitter.send(SseEmitter.event().data(content).name("answer"));
|
||||
log.debug("Answer Chunk:{}", content);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
} catch (Exception e) {
|
||||
log.warn("解析JSON数据失败，忽略本行: {}", jsonData, e);
|
||||
}
|
||||
}
|
||||
|
||||
emitter.complete();
|
||||
log.info("深度思考流式响应完成");
|
||||
} catch (IOException e) {
|
||||
log.error("读取响应流时出错: {}", e.getMessage(), e);
|
||||
try {
|
||||
emitter.send(SseEmitter.event().data("读取响应流时出错: " + e.getMessage()).name("error"));
|
||||
emitter.complete();
|
||||
} catch (IOException ioException) {
|
||||
log.error("发送错误消息失败: {}", ioException.getMessage(), ioException);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
@Override
|
||||
public String getCategory() {
|
||||
return ChatModeType.DEEPSEEK.getCode();
|
||||
}
|
||||
}
|
||||
}
|
||||
@@ -91,45 +91,66 @@ public class KnowledgeInfoServiceImpl implements IKnowledgeInfoService {
|
||||
public TableDataInfo<KnowledgeInfoVo> queryPageListByRole(KnowledgeInfoBo bo, PageQuery pageQuery) {
|
||||
// 查询用户关联角色
|
||||
LoginUser loginUser = LoginHelper.getLoginUser();
|
||||
if (StringUtils.isEmpty(loginUser.getKroleGroupIds()) || StringUtils.isEmpty(loginUser.getKroleGroupType())) {
|
||||
return new TableDataInfo<>();
|
||||
}
|
||||
|
||||
// 角色/角色组id列表
|
||||
List<String> groupIdList = Arrays.stream(loginUser.getKroleGroupIds().split(","))
|
||||
.filter(StringUtils::isNotEmpty)
|
||||
.toList();
|
||||
List<KnowledgeRole> knowledgeRoles;
|
||||
LambdaQueryWrapper<KnowledgeRole> roleLqw = Wrappers.lambdaQuery();
|
||||
if ("role".equals(loginUser.getKroleGroupType())) {
|
||||
roleLqw.in(KnowledgeRole::getId, groupIdList);
|
||||
} else {
|
||||
roleLqw.in(KnowledgeRole::getGroupId, groupIdList);
|
||||
}
|
||||
knowledgeRoles = knowledgeRoleMapper.selectList(roleLqw);
|
||||
|
||||
if (CollectionUtils.isEmpty(knowledgeRoles)) {
|
||||
return new TableDataInfo<>();
|
||||
}
|
||||
|
||||
// 查询知识库id列表
|
||||
LambdaQueryWrapper<KnowledgeRoleRelation> relationLqw = Wrappers.lambdaQuery();
|
||||
relationLqw.in(KnowledgeRoleRelation::getKnowledgeRoleId, knowledgeRoles.stream().map(KnowledgeRole::getId).filter(Objects::nonNull).collect(Collectors.toList()));
|
||||
List<KnowledgeRoleRelation> knowledgeRoleRelations = knowledgeRoleRelationMapper.selectList(relationLqw);
|
||||
|
||||
if (CollectionUtils.isEmpty(knowledgeRoleRelations)) {
|
||||
return new TableDataInfo<>();
|
||||
}
|
||||
|
||||
|
||||
// 构建查询条件
|
||||
LambdaQueryWrapper<KnowledgeInfo> lqw = buildQueryWrapper(bo);
|
||||
// 在查询用户创建的知识库条件下，拼接角色分配知识库
|
||||
lqw.or(q -> q.in(
|
||||
KnowledgeInfo::getId,
|
||||
knowledgeRoleRelations.stream()
|
||||
.map(KnowledgeRoleRelation::getKnowledgeId)
|
||||
.filter(Objects::nonNull)
|
||||
.collect(Collectors.toList())
|
||||
));
|
||||
|
||||
// 管理员用户直接查询所有数据
|
||||
if (Objects.equals(loginUser.getUserId(), 1L)) {
|
||||
Page<KnowledgeInfoVo> result = baseMapper.selectVoPage(pageQuery.build(), lqw);
|
||||
return TableDataInfo.build(result);
|
||||
}
|
||||
|
||||
// 检查用户是否配置了角色信息
|
||||
if (StringUtils.isNotEmpty(loginUser.getKroleGroupIds()) && StringUtils.isNotEmpty(loginUser.getKroleGroupType())) {
|
||||
// 角色/角色组id列表
|
||||
List<String> groupIdList = Arrays.stream(loginUser.getKroleGroupIds().split(","))
|
||||
.filter(StringUtils::isNotEmpty)
|
||||
.toList();
|
||||
|
||||
// 查询用户关联的角色
|
||||
List<KnowledgeRole> knowledgeRoles = new ArrayList<>();
|
||||
LambdaQueryWrapper<KnowledgeRole> roleLqw = Wrappers.lambdaQuery();
|
||||
if ("role".equals(loginUser.getKroleGroupType())) {
|
||||
roleLqw.in(KnowledgeRole::getId, groupIdList);
|
||||
} else {
|
||||
roleLqw.in(KnowledgeRole::getGroupId, groupIdList);
|
||||
}
|
||||
knowledgeRoles = knowledgeRoleMapper.selectList(roleLqw);
|
||||
|
||||
// 如果用户有关联角色
|
||||
if (!CollectionUtils.isEmpty(knowledgeRoles)) {
|
||||
// 查询这些角色关联的知识库
|
||||
LambdaQueryWrapper<KnowledgeRoleRelation> relationLqw = Wrappers.lambdaQuery();
|
||||
relationLqw.in(KnowledgeRoleRelation::getKnowledgeRoleId,
|
||||
knowledgeRoles.stream().map(KnowledgeRole::getId).filter(Objects::nonNull).collect(Collectors.toList()));
|
||||
List<KnowledgeRoleRelation> knowledgeRoleRelations = knowledgeRoleRelationMapper.selectList(relationLqw);
|
||||
|
||||
// 如果角色关联了知识库
|
||||
if (!CollectionUtils.isEmpty(knowledgeRoleRelations)) {
|
||||
// 查询用户自己的知识库和角色分配的知识库
|
||||
lqw.and(q -> q.eq(KnowledgeInfo::getUid, loginUser.getUserId())
|
||||
.or()
|
||||
.in(KnowledgeInfo::getId,
|
||||
knowledgeRoleRelations.stream()
|
||||
.map(KnowledgeRoleRelation::getKnowledgeId)
|
||||
.filter(Objects::nonNull)
|
||||
.collect(Collectors.toList())
|
||||
)
|
||||
);
|
||||
} else {
|
||||
// 用户没有关联任何知识库，只显示自己的
|
||||
lqw.eq(KnowledgeInfo::getUid, loginUser.getUserId());
|
||||
}
|
||||
} else {
|
||||
// 用户没有关联角色，只显示自己的
|
||||
lqw.eq(KnowledgeInfo::getUid, loginUser.getUserId());
|
||||
}
|
||||
} else {
|
||||
// 用户没有配置角色信息，只显示自己的
|
||||
lqw.eq(KnowledgeInfo::getUid, loginUser.getUserId());
|
||||
}
|
||||
|
||||
Page<KnowledgeInfoVo> result = baseMapper.selectVoPage(pageQuery.build(), lqw);
|
||||
return TableDataInfo.build(result);
|
||||
}
|
||||
|
||||
Reference in New Issue
Block a user