接上回--综合AIDemo测试

一，前言

上回外挂了知识库之后，我们需要使用知识库中的信息，让AI为我们实际处理业务上的需求。

这里我们让AI扮演公司的人事助手，帮我们处理员工请假的业务。

具体流程如下

• 感知用户需要请假
• 提取用户请假信息
• 获取用户数据库中的信息
• 根据用户角色判断请假时长是否符合公司要求（或者是否需要推送给上级审批）
• 不符合返回理由，符合返回确认信息
• 用户确认后往数据库中写入请假信息

在获取请假信息这块上，如果我们手动去提取用户的msg，那么会有很大的可能会提取失败，所以这块尽可能让ai从我们的对话上下文去提取所需要的信息。

二，需要注意的问题

另外有个很关键的点，就是在使用fuctions处理问题的时候，需要注意，fuctions返回的应该是严格的数据格式，而不能是一段对话。因为这段代码

		if (!isProxyToolCalls(prompt, this.defaultOptions)&& isToolCall(response, Set.of(OpenAiApi.ChatCompletionFinishReason.TOOL_CALLS.name(),OpenAiApi.ChatCompletionFinishReason.STOP.name()))) {var toolCallConversation = handleToolCalls(prompt, response);// Recursively call the call method with the tool call message// conversation that contains the call responses.return this.internalCall(new Prompt(toolCallConversation, prompt.getOptions()), response);}

这段代码中调用完fuction方法后会再次调用aiCall进行对话请求，这个时候你可能就会陷入无尽的loop中了。

方法的完整代码如下

	public ChatResponse internalCall(Prompt prompt, ChatResponse previousChatResponse) {ChatCompletionRequest request = createRequest(prompt, false);ChatModelObservationContext observationContext = ChatModelObservationContext.builder().prompt(prompt).provider(OpenAiApiConstants.PROVIDER_NAME).requestOptions(buildRequestOptions(request)).build();ChatResponse response = ChatModelObservationDocumentation.CHAT_MODEL_OPERATION.observation(this.observationConvention, DEFAULT_OBSERVATION_CONVENTION, () -> observationContext,this.observationRegistry).observe(() -> {ResponseEntity<ChatCompletion> completionEntity = this.retryTemplate.execute(ctx -> this.openAiApi.chatCompletionEntity(request, getAdditionalHttpHeaders(prompt)));var chatCompletion = completionEntity.getBody();if (chatCompletion == null) {logger.warn("No chat completion returned for prompt: {}", prompt);return new ChatResponse(List.of());}List<Choice> choices = chatCompletion.choices();if (choices == null) {logger.warn("No choices returned for prompt: {}", prompt);return new ChatResponse(List.of());}List<Generation> generations = choices.stream().map(choice -> {// @formatter:offMap<String, Object> metadata = Map.of("id", chatCompletion.id() != null ? chatCompletion.id() : "","role", choice.message().role() != null ? choice.message().role().name() : "","index", choice.index(),"finishReason", choice.finishReason() != null ? choice.finishReason().name() : "","refusal", StringUtils.hasText(choice.message().refusal()) ? choice.message().refusal() : "");// @formatter:onreturn buildGeneration(choice, metadata, request);}).toList();// Non function calling.RateLimit rateLimit = OpenAiResponseHeaderExtractor.extractAiResponseHeaders(completionEntity);// Current usageOpenAiApi.Usage usage = completionEntity.getBody().usage();Usage currentChatResponseUsage = usage != null ? OpenAiUsage.from(usage) : new EmptyUsage();Usage accumulatedUsage = UsageUtils.getCumulativeUsage(currentChatResponseUsage, previousChatResponse);ChatResponse chatResponse = new ChatResponse(generations,from(completionEntity.getBody(), rateLimit, accumulatedUsage));observationContext.setResponse(chatResponse);return chatResponse;});if (!isProxyToolCalls(prompt, this.defaultOptions)&& isToolCall(response, Set.of(OpenAiApi.ChatCompletionFinishReason.TOOL_CALLS.name(),OpenAiApi.ChatCompletionFinishReason.STOP.name()))) {var toolCallConversation = handleToolCalls(prompt, response);// Recursively call the call method with the tool call message// conversation that contains the call responses.return this.internalCall(new Prompt(toolCallConversation, prompt.getOptions()), response);}return response;}

我们来完善一下这块逻辑代码。

三，主要代码

首先我们需要建立拦截，去拦截用户信息，为了简便我就不进行加密token的处理了。有需要可以看我springboot下对这块的处理。

package org.example.springaidemo.filter;import jakarta.servlet.*;
import jakarta.servlet.http.HttpServletRequest;
import jakarta.servlet.http.HttpServletResponse;
import org.example.springaidemo.context.UserContext;
import org.example.springaidemo.moudle.UserModel;
import org.example.springaidemo.service.UserService;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.stereotype.Component;
import org.springframework.util.StringUtils;import java.io.IOException;@Component
public class UserAuthFilter implements Filter {@Autowiredprivate UserService userService;@Overridepublic void doFilter(ServletRequest request, ServletResponse response, FilterChain chain)throws IOException, ServletException {HttpServletRequest httpRequest = (HttpServletRequest) request;HttpServletResponse httpResponse = (HttpServletResponse) response;try {// 获取tokenString token = httpRequest.getHeader("token");if (!StringUtils.hasText(token)) {httpResponse.setStatus(HttpServletResponse.SC_UNAUTHORIZED);httpResponse.getWriter().write("Missing token");return;}// 获取用户信息UserModel user = userService.getUserByToken(token);if (user == null) {httpResponse.setStatus(HttpServletResponse.SC_UNAUTHORIZED);httpResponse.getWriter().write("Invalid token");return;}// 设置用户上下文UserContext.setCurrentUser(user);// 继续处理请求chain.doFilter(request, response);} finally {// 清理用户上下文UserContext.clear();}}
} 

这样我们可以从内存中快速提取用户信息。

核心业务逻辑

package org.example.springaidemo.impl;import lombok.extern.slf4j.Slf4j;
import org.example.springaidemo.config.AiMyFunction;
import org.example.springaidemo.config.knowledge.TextKnowledgeBase;
import org.example.springaidemo.context.UserContext;
import org.example.springaidemo.dao.mapper.Mapper1;
import org.example.springaidemo.moudle.UserModel;
import org.example.springaidemo.moudle.UserRestModel;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.scheduling.annotation.Async;
import org.springframework.stereotype.Service;import java.time.LocalDateTime;
import java.time.format.DateTimeFormatter;
import java.util.List;
import java.util.concurrent.CompletableFuture;@Service
@Slf4j
public class SimpleControllerImpl {@Autowiredprivate AiMyFunction aiMyFunction;@Autowiredprivate TextKnowledgeBase textKnowledgeBase;@Autowiredprivate Mapper1 mapper1;public String chat(String message, String token) {// 如果是休假相关的请求，使用特殊处理if (message.contains("休假") || message.contains("请假")) {return handleRestRequest(message, token);}// 普通对话处理List<String> knowledgeResults = textKnowledgeBase.searchKnowledgeAll();StringBuilder promptBuilder = new StringBuilder();if (!knowledgeResults.isEmpty()) {promptBuilder.append("根据以下参考知识回答问题：\n\n");for (String knowledge : knowledgeResults) {promptBuilder.append("- ").append(knowledge).append("\n");}promptBuilder.append("\n问题是：").append(message);} else {promptBuilder.append(message);}return aiMyFunction.generateNormal(promptBuilder.toString(), token);}@Asyncpublic CompletableFuture<Void> processRestAsync(String userId, LocalDateTime startTime, LocalDateTime endTime, String restType, String status, String reason) {return CompletableFuture.runAsync(() -> {try {UserRestModel restModel = new UserRestModel();restModel.setUserid(userId);restModel.setStartTime(startTime);restModel.setEndTime(endTime);restModel.setRestType(restType);restModel.setStatus(status);if ("REJECTED".equals(status)) {restModel.setRejectReason(reason);}mapper1.addUserRest(restModel);log.info("Successfully processed rest request for user: {}", userId);} catch (Exception e) {log.error("Error processing rest request for user: {}", userId, e);}});}private String handleRestRequest(String message, String token) {UserModel user = UserContext.getCurrentUser();if (user == null) {return "未找到用户信息";}// 构建提示词，包含用户角色和规则信息StringBuilder promptBuilder = new StringBuilder();promptBuilder.append("当前用户角色：").append(user.getRoles()).append("\n");promptBuilder.append("请根据以下规则分析休假申请：\n");List<String> rules = textKnowledgeBase.searchKnowledgeAll();for (String rule : rules) {promptBuilder.append("- ").append(rule).append("\n");}promptBuilder.append("\n请分析以下休假申请，并从中提取开始时间、结束时间和休假类型：\n");promptBuilder.append(message).append("\n");promptBuilder.append("\n如果信息不完整或格式不正确，请直接回复提示用户正确的格式。");promptBuilder.append("\n如果信息完整，请按以下格式回复：");promptBuilder.append("\n批准：[分析理由]|[开始时间]|[结束时间]|[休假类型]");promptBuilder.append("\n或");promptBuilder.append("\n拒绝：[拒绝原因]|[开始时间]|[结束时间]|[休假类型]");// 调用AI处理休假请求String aiResponse = aiMyFunction.generateNormal(promptBuilder.toString(), token);log.info("AI Response for rest request: {}", aiResponse);// 解析AI响应并异步处理数据库操作if (aiResponse.startsWith("批准：")) {String[] parts = aiResponse.substring(3).split("\\|");if (parts.length != 4) {return "AI响应格式错误，请重试";}String reason = parts[0];LocalDateTime startTime = LocalDateTime.parse(parts[1].trim() + " 00:00:00", DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"));LocalDateTime endTime = LocalDateTime.parse(parts[2].trim() + " 23:59:59", DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"));String restType = parts[3].trim();processRestAsync(user.getUserid(), startTime, endTime, restType, "PENDING", null);return String.format("批准：%s\n请确认以下休假信息：\n开始时间：%s\n结束时间：%s\n休假类型：%s", reason, startTime.toLocalDate(), endTime.toLocalDate(), restType);} else if (aiResponse.startsWith("拒绝：")) {String[] parts = aiResponse.substring(3).split("\\|");if (parts.length != 4) {return "AI响应格式错误，请重试";}String reason = parts[0];LocalDateTime startTime = LocalDateTime.parse(parts[1].trim() + " 00:00:00", DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"));LocalDateTime endTime = LocalDateTime.parse(parts[2].trim() + " 23:59:59", DateTimeFormatter.ofPattern("yyyy-MM-dd HH:mm:ss"));String restType = parts[3].trim();processRestAsync(user.getUserid(), startTime, endTime, restType, "REJECTED", reason);return String.format("拒绝：%s", reason);}return aiResponse;}public Double priceAll(int count) {double price = 3.25;return price * count;}
}

1. 这里首先从对话中catch到用户的请假需求。
2. 由ai自动提取所需信息，结合我们外挂的知识库进行分析用户的请假需求是否合理。
3. 在ai给出回答后，异步提取回答内容，写入我们业务数据库

注意异步任务的开启需要使用注解

@EnableAsync

你可以新建一个类

package org.example.springaidemo.config;import org.springframework.context.annotation.Configuration;
import org.springframework.scheduling.annotation.EnableAsync;@Configuration
@EnableAsync
public class AsyncConfig {
} 

或者直接加到启动类注解上

package org.example.springaidemo;import org.springframework.boot.SpringApplication;
import org.springframework.boot.autoconfigure.SpringBootApplication;
import org.springframework.context.annotation.ComponentScan;
import org.springframework.scheduling.annotation.EnableAsync;@SpringBootApplication
@ComponentScan({"org.springframework.ai.chat", "org.example.springaidemo"})
@EnableAsync
public class SpringAiDemoApplication {public static void main(String[] args) {SpringApplication.run(SpringAiDemoApplication.class, args);}
}

其它边缘性的代码，就不截图了。相信能看到这块的都是老程序员，这点小case轻轻松松。

另外之前 我都是使用英特

现在使用的是

DeepSeek | 深度求索

DeepSeek

切换也很平滑。改几个配置项即可。

spring.application.name=springAiDemo
spring.ai.openai.api-key=密钥 //改一下密钥
spring.ai.openai.base-url=https://api.deepseek.com // 改一下urlspring.ai.openai.chat.options.model=deepseek-chat  // 改一下使用模型

我们来看看测试如何

四，测试

开启对话

按ai给的提示再次对话

查看数据库写入情况