feat(ai): 实现对话记忆与消息持久化功能

- 移除 Cassandra 相关配置及依赖
- 新增 CustomChatMemoryAdvisor 实现对话记忆管理
-重命名并扩展 CustomStreamLoggerAdvisor 为 CustomStreamLoggerAndMessage2DBAdvisor，增加消息入库逻辑
- 在 ChatController 中集成新的 Advisor 并注入相关依赖
- 使用 TransactionTemplate 管理消息存储事务
-限制记忆消息数量为最新 50 条
- 支持将用户消息与 AI 回答同步写入数据库

pom.xml

@@ -43,11 +43,6 @@
             <groupId>com.alibaba.cloud.ai</groupId>
             <artifactId>spring-ai-alibaba-starter-dashscope</artifactId>
         </dependency>
-        <!-- Cassandra -->
-        <dependency>
-            <groupId>org.springframework.ai</groupId>
-            <artifactId>spring-ai-starter-model-chat-memory-repository-cassandra</artifactId>
-        </dependency>
         <dependency>
             <groupId>org.springframework.boot</groupId>
             <artifactId>spring-boot-starter-test</artifactId>

src/main/java/com/hanserwei/airobot/advisor/CustomChatMemoryAdvisor.java

@@ -0,0 +1,100 @@
+package com.hanserwei.airobot.advisor;
+
+import com.baomidou.mybatisplus.core.toolkit.Wrappers;
+import com.google.common.collect.Lists;
+import com.hanserwei.airobot.domain.dos.ChatMessageDO;
+import com.hanserwei.airobot.domain.mapper.ChatMessageMapper;
+import com.hanserwei.airobot.model.vo.chat.AiChatReqVO;
+import lombok.extern.slf4j.Slf4j;
+import org.jetbrains.annotations.NotNull;
+import org.springframework.ai.chat.client.ChatClientRequest;
+import org.springframework.ai.chat.client.ChatClientResponse;
+import org.springframework.ai.chat.client.advisor.api.StreamAdvisor;
+import org.springframework.ai.chat.client.advisor.api.StreamAdvisorChain;
+import org.springframework.ai.chat.messages.AssistantMessage;
+import org.springframework.ai.chat.messages.Message;
+import org.springframework.ai.chat.messages.MessageType;
+import org.springframework.ai.chat.messages.UserMessage;
+import reactor.core.publisher.Flux;
+
+import java.util.Comparator;
+import java.util.List;
+import java.util.Objects;
+
+@Slf4j
+public class CustomChatMemoryAdvisor implements StreamAdvisor {
+
+    private final ChatMessageMapper chatMessageMapper;
+    private final AiChatReqVO aiChatReqVO;
+    private final int limit;
+
+    public CustomChatMemoryAdvisor(ChatMessageMapper chatMessageMapper, AiChatReqVO aiChatReqVO, int limit) {
+        this.chatMessageMapper = chatMessageMapper;
+        this.aiChatReqVO = aiChatReqVO;
+        this.limit = limit;
+    }
+
+    @Override
+    public int getOrder() {
+        return 2; // order 值越小，越先执行
+    }
+
+    @NotNull
+    @Override
+    public String getName() {
+        return this.getClass().getSimpleName();
+    }
+
+    @NotNull
+    @Override
+    public Flux<ChatClientResponse> adviseStream(ChatClientRequest chatClientRequest, StreamAdvisorChain streamAdvisorChain) {
+        log.info("## 自定义聊天记忆 Advisor...");
+
+        // 对话 UUID
+        String chatUuid = aiChatReqVO.getChatId();
+
+        // 查询数据库拉取最新的聊天消息
+        List<ChatMessageDO> messages = chatMessageMapper.selectList(Wrappers.<ChatMessageDO>lambdaQuery()
+                .eq(ChatMessageDO::getChatUuid, chatUuid) // 查询指定对话 UUID 下的聊天记录
+                .orderByDesc(ChatMessageDO::getCreateTime) // 查询最新的消息
+                .last(String.format("LIMIT %d", limit))); // 仅查询 LIMIT 条
+
+        // 按发布时间升序排列
+        List<ChatMessageDO> sortedMessages = messages.stream()
+                 .sorted(Comparator.comparing(ChatMessageDO::getCreateTime)) // 升序排列
+                 .toList();
+
+        // 所有消息
+        List<Message> messageList = getMessageList(sortedMessages);
+
+        // 除了记忆消息，还需要添加当前用户消息
+        messageList.addAll(chatClientRequest.prompt().getInstructions());
+
+        // 构建一个新的 ChatClientRequest 请求对象
+        ChatClientRequest processedChatClientRequest = chatClientRequest
+                .mutate()
+                .prompt(chatClientRequest.prompt().mutate().messages(messageList).build())
+                .build();
+
+        return streamAdvisorChain.nextStream(processedChatClientRequest);
+    }
+
+    @NotNull
+    private static List<Message> getMessageList(List<ChatMessageDO> sortedMessages) {
+        List<Message> messageList = Lists.newArrayList();
+
+        // 将数据库记录转换为对应类型的消息
+        for (ChatMessageDO chatMessageDO : sortedMessages) {
+            // 消息类型
+            String type  = chatMessageDO.getRole();
+            if (Objects.equals(type, MessageType.USER.getValue())) { // 用户消息
+                Message userMessage = new UserMessage(chatMessageDO.getContent());
+                messageList.add(userMessage);
+            } else if (Objects.equals(type, MessageType.ASSISTANT.getValue())) { // AI 助手消息
+                Message assistantMessage = new AssistantMessage(chatMessageDO.getContent());
+                messageList.add(assistantMessage);
+            }
+        }
+        return messageList;
+    }
+}

src/main/java/com/hanserwei/airobot/advisor/CustomStreamLoggerAdvisor.java

@@ -1,63 +0,0 @@
-package com.hanserwei.airobot.advisor;
-
-import lombok.extern.slf4j.Slf4j;
-import org.jetbrains.annotations.NotNull;
-import org.springframework.ai.chat.client.ChatClientRequest;
-import org.springframework.ai.chat.client.ChatClientResponse;
-import org.springframework.ai.chat.client.advisor.api.StreamAdvisor;
-import org.springframework.ai.chat.client.advisor.api.StreamAdvisorChain;
-import reactor.core.publisher.Flux;
-
-import java.util.concurrent.atomic.AtomicReference;
-
-@Slf4j
-public class CustomStreamLoggerAdvisor implements StreamAdvisor {
-
-    @Override
-    public int getOrder() {
-        return 99; // order 值越小，越先执行
-    }
-
-    @NotNull
-    @Override
-    public String getName() {
-        return this.getClass().getSimpleName();
-    }
-
-    @NotNull
-    @Override
-    public Flux<ChatClientResponse> adviseStream(@NotNull ChatClientRequest chatClientRequest, StreamAdvisorChain streamAdvisorChain) {
-		
-        Flux<ChatClientResponse> chatClientResponseFlux = streamAdvisorChain.nextStream(chatClientRequest);
-
-        // 创建 AI 流式回答聚合容器（线程安全）
-        AtomicReference<StringBuilder> fullContent = new AtomicReference<>(new StringBuilder());
-
-        // 返回处理后的流
-        return chatClientResponseFlux
-                .doOnNext(response -> {
-                    // 逐块收集内容
-                    String chunk = null;
-                    if (response.chatResponse() != null) {
-                        chunk = response.chatResponse().getResult().getOutput().getText();
-                    }
-
-                    log.info("## chunk: {}", chunk);
-
-                    // 若 chunk 块不为空，则追加到 fullContent 中
-                    if (chunk != null) {
-                        fullContent.get().append(chunk);
-                    }
-                })
-                .doOnComplete(() -> {
-                    // 流完成后打印完整回答
-                    String completeResponse = fullContent.get().toString();
-                    log.info("\n==== FULL AI RESPONSE ====\n{}\n========================", completeResponse);
-                })
-                .doOnError(error -> {
-                    // 出错时打印已收集的部分
-                    String partialResponse = fullContent.get().toString();
-                    log.error("## Stream 流出现错误，已收集回答如下: {}", partialResponse, error);
-                });
-    }
-}

src/main/java/com/hanserwei/airobot/advisor/CustomStreamLoggerAndMessage2DBAdvisor.java

@@ -0,0 +1,108 @@
+package com.hanserwei.airobot.advisor;
+
+import com.hanserwei.airobot.domain.dos.ChatMessageDO;
+import com.hanserwei.airobot.domain.mapper.ChatMessageMapper;
+import com.hanserwei.airobot.model.vo.chat.AiChatReqVO;
+import lombok.extern.slf4j.Slf4j;
+import org.jetbrains.annotations.NotNull;
+import org.springframework.ai.chat.client.ChatClientRequest;
+import org.springframework.ai.chat.client.ChatClientResponse;
+import org.springframework.ai.chat.client.advisor.api.StreamAdvisor;
+import org.springframework.ai.chat.client.advisor.api.StreamAdvisorChain;
+import org.springframework.ai.chat.messages.MessageType;
+import org.springframework.transaction.support.TransactionTemplate;
+import reactor.core.publisher.Flux;
+
+import java.time.LocalDateTime;
+import java.util.concurrent.atomic.AtomicReference;
+
+@Slf4j
+public class CustomStreamLoggerAndMessage2DBAdvisor implements StreamAdvisor {
+
+    private final ChatMessageMapper chatMessageMapper;
+    private final AiChatReqVO aiChatReqVO;
+    private final TransactionTemplate transactionTemplate;
+
+    public CustomStreamLoggerAndMessage2DBAdvisor(ChatMessageMapper chatMessageMapper,
+                                                  AiChatReqVO aiChatReqVO,
+                                                  TransactionTemplate transactionTemplate) {
+        this.chatMessageMapper = chatMessageMapper;
+        this.aiChatReqVO = aiChatReqVO;
+        this.transactionTemplate = transactionTemplate;
+    }
+
+    @Override
+    public int getOrder() {
+        return 99; // order 值越小，越先执行
+    }
+
+    @NotNull
+    @Override
+    public String getName() {
+        return this.getClass().getSimpleName();
+    }
+
+    @NotNull
+    @Override
+    public Flux<ChatClientResponse> adviseStream(@NotNull ChatClientRequest chatClientRequest, StreamAdvisorChain streamAdvisorChain) {
+		
+        Flux<ChatClientResponse> chatClientResponseFlux = streamAdvisorChain.nextStream(chatClientRequest);
+
+        // 对话 UUID
+        String chatUuid = aiChatReqVO.getChatId();
+        // 用户消息
+        String userMessage = aiChatReqVO.getMessage();
+
+        // 创建 AI 流式回答聚合容器（线程安全）
+        AtomicReference<StringBuilder> fullContent = new AtomicReference<>(new StringBuilder());
+
+        // 返回处理后的流
+        return chatClientResponseFlux
+                .doOnNext(response -> {
+                    // 逐块收集内容
+                    String chunk = null;
+                    if (response.chatResponse() != null) {
+                        chunk = response.chatResponse().getResult().getOutput().getText();
+                    }
+
+                    log.info("## chunk: {}", chunk);
+
+                    // 若 chunk 块不为空，则追加到 fullContent 中
+                    if (chunk != null) {
+                        fullContent.get().append(chunk);
+                    }
+                })
+                .doOnComplete(() -> {
+                    // 流完成后打印完整回答
+                    String completeResponse = fullContent.get().toString();
+                    log.info("\n==== FULL AI RESPONSE ====\n{}\n========================", completeResponse);
+
+                    // 开启编程式事务
+                    transactionTemplate.execute(status -> {
+                        try {
+                            // 1. 存储用户消息
+                            chatMessageMapper.insert(ChatMessageDO.builder()
+                                    .chatUuid(chatUuid)
+                                    .content(userMessage)
+                                    .role(MessageType.USER.getValue()) // 用户消息
+                                    .createTime(LocalDateTime.now())
+                                    .build());
+
+                            // 2. 存储 AI 回答
+                            chatMessageMapper.insert(ChatMessageDO.builder()
+                                    .chatUuid(chatUuid)
+                                    .content(completeResponse)
+                                    .role(MessageType.ASSISTANT.getValue()) // AI 回答
+                                    .createTime(LocalDateTime.now())
+                                    .build());
+
+                            return true;
+                        } catch (Exception ex) {
+                            status.setRollbackOnly(); // 标记事务为回滚
+                            log.error("", ex);
+                        }
+                        return false;
+                    });
+                });
+    }
+}

src/main/java/com/hanserwei/airobot/controller/ChatController.java

@@ -4,8 +4,10 @@ import com.alibaba.cloud.ai.dashscope.api.DashScopeApi;
 import com.alibaba.cloud.ai.dashscope.chat.DashScopeChatModel;
 import com.alibaba.cloud.ai.dashscope.chat.DashScopeChatOptions;
 import com.google.common.collect.Lists;
-import com.hanserwei.airobot.advisor.CustomStreamLoggerAdvisor;
+import com.hanserwei.airobot.advisor.CustomChatMemoryAdvisor;
+import com.hanserwei.airobot.advisor.CustomStreamLoggerAndMessage2DBAdvisor;
 import com.hanserwei.airobot.aspect.ApiOperationLog;
+import com.hanserwei.airobot.domain.mapper.ChatMessageMapper;
 import com.hanserwei.airobot.model.vo.chat.AiChatReqVO;
 import com.hanserwei.airobot.model.vo.chat.AiResponse;
 import com.hanserwei.airobot.model.vo.chat.NewChatReqVO;
@@ -18,6 +20,7 @@ import org.springframework.ai.chat.client.advisor.api.Advisor;
 import org.springframework.ai.chat.model.ChatModel;
 import org.springframework.beans.factory.annotation.Value;
 import org.springframework.http.MediaType;
+import org.springframework.transaction.support.TransactionTemplate;
 import org.springframework.validation.annotation.Validated;
 import org.springframework.web.bind.annotation.PostMapping;
 import org.springframework.web.bind.annotation.RequestBody;
@@ -34,6 +37,10 @@ public class ChatController {
 
     @Resource
     private ChatService chatService;
+    @Resource
+    private ChatMessageMapper chatMessageMapper;
+    @Resource
+    private TransactionTemplate transactionTemplate;
 
     @Value("${spring.ai.dashscope.api-key}")
     private String apiKey;
@@ -72,8 +79,10 @@ public class ChatController {
 
         // Advisor 集合
         List<Advisor> advisors = Lists.newArrayList();
+        // 添加自定义对话记忆 Advisor（以最新的 50 条消息作为记忆）
+        advisors.add(new CustomChatMemoryAdvisor(chatMessageMapper, aiChatReqVO, 50));
         // 添加自定义打印流式对话日志 Advisor
-        advisors.add(new CustomStreamLoggerAdvisor());
+        advisors.add(new CustomStreamLoggerAndMessage2DBAdvisor(chatMessageMapper, aiChatReqVO, transactionTemplate));
 
         // 应用 Advisor 集合
         chatClientRequestSpec.advisors(advisors);

src/main/resources/application.yml

@@ -17,10 +17,6 @@ spring:
       maximum-pool-size: 20 # 最大连接池大小
       connection-test-query: SELECT 1 # 连接测试查询
       validation-timeout: 5000 # 验证连接的有效性
-  cassandra:
-    contact-points: 127.0.0.1 # Cassandra 集群节点地址（可配置多个，用逗号分隔）
-    port: 9042 # 端口号
-    local-datacenter: datacenter1 # 必须与集群配置的数据中心名称一致（大小写敏感）
   ai:
     dashscope:
       api-key: ENC(cMgcKZkFllyE88DIbGwLKot9Vg02co+gsmY8L8o4/o3UjhcmqO4lJzFU35Sx0n+qFG8pDL0wBjoWrT8X6BuRw9vNlQhY1LgRWHaF9S1zzyM=)