【代码评审】AI 大模型：知识库的逻辑

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/controller/admin/knowledge/vo/AiKnowledgeCreateMyReqVO.java

@@ -7,9 +7,6 @@ import lombok.Data;
 
 import java.util.List;
 
-/**
- * @author xiaoxin
- */
 @Schema(description = "管理后台 - AI 知识库创建【我的】 Request VO")
 @Data
 public class AiKnowledgeCreateMyReqVO {
@@ -21,10 +18,10 @@ public class AiKnowledgeCreateMyReqVO {
     @Schema(description = "知识库描述", requiredMode = Schema.RequiredMode.REQUIRED, example = "存储 ruoyi-vue-pro 操作文档")
     private String description;
 
-    @Schema(description = "可见权限,只能选择哪些人可见", requiredMode = Schema.RequiredMode.REQUIRED, example = "[1]")
+    @Schema(description = "可见权限，只能选择哪些人可见", requiredMode = Schema.RequiredMode.REQUIRED, example = "[1]")
     private List<Long> visibilityPermissions;
 
-    @Schema(description = "嵌入模型 id", requiredMode = Schema.RequiredMode.REQUIRED, example = "1")
+    @Schema(description = "嵌入模型编号", requiredMode = Schema.RequiredMode.REQUIRED, example = "1")
     @NotNull(message = "嵌入模型不能为空")
     private Long modelId;
 

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/controller/admin/knowledge/vo/AiKnowledgeDocumentCreateReqVO.java

@@ -4,6 +4,7 @@ import io.swagger.v3.oas.annotations.media.Schema;
 import jakarta.validation.constraints.NotBlank;
 import jakarta.validation.constraints.NotNull;
 import lombok.Data;
+import org.hibernate.validator.constraints.URL;
 
 /**
  * @author xiaoxin
@@ -12,7 +13,6 @@ import lombok.Data;
 @Data
 public class AiKnowledgeDocumentCreateReqVO {
 
-
     @Schema(description = "知识库编号", requiredMode = Schema.RequiredMode.REQUIRED, example = "1204")
     @NotNull(message = "知识库编号不能为空")
     private Long knowledgeId;
@@ -21,7 +21,8 @@ public class AiKnowledgeDocumentCreateReqVO {
     @NotBlank(message = "文档名称不能为空")
     private String name;
 
-    @Schema(description = "文档 url", requiredMode = Schema.RequiredMode.REQUIRED, example = "https://doc.iocoder.cn")
+    @Schema(description = "文档 URL", requiredMode = Schema.RequiredMode.REQUIRED, example = "https://doc.iocoder.cn")
+    @URL(message = "文档 URL 格式不正确")
     private String url;
 
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/controller/admin/knowledge/vo/AiKnowledgeUpdateMyReqVO.java

@@ -7,14 +7,10 @@ import lombok.Data;
 
 import java.util.List;
 
-/**
- * @author xiaoxin
- */
 @Schema(description = "管理后台 - AI 知识库创建【我的】 Request VO")
 @Data
 public class AiKnowledgeUpdateMyReqVO {
 
-
     @Schema(description = "对话编号", requiredMode = Schema.RequiredMode.REQUIRED, example = "1204")
     @NotNull(message = "知识库编号不能为空")
     private Long id;
@@ -26,10 +22,10 @@ public class AiKnowledgeUpdateMyReqVO {
     @Schema(description = "知识库描述", requiredMode = Schema.RequiredMode.REQUIRED, example = "")
     private String description;
 
-    @Schema(description = "可见权限,只能选择哪些人可见", requiredMode = Schema.RequiredMode.REQUIRED, example = "[1]")
+    @Schema(description = "可见权限，只能选择哪些人可见", requiredMode = Schema.RequiredMode.REQUIRED, example = "[1]")
     private List<Long> visibilityPermissions;
 
-    @Schema(description = "嵌入模型 id", requiredMode = Schema.RequiredMode.REQUIRED, example = "1")
+    @Schema(description = "嵌入模型编号", requiredMode = Schema.RequiredMode.REQUIRED, example = "1")
     @NotNull(message = "嵌入模型不能为空")
     private Long modelId;
 

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/image/AiImageDO.java

@@ -32,7 +32,7 @@ public class AiImageDO extends BaseDO {
     /**
      * 编号
      */
-    @TableId(type = IdType.AUTO)
+    @TableId
     private Long id;
 
     /**

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/knowledge/AiKnowledgeBaseDO.java

@@ -1,9 +1,7 @@
 package cn.iocoder.yudao.module.ai.dal.dataobject.knowledge;
 
-
 import cn.iocoder.yudao.framework.common.enums.CommonStatusEnum;
 import cn.iocoder.yudao.framework.mybatis.core.dataobject.BaseDO;
-import com.baomidou.mybatisplus.annotation.IdType;
 import com.baomidou.mybatisplus.annotation.TableField;
 import com.baomidou.mybatisplus.annotation.TableId;
 import com.baomidou.mybatisplus.annotation.TableName;
@@ -12,19 +10,20 @@ import lombok.Data;
 
 import java.util.List;
 
+// TODO @xin：要不把 AiKnowledgeBaseDO 改成 AiKnowledgeDO。感觉 base 后缀，感觉有点奇怪（让人以为是基类）。然后，我们很多地方的外键编号，都是 knowledgeId
 /**
  * AI 知识库 DO
  *
  * @author xiaoxin
  */
-@TableName(value = "ai_knowledge_base")
+@TableName(value = "ai_knowledge_base", autoResultMap = true)
 @Data
 public class AiKnowledgeBaseDO extends BaseDO {
 
     /**
      * 编号
      */
-    @TableId(type = IdType.AUTO)
+    @TableId
     private Long id;
     /**
      * 用户编号
@@ -40,6 +39,7 @@ public class AiKnowledgeBaseDO extends BaseDO {
      * 知识库描述
      */
     private String description;
+    // TODO @新：如果全部可见，需要怎么设置？
     /**
      * 可见权限,只能选择哪些人可见
      */

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/knowledge/AiKnowledgeDocumentDO.java

@@ -3,7 +3,6 @@ package cn.iocoder.yudao.module.ai.dal.dataobject.knowledge;
 import cn.iocoder.yudao.framework.common.enums.CommonStatusEnum;
 import cn.iocoder.yudao.framework.mybatis.core.dataobject.BaseDO;
 import cn.iocoder.yudao.module.ai.enums.knowledge.AiKnowledgeDocumentStatusEnum;
-import com.baomidou.mybatisplus.annotation.IdType;
 import com.baomidou.mybatisplus.annotation.TableId;
 import com.baomidou.mybatisplus.annotation.TableName;
 import lombok.Data;
@@ -20,10 +19,12 @@ public class AiKnowledgeDocumentDO extends BaseDO {
     /**
      * 编号
      */
-    @TableId(type = IdType.AUTO)
+    @TableId
     private Long id;
     /**
      * 知识库编号
+     *
+     * 关联 {@link AiKnowledgeBaseDO#getId()}
      */
     private Long knowledgeId;
     /**
@@ -59,4 +60,5 @@ public class AiKnowledgeDocumentDO extends BaseDO {
      * 枚举 {@link CommonStatusEnum}
      */
     private Integer status;
+
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/knowledge/AiKnowledgeSegmentDO.java

@@ -2,7 +2,6 @@ package cn.iocoder.yudao.module.ai.dal.dataobject.knowledge;
 
 import cn.iocoder.yudao.framework.common.enums.CommonStatusEnum;
 import cn.iocoder.yudao.framework.mybatis.core.dataobject.BaseDO;
-import com.baomidou.mybatisplus.annotation.IdType;
 import com.baomidou.mybatisplus.annotation.TableId;
 import com.baomidou.mybatisplus.annotation.TableName;
 import lombok.Data;
@@ -19,14 +18,17 @@ public class AiKnowledgeSegmentDO extends BaseDO {
     /**
      * 编号
      */
-    @TableId(type = IdType.AUTO)
+    @TableId
     private Long id;
     /**
-     * 向量库的id
+     * 向量库的编号
      */
     private String vectorId;
+    // TODO @新：knowledgeId 加个，会方便点
     /**
      * 文档编号
+     *
+     * 关联 {@link AiKnowledgeDocumentDO#getId()}
      */
     private Long documentId;
     /**

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/mindmap/AiMindMapDO.java

@@ -19,7 +19,7 @@ public class AiMindMapDO extends BaseDO {
     /**
      * 编号
      */
-    @TableId(type = IdType.AUTO)
+    @TableId
     private Long id;
 
     /**

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/music/AiMusicDO.java

@@ -25,7 +25,7 @@ public class AiMusicDO extends BaseDO {
     /**
      * 编号
      */
-    @TableId(type = IdType.AUTO)
+    @TableId
     private Long id;
 
     /**

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/dal/dataobject/write/AiWriteDO.java

@@ -20,7 +20,7 @@ public class AiWriteDO extends BaseDO {
     /**
      * 编号
      */
-    @TableId(type = IdType.AUTO)
+    @TableId
     private Long id;
 
     /**

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiEmbeddingService.java

@@ -17,11 +17,11 @@ public interface AiEmbeddingService {
      */
     void add(List<Document> documents);
 
-
     /**
      * 相似查询
      *
      * @param request 查询实体
      */
     List<Document> similaritySearch(SearchRequest request);
+
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiEmbeddingServiceImpl.java

@@ -8,6 +8,7 @@ import org.springframework.stereotype.Service;
 
 import java.util.List;
 
+// TODO @xin：是不是不用 AiEmbeddingServiceImpl，直接 vectorStore 注入到需要的地方就好啦。通过 KnowledgeDocumentService 返回就好。
 /**
  * AI 嵌入 Service 实现类
  *
@@ -30,4 +31,5 @@ public class AiEmbeddingServiceImpl implements AiEmbeddingService {
     public List<Document> similaritySearch(SearchRequest request) {
         return vectorStore.similaritySearch(request);
     }
+
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiKnowledgeBaseService.java

@@ -9,7 +9,6 @@ import cn.iocoder.yudao.module.ai.controller.admin.knowledge.vo.AiKnowledgeUpdat
  */
 public interface AiKnowledgeBaseService {
 
-
     /**
      * 创建【我的】知识库
      *
@@ -27,4 +26,5 @@ public interface AiKnowledgeBaseService {
      * @param userId 用户编号
      */
     void updateKnowledgeMy(AiKnowledgeUpdateMyReqVO updateReqVO, Long userId);
+
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiKnowledgeBaseServiceImpl.java

@@ -32,33 +32,36 @@ public class AiKnowledgeBaseServiceImpl implements AiKnowledgeBaseService {
     @Resource
     private AiKnowledgeBaseMapper knowledgeBaseMapper;
 
-
     @Override
     public Long createKnowledgeMy(AiKnowledgeCreateMyReqVO createReqVO, Long userId) {
+        // TODO @xin：貌似直接调用 chatModalService.validateChatModel(id) 完事，不用搞个方法
+        // 1. 校验模型配置
         AiChatModelDO model = validateChatModel(createReqVO.getModelId());
 
-        AiKnowledgeBaseDO knowledgeBaseDO = BeanUtils.toBean(createReqVO, AiKnowledgeBaseDO.class);
+        // 2. 插入知识库
-        knowledgeBaseDO.setModel(model.getModel()).setUserId(userId).setStatus(CommonStatusEnum.ENABLE.getStatus());
+        // TODO @xin：不用 DO 结尾
-
+        AiKnowledgeBaseDO knowledgeBaseDO = BeanUtils.toBean(createReqVO, AiKnowledgeBaseDO.class)
+                .setModel(model.getModel()).setUserId(userId).setStatus(CommonStatusEnum.ENABLE.getStatus());
         knowledgeBaseMapper.insert(knowledgeBaseDO);
         return knowledgeBaseDO.getId();
     }
 
     @Override
     public void updateKnowledgeMy(AiKnowledgeUpdateMyReqVO updateReqVO, Long userId) {
-
+        // 1.1 校验知识库存在
         AiKnowledgeBaseDO knowledgeBaseDO = validateKnowledgeExists(updateReqVO.getId());
         if (ObjUtil.notEqual(knowledgeBaseDO.getUserId(), userId)) {
             throw exception(KNOWLEDGE_NOT_EXISTS);
         }
+        // 1.2 校验模型配置
         AiChatModelDO model = validateChatModel(updateReqVO.getModelId());
+
+        // 2. 更新知识库
         AiKnowledgeBaseDO updateDO = BeanUtils.toBean(updateReqVO, AiKnowledgeBaseDO.class);
         updateDO.setModel(model.getModel());
-
         knowledgeBaseMapper.updateById(updateDO);
     }
 
-
     private AiChatModelDO validateChatModel(Long id) {
         AiChatModelDO model = chatModalService.validateChatModel(id);
         Assert.notNull(model, "未找到对应嵌入模型");
@@ -72,4 +75,5 @@ public class AiKnowledgeBaseServiceImpl implements AiKnowledgeBaseService {
         }
         return knowledgeBase;
     }
+
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiKnowledgeDocumentService.java

@@ -9,7 +9,6 @@ import cn.iocoder.yudao.module.ai.controller.admin.knowledge.vo.AiKnowledgeDocum
  */
 public interface AiKnowledgeDocumentService {
 
-
     /**
      * 创建文档
      *

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiKnowledgeDocumentServiceImpl.java

@@ -43,28 +43,30 @@ public class AiKnowledgeDocumentServiceImpl implements AiKnowledgeDocumentServic
     @Resource
     private AiEmbeddingService embeddingService;
 
+    // TODO @xin：@Resource 注入
     private static final JTokkitTokenCountEstimator TOKEN_COUNT_ESTIMATOR = new JTokkitTokenCountEstimator();
 
     // TODO xiaoxin 临时测试用，后续删
     @Value("classpath:/webapp/test/Fel.pdf")
     private org.springframework.core.io.Resource data;
 
-
+    // TODO 芋艿：需要 review 下，代码格式；
+    // TODO @xin：最好有 1、/2、/3 这种，让代码更有层次感
     @Override
     @Transactional(rollbackFor = Exception.class)
     public Long createKnowledgeDocument(AiKnowledgeDocumentCreateReqVO createReqVO) {
-
         // TODO xiaoxin 后续从 url 加载
         TikaDocumentReader loader = new TikaDocumentReader(data);
         // 加载文档
         List<Document> documents = loader.get();
         Document document = CollUtil.getFirst(documents);
-        // TODO 芋艿 文档层面有没有可能会比较大，这两个字段是否可以从分段表计算得出？
+        // TODO @xin：是不是不存在，就抛出异常呀；厚泽 return 呀；
+        // TODO 芋艿 文档层面有没有可能会比较大，这两个字段是否可以从分段表计算得出？回复：先直接算；
         Integer tokens = Objects.nonNull(document) ? TOKEN_COUNT_ESTIMATOR.estimate(document.getContent()) : 0;
         Integer wordCount = Objects.nonNull(document) ? document.getContent().length() : 0;
 
-        AiKnowledgeDocumentDO documentDO = BeanUtils.toBean(createReqVO, AiKnowledgeDocumentDO.class);
+        AiKnowledgeDocumentDO documentDO = BeanUtils.toBean(createReqVO, AiKnowledgeDocumentDO.class)
-        documentDO.setTokens(tokens).setWordCount(wordCount)
+                .setTokens(tokens).setWordCount(wordCount)
                 .setStatus(CommonStatusEnum.ENABLE.getStatus()).setSliceStatus(AiKnowledgeDocumentStatusEnum.SUCCESS.getStatus());
         // 文档记录入库
         documentMapper.insert(documentDO);
@@ -75,17 +77,15 @@ public class AiKnowledgeDocumentServiceImpl implements AiKnowledgeDocumentServic
 
         // 文档分段
         List<Document> segments = tokenTextSplitter.apply(documents);
-
+        // 分段内容入库
         List<AiKnowledgeSegmentDO> segmentDOList = CollectionUtils.convertList(segments,
                 segment -> new AiKnowledgeSegmentDO().setContent(segment.getContent()).setDocumentId(documentId)
                         .setTokens(TOKEN_COUNT_ESTIMATOR.estimate(segment.getContent())).setWordCount(segment.getContent().length())
                         .setStatus(CommonStatusEnum.ENABLE.getStatus()));
-        // 分段内容入库
         segmentMapper.insertBatch(segmentDOList);
-
         // 向量化并存储
         embeddingService.add(segments);
-
         return documentId;
     }
+
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiKnowledgeSegmentService.java

@@ -1,11 +1,10 @@
 package cn.iocoder.yudao.module.ai.service.knowledge;
 
 /**
- * AI 知识库-分片 Service 接口
+ * AI 知识库分片 Service 接口
  *
  * @author xiaoxin
  */
 public interface AiKnowledgeSegmentService {
 
-
 }

yudao-module-ai/yudao-module-ai-biz/src/main/java/cn/iocoder/yudao/module/ai/service/knowledge/AiKnowledgeSegmentServiceImpl.java

@@ -4,7 +4,7 @@ import lombok.extern.slf4j.Slf4j;
 import org.springframework.stereotype.Service;
 
 /**
- * AI 知识库-基础信息 Service 实现类
+ * AI 知识库分片 Service 实现类
  *
  * @author xiaoxin
  */
@@ -12,5 +12,4 @@ import org.springframework.stereotype.Service;
 @Slf4j
 public class AiKnowledgeSegmentServiceImpl implements AiKnowledgeSegmentService {
 
-
 }