优化了es索引分数，跟关键字的相关度更高，会随着关键字的多少进行匹配，增加当删除知识库文件时，该文件索引也会被删除

chat-client/src/views/datasets/components/DocumentList.vue

@@ -882,6 +882,7 @@ if (!newVal) {
 import { useUserStore } from '@/store/modules/user'
 import { useAclStore } from '@/store/modules/acl'
 import vab from '~/library/plugins/vab'
+import {log} from "echarts/types/src/util/log";
 
 // 事件处理函数
 const renderedHandler = () => {
@@ -1225,7 +1226,13 @@ try {
 
   // 调用删除接口
   await deleteDocument(parseInt(row.id))
-
+  try{
+    //删除es索引
+    await deleteFileIndex(row.difyDocId,datasetId.value)
+    console.log("删除es索引成功,文档id:{}",row.difyDocId)
+  } catch (error) {
+    console.error("删除es索引失败,文档id:{}",row.difyDocId,error)
+  }
 
   // 成功处理
   ElNotification({

chat-server/src/main/java/com/bjtds/brichat/controller/DatasetManageController.java

@@ -70,6 +70,14 @@ public class DatasetManageController {
         datasetManagerService.deleteDataset(id);
 
         difyDatasetService.delete(id);
+
+        //删除es索引
+        try {
+            esTDatasetFilesService.deleteIndex(id);
+            log.info("删除es索引成功,知识库id:{}",id);
+        } catch (IOException e) {
+            log.error("删除es索引失败,知识库id:{}",id,e);
+        }
     }
 
     @PostMapping("/create")

chat-server/src/main/java/com/bjtds/brichat/service/impl/EsTDatasetFilesServiceImpl.java

@@ -37,6 +37,7 @@ public class EsTDatasetFilesServiceImpl implements EsTDatasetFilesService {
     private ElasticsearchClient client;
     @Autowired
     private TUserDatasetMapper tUserDatasetMapper;
+
     @Autowired
     private DifyDatasetsMapper difyDatasetsMapper;
 
@@ -73,6 +74,7 @@ public class EsTDatasetFilesServiceImpl implements EsTDatasetFilesService {
                             .properties("source_url", p -> p.keyword(k -> k))
                             .properties("dataset_name", p -> p.keyword(k -> k))
                             .properties("document_id", p -> p.keyword(k -> k))
+                            .properties("difyDocId", p -> p.keyword(k -> k))
 
                     )
             );
@@ -263,12 +265,15 @@ public class EsTDatasetFilesServiceImpl implements EsTDatasetFilesService {
         double epsilon = 1e-6;
         double lower = 0.05, upper = 0.98;
 
+        Random random = new Random();
         int index = 0;
         for (Hit<TDatasetFiles> hit : response.hits().hits()) {
             TDatasetFiles d = hit.source();
             double rawScore = scores.get(index++);
-            double normalizedScore = (maxScore - minScore < epsilon) ? upper
+            log.info("Score: {}", rawScore);
-                    : lower + (rawScore - minScore) / (maxScore - minScore) * (upper - lower);
+            double normalizedScore = normalizeScore(rawScore, minScore, maxScore, lower, upper, epsilon, random);
+
+            // 高亮内容
 
             String content = String.join(" ... ", hit.highlight().getOrDefault("content", Collections.emptyList()));
             System.out.println("content: " + content);
@@ -340,11 +345,11 @@ public class EsTDatasetFilesServiceImpl implements EsTDatasetFilesService {
             try {
                 DeleteByQueryResponse response = client.deleteByQuery(d -> d
                         .index(DatasetId)
-                        .query(q -> q.term(t -> t.field("document_id").value(documentId))));
+                        .query(q -> q.term(t -> t.field("difyDocId").value(documentId))));
-                datasetFilesService.updateByDatasetIdAndDocId(DatasetId, documentId);
                 log.info("删除文档索引成功: {}, documentId: {}", DatasetId, documentId);
+                log.info(" deleted: {}", response.deleted());
             } catch (IOException e) {
-                log.error("删除文档索引失败: {}, documentId: {}", DatasetId, documentId, e);
+                log.error("删除文档索引失败: {}, documentId: {}, error: {}", DatasetId, documentId, e.getMessage());
             }
         } else {
             log.warn("索引不存在: {}", DatasetId);
@@ -365,6 +370,33 @@ public class EsTDatasetFilesServiceImpl implements EsTDatasetFilesService {
         });
     }
 
+    private double normalizeScore(double rawScore, double minScore, double maxScore, double lower, double upper, double epsilon, Random random) {
+        // 计算分数范围
+        double scoreRange = maxScore - minScore;
+        log.warn("Score range: {}", scoreRange);
+
+        // 如果最大最小分数相差小于 epsilon，直接使用 upper
+        double normalizedScore = (scoreRange < epsilon) ? upper
+                : lower + (rawScore - minScore) / scoreRange * (upper - lower);
+
+        log.warn("rawScore: {}, normalizedScore before fluctuation: {}", rawScore, normalizedScore);
+
+        // 获取原始分数的小数点后两位
+        double integerPart = Math.floor(rawScore);  // 获取整数部分
+        double decimalPart = rawScore - integerPart;  // 获取小数部分
+        double decimalPartOneDigit = Math.floor(decimalPart * 10) / 10.0;  // 获取小数部分的第一位
+        double result = integerPart + decimalPartOneDigit;  // 将整数部分和小数点后一位合成
+        double decimalPartTwoDigits = (result*10)/ 1000; // 获取小数部分的两位
+        log.warn("Raw score decimal part (2 digits): {}", decimalPartTwoDigits);
+        if(normalizedScore==upper){
+            normalizedScore -= (0.1-decimalPartTwoDigits);
+        }else if(normalizedScore==lower){
+            normalizedScore += decimalPartTwoDigits;
+        }
+        return normalizedScore;
+    }
+
+
 }