feat(ai): 优化AI模块路由，支持文档分片管理功能

- 将AI对话与知识库控制器基础路径改为/api/v1/前缀
- 修改ChatRequest中sessionId类型为String，统一会话ID处理
- 新增DocumentChunkVO用于表示文档分片实体
- 在AiKnowledgeBaseService及实现类中添加获取文档分片列表及详情接口
- 在AiKnowledgeBaseController新增两个REST API：获取文档分片列表和获取分片详情
- AiDocumentMapper新增数据库操作方法selectDocumentChunks和selectChunkById
- 修改向量存储切片保存逻辑，确保分片含完整项目和文档属性元数据
- 优化文档分片持久化时更新原始

src/main/java/cn/yinlihupo/controller/ai/AiChatController.java

@@ -25,7 +25,7 @@ import java.util.UUID;
  */
 @Slf4j
 @RestController
-@RequestMapping("/ai/chat")
+@RequestMapping("/api/v1/ai/chat")
 @RequiredArgsConstructor
 @Tag(name = "AI对话", description = "AI智能问答相关接口")
 public class AiChatController {

src/main/java/cn/yinlihupo/controller/ai/AiKnowledgeBaseController.java

@@ -3,6 +3,7 @@ package cn.yinlihupo.controller.ai;
 import cn.yinlihupo.common.core.BaseResponse;
 import cn.yinlihupo.common.util.ResultUtils;
 import cn.yinlihupo.common.util.SecurityUtils;
+import cn.yinlihupo.domain.vo.DocumentChunkVO;
 import cn.yinlihupo.domain.vo.KbDocumentVO;
 import cn.yinlihupo.service.ai.AiKnowledgeBaseService;
 import io.swagger.v3.oas.annotations.Operation;
@@ -20,7 +21,7 @@ import java.util.List;
  */
 @Slf4j
 @RestController
-@RequestMapping("/ai/kb")
+@RequestMapping("/api/v1/ai/kb")
 @RequiredArgsConstructor
 @Tag(name = "AI知识库", description = "AI知识库文档管理相关接口")
 public class AiKnowledgeBaseController {
@@ -134,4 +135,53 @@ public class AiKnowledgeBaseController {
             return ResultUtils.error("重新索引失败: " + e.getMessage());
         }
     }
+
+    /**
+     * 获取文档分片列表
+     *
+     * @param docId 文档UUID
+     * @return 分片列表
+     */
+    @GetMapping("/document/{docId}/chunks")
+    @Operation(summary = "获取文档分片列表", description = "获取指定文档的所有分片信息")
+    public BaseResponse<List<DocumentChunkVO>> getDocumentChunks(@PathVariable String docId) {
+        Long userId = SecurityUtils.getCurrentUserId();
+        if (userId == null) {
+            return ResultUtils.error("用户未登录");
+        }
+
+        try {
+            List<DocumentChunkVO> chunks = knowledgeBaseService.getDocumentChunks(docId);
+            return ResultUtils.success("查询成功", chunks);
+        } catch (Exception e) {
+            log.error("获取文档分片失败: {}", e.getMessage(), e);
+            return ResultUtils.error("获取文档分片失败: " + e.getMessage());
+        }
+    }
+
+    /**
+     * 获取分片详情
+     *
+     * @param chunkId 分片ID
+     * @return 分片详情
+     */
+    @GetMapping("/chunk/{chunkId}")
+    @Operation(summary = "获取分片详情", description = "获取指定分片的详细信息")
+    public BaseResponse<DocumentChunkVO> getChunkDetail(@PathVariable String chunkId) {
+        Long userId = SecurityUtils.getCurrentUserId();
+        if (userId == null) {
+            return ResultUtils.error("用户未登录");
+        }
+
+        try {
+            DocumentChunkVO chunk = knowledgeBaseService.getChunkDetail(chunkId);
+            if (chunk == null) {
+                return ResultUtils.error("分片不存在");
+            }
+            return ResultUtils.success("查询成功", chunk);
+        } catch (Exception e) {
+            log.error("获取分片详情失败: {}", e.getMessage(), e);
+            return ResultUtils.error("获取分片详情失败: " + e.getMessage());
+        }
+    }
 }

src/main/java/cn/yinlihupo/domain/dto/ChatRequest.java

@@ -2,8 +2,6 @@ package cn.yinlihupo.domain.dto;
 
 import lombok.Data;
 
-import java.util.UUID;
-
 /**
  * AI对话请求DTO
  */
@@ -13,7 +11,7 @@ public class ChatRequest {
     /**
      * 会话ID（为空则新建会话）
      */
-    private UUID sessionId;
+    private String sessionId;
 
     /**
      * 项目ID（必填）

src/main/java/cn/yinlihupo/domain/vo/DocumentChunkVO.java

@@ -0,0 +1,55 @@
+package cn.yinlihupo.domain.vo;
+
+import lombok.Data;
+
+/**
+ * 文档分片VO
+ */
+@Data
+public class DocumentChunkVO {
+
+    /**
+     * 分片ID
+     */
+    private String id;
+
+    /**
+     * 原始文档ID
+     */
+    private String docId;
+
+    /**
+     * 分片内容
+     */
+    private String content;
+
+    /**
+     * 分片序号
+     */
+    private Integer chunkIndex;
+
+    /**
+     * 总分片数
+     */
+    private Integer chunkTotal;
+
+    /**
+     * 文档标题
+     */
+    private String title;
+
+    /**
+     * 文档类型
+     */
+    private String docType;
+
+    /**
+     * 来源类型
+     */
+    private String sourceType;
+
+    /**
+     * 状态
+     */
+    private String status;
+}

src/main/java/cn/yinlihupo/mapper/AiDocumentMapper.java

@@ -1,6 +1,7 @@
 package cn.yinlihupo.mapper;
 
 import cn.yinlihupo.domain.entity.AiDocument;
+import cn.yinlihupo.domain.vo.DocumentChunkVO;
 import cn.yinlihupo.domain.vo.KbDocumentVO;
 import cn.yinlihupo.domain.vo.ReferencedDocVO;
 import com.baomidou.mybatisplus.core.mapper.BaseMapper;
@@ -88,4 +89,21 @@ public interface AiDocumentMapper extends BaseMapper<AiDocument> {
      * @return 影响行数
      */
     int incrementQueryCount(@Param("id") String id);
+
+    /**
+     * 查询文档分片列表
+     * 通过 metadata 中的 doc_id 字段查询关联的分片
+     *
+     * @param docId 文档ID
+     * @return 分片列表
+     */
+    List<DocumentChunkVO> selectDocumentChunks(@Param("docId") String docId);
+
+    /**
+     * 查询文档分片详情
+     *
+     * @param chunkId 分片ID
+     * @return 分片详情
+     */
+    DocumentChunkVO selectChunkById(@Param("chunkId") String chunkId);
 }

src/main/java/cn/yinlihupo/service/ai/AiKnowledgeBaseService.java

@@ -1,5 +1,6 @@
 package cn.yinlihupo.service.ai;
 
+import cn.yinlihupo.domain.vo.DocumentChunkVO;
 import cn.yinlihupo.domain.vo.KbDocumentVO;
 import org.springframework.web.multipart.MultipartFile;
 
@@ -57,4 +58,20 @@ public interface AiKnowledgeBaseService {
      * @param docId 文档ID
      */
     void processDocumentAsync(String docId);
+
+    /**
+     * 获取文档分片列表
+     *
+     * @param docId 文档ID
+     * @return 分片列表
+     */
+    List<DocumentChunkVO> getDocumentChunks(String docId);
+
+    /**
+     * 获取分片详情
+     *
+     * @param chunkId 分片ID
+     * @return 分片详情
+     */
+    DocumentChunkVO getChunkDetail(String chunkId);
 }

src/main/java/cn/yinlihupo/service/ai/impl/AiChatServiceImpl.java

@@ -67,24 +67,24 @@ public class AiChatServiceImpl implements AiChatService {
     @Override
     public void streamChat(ChatRequest request, Long userId, SseEmitter emitter) {
         long startTime = System.currentTimeMillis();
-        UUID sessionId = request.getSessionId();
+        String sessionId = request.getSessionId();
         boolean isNewSession = (sessionId == null);
 
         try {
             // 1. 获取或创建会话
             if (isNewSession) {
-                sessionId = UUID.randomUUID();
+                sessionId = UUID.randomUUID().toString();
                 String title = generateSessionTitle(request.getMessage());
                 createSession(userId, request.getProjectId(), request.getTimelineNodeId(), request.getMessage(), title);
             } else {
                 // 验证会话权限
-                if (!hasSessionAccess(sessionId, userId)) {
+                if (!hasSessionAccess(UUID.fromString(sessionId), userId)) {
                     sendError(emitter, "无权访问该会话");
                     return;
                 }
             }
 
-            final UUID finalSessionId = sessionId;
+            final String finalSessionId = sessionId;
 
             // 发送开始消息
             sendEvent(emitter, "start", Map.of(
@@ -107,7 +107,7 @@ public class AiChatServiceImpl implements AiChatService {
 
             // 4. 构建Prompt
             String systemPrompt = buildSystemPrompt(request.getProjectId(), retrievedDocs);
-            List<Message> messages = buildMessages(finalSessionId, request.getContextWindow(),
+            List<Message> messages = buildMessages(UUID.fromString(finalSessionId) , request.getContextWindow(),
                     systemPrompt, request.getMessage());
 
             // 5. 流式调用LLM
@@ -331,15 +331,15 @@ public class AiChatServiceImpl implements AiChatService {
     /**
      * 保存消息
      */
-    private Long saveMessage(UUID sessionId, Long userId, Long projectId,
+    private Long saveMessage(String sessionId, Long userId, Long projectId,
                              Long timelineNodeId, String role, String content,
                              String referencedDocIds) {
         // 获取当前最大序号
-        Integer maxIndex = chatHistoryMapper.selectMaxMessageIndex(sessionId);
+        Integer maxIndex = chatHistoryMapper.selectMaxMessageIndex(UUID.fromString(sessionId));
         int nextIndex = (maxIndex != null ? maxIndex : 0) + 1;
 
         AiChatMessage message = new AiChatMessage();
-        message.setSessionId(sessionId);
+        message.setSessionId(UUID.fromString(sessionId));
         message.setUserId(userId);
         message.setProjectId(projectId);
         message.setTimelineNodeId(timelineNodeId);

src/main/java/cn/yinlihupo/service/ai/impl/AiKnowledgeBaseServiceImpl.java

@@ -1,6 +1,7 @@
 package cn.yinlihupo.service.ai.impl;
 
 import cn.yinlihupo.domain.entity.AiDocument;
+import cn.yinlihupo.domain.vo.DocumentChunkVO;
 import cn.yinlihupo.domain.vo.KbDocumentVO;
 import cn.yinlihupo.mapper.AiDocumentMapper;
 import cn.yinlihupo.service.ai.AiKnowledgeBaseService;
@@ -209,4 +210,14 @@ public class AiKnowledgeBaseServiceImpl implements AiKnowledgeBaseService {
         vo.setCreateTime(doc.getCreateTime());
         return vo;
     }
+
+    @Override
+    public List<DocumentChunkVO> getDocumentChunks(String docId) {
+        return documentMapper.selectDocumentChunks(docId);
+    }
+
+    @Override
+    public DocumentChunkVO getChunkDetail(String chunkId) {
+        return documentMapper.selectChunkById(chunkId);
+    }
 }

src/main/java/cn/yinlihupo/service/ai/rag/DocumentProcessor.java

@@ -14,6 +14,7 @@ import org.springframework.stereotype.Component;
 
 import java.io.InputStream;
 import java.util.Collections;
+import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
 import java.util.UUID;
@@ -165,44 +166,51 @@ public class DocumentProcessor {
 
     /**
      * 存储切片到向量库
+     * 每个分片都是独立记录，包含完整的项目属性用于检索
      *
-     * @param parentDoc 父文档
+     * @param doc    文档实体（包含项目属性）
-     * @param chunks    切片列表
+     * @param chunks 切片列表
      */
-    private void storeChunks(AiDocument parentDoc, List<String> chunks) {
+    private void storeChunks(AiDocument doc, List<String> chunks) {
-        String parentId = parentDoc.getId();
+        String docId = doc.getId();
 
         for (int i = 0; i < chunks.size(); i++) {
             String chunkContent = chunks.get(i);
             // 使用UUID生成唯一的chunk ID，确保格式正确
             String chunkId = UUID.randomUUID().toString();
 
-            // 创建向量文档
+            // 创建向量文档，每个分片都包含完整的项目属性
-            Document vectorDoc = new Document(
+            Map<String, Object> metadata = new HashMap<>();
-                    chunkId,
+            // 项目关联属性（用于检索过滤）
-                    chunkContent,
+            metadata.put("project_id", doc.getProjectId() != null ? doc.getProjectId().toString() : "");
-                    Map.of(
+            metadata.put("timeline_node_id", doc.getTimelineNodeId() != null ? doc.getTimelineNodeId().toString() : "");
-                            "project_id", parentDoc.getProjectId() != null ? parentDoc.getProjectId().toString() : "",
+            metadata.put("kb_id", doc.getKbId() != null ? doc.getKbId().toString() : "");
-                            "timeline_node_id", parentDoc.getTimelineNodeId() != null ? parentDoc.getTimelineNodeId().toString() : "",
+            // 文档来源信息
-                            "chunk_index", i,
+            metadata.put("source_type", doc.getSourceType() != null ? doc.getSourceType() : "");
-                            "chunk_total", chunks.size(),
+            metadata.put("source_id", doc.getSourceId() != null ? doc.getSourceId().toString() : "");
-                            "chunk_parent_id", parentId,
+            // 文档信息
-                            "title", parentDoc.getTitle() != null ? parentDoc.getTitle() : "",
+            metadata.put("doc_id", docId);  // 原始文档ID，用于关联查询
-                            "source_type", parentDoc.getSourceType() != null ? parentDoc.getSourceType() : "",
+            metadata.put("title", doc.getTitle() != null ? doc.getTitle() : "");
-                            "status", "active"
+            metadata.put("doc_type", doc.getDocType() != null ? doc.getDocType() : "");
-                    )
+            metadata.put("file_type", doc.getFileType() != null ? doc.getFileType() : "");
-            );
+            // 分片信息
+            metadata.put("chunk_index", i);
+            metadata.put("chunk_total", chunks.size());
+            // 状态
+            metadata.put("status", "active");
+
+            Document vectorDoc = new Document(chunkId, chunkContent, metadata);
 
             // 存储到向量库
             vectorStore.add(List.of(vectorDoc));
 
-            // 如果是第一个切片，更新父文档内容
+            log.debug("存储切片: {}/{}, docId: {}, chunkId: {}", i + 1, chunks.size(), docId, chunkId);
-            if (i == 0) {
+        }
-                parentDoc.setContent(chunkContent);
-                documentMapper.updateById(parentDoc);
-            }
 
-            log.debug("存储切片: {}/{}, parentId: {}, chunkId: {}", i + 1, chunks.size(), parentId, chunkId);
+        // 更新文档内容为第一个分片（用于预览）
+        if (!chunks.isEmpty()) {
+            doc.setContent(chunks.get(0));
+            documentMapper.updateById(doc);
         }
     }
 

src/main/resources/mapper/AiDocumentMapper.xml

@@ -2,7 +2,7 @@
 <!DOCTYPE mapper PUBLIC "-//mybatis.org//DTD Mapper 3.0//EN" "http://mybatis.org/dtd/mybatis-3-mapper.dtd">
 <mapper namespace="cn.yinlihupo.mapper.AiDocumentMapper">
 
-    <!-- 获取项目文档列表 -->
+    <!-- 获取项目文档列表（只查询原始文档，不包含分片） -->
     <select id="selectProjectDocuments" resultType="cn.yinlihupo.domain.vo.KbDocumentVO">
         SELECT
             vs.id,
@@ -20,7 +20,7 @@
         LEFT JOIN sys_user su ON vs.create_by = su.id
         WHERE vs.project_id = #{projectId}
           AND vs.deleted = 0
-          AND vs.chunk_parent_id IS NULL
+          AND (vs.metadata->>'doc_id') IS NULL
         ORDER BY vs.create_time DESC
     </select>
 
@@ -58,11 +58,11 @@
           AND deleted = 0
     </select>
 
-    <!-- 获取父文档的分块数量 -->
+    <!-- 获取文档的分片数量 -->
     <select id="selectChunkCount" resultType="java.lang.Integer">
         SELECT COUNT(*)
         FROM vector_store
-        WHERE chunk_parent_id = #{docId}
+        WHERE metadata->>'doc_id' = #{docId}
           AND deleted = 0
     </select>
 
@@ -98,4 +98,40 @@
         WHERE id = #{id}
     </update>
 
+    <!-- 查询文档分片列表 -->
+    <select id="selectDocumentChunks" resultType="cn.yinlihupo.domain.vo.DocumentChunkVO">
+        SELECT
+            id,
+            metadata->>'doc_id' as docId,
+            content,
+            CAST(metadata->>'chunk_index' AS INTEGER) as chunkIndex,
+            CAST(metadata->>'chunk_total' AS INTEGER) as chunkTotal,
+            metadata->>'title' as title,
+            metadata->>'doc_type' as docType,
+            metadata->>'source_type' as sourceType,
+            metadata->>'status' as status
+        FROM vector_store
+        WHERE metadata->>'doc_id' = #{docId}
+          AND deleted = 0
+        ORDER BY CAST(metadata->>'chunk_index' AS INTEGER)
+    </select>
+
+    <!-- 查询分片详情 -->
+    <select id="selectChunkById" resultType="cn.yinlihupo.domain.vo.DocumentChunkVO">
+        SELECT
+            id,
+            metadata->>'doc_id' as docId,
+            content,
+            CAST(metadata->>'chunk_index' AS INTEGER) as chunkIndex,
+            CAST(metadata->>'chunk_total' AS INTEGER) as chunkTotal,
+            metadata->>'title' as title,
+            metadata->>'doc_type' as docType,
+            metadata->>'source_type' as sourceType,
+            metadata->>'status' as status
+        FROM vector_store
+        WHERE id = #{chunkId}
+          AND deleted = 0
+        LIMIT 1
+    </select>
+
 </mapper>