diff --git a/apps/application/serializers/application_serializers.py b/apps/application/serializers/application_serializers.py
index 64ff19580..189b70d48 100644
--- a/apps/application/serializers/application_serializers.py
+++ b/apps/application/serializers/application_serializers.py
@@ -533,6 +533,7 @@ class ApplicationSerializer(serializers.Serializer):
             QuerySet(Application).filter(id=self.data.get('application_id')).delete()
             return True
 
+        @transaction.atomic
         def publish(self, instance, with_valid=True):
             if with_valid:
                 self.is_valid()
@@ -549,9 +550,12 @@ class ApplicationSerializer(serializers.Serializer):
                     application.desc = node_data.get('desc')
                     application.prologue = node_data.get('prologue')
             dataset_list = self.list_dataset(with_valid=False)
-            self.update_reverse_search_node(work_flow, [str(dataset.get('id')) for dataset in dataset_list])
+            dataset_id_list = self.update_reverse_search_node(work_flow,
+                                                              [str(dataset.get('id')) for dataset in dataset_list])
             application.work_flow = work_flow
             application.save()
+            # 插入知识库关联关系
+            self.save_application_mapping(dataset_id_list, application.id)
             work_flow_version = WorkFlowVersion(work_flow=work_flow, application=application)
             work_flow_version.save()
             return True
@@ -585,6 +589,7 @@ class ApplicationSerializer(serializers.Serializer):
 
         def update_reverse_search_node(self, work_flow, user_dataset_id_list: List):
             search_node_list = self.get_search_node(work_flow)
+            dataset_id_list = []
             for search_node in search_node_list:
                 node_data = search_node.get('properties', {}).get('node_data', {})
                 dataset_id_list = node_data.get('dataset_id_list', [])
@@ -598,6 +603,8 @@ class ApplicationSerializer(serializers.Serializer):
                 source_dataset_id_list = list({*source_dataset_id_list, *dataset_id_list})
                 node_data['source_dataset_id_list'] = []
                 node_data['dataset_id_list'] = source_dataset_id_list
+                dataset_id_list = [*source_dataset_id_list, *dataset_id_list]
+            return list(set(dataset_id_list))
 
         def profile(self, with_valid=True):
             if with_valid:
@@ -611,6 +618,7 @@ class ApplicationSerializer(serializers.Serializer):
                 {**ApplicationSerializer.ApplicationModel(application).data,
                  'show_source': application_access_token.show_source})
 
+        @transaction.atomic
         def edit(self, instance: Dict, with_valid=True):
             if with_valid:
                 self.is_valid()
@@ -653,16 +661,20 @@ class ApplicationSerializer(serializers.Serializer):
                     if not application_dataset_id_list.__contains__(dataset_id):
                         raise AppApiException(500, f"未知的知识库id${dataset_id},无法关联")
 
-                # 删除已经关联的id
-                QuerySet(ApplicationDatasetMapping).filter(dataset_id__in=application_dataset_id_list,
-                                                           application_id=application_id).delete()
-                # 插入
-                QuerySet(ApplicationDatasetMapping).bulk_create(
-                    [ApplicationDatasetMapping(application_id=application_id, dataset_id=dataset_id) for dataset_id in
-                     dataset_id_list]) if len(dataset_id_list) > 0 else None
+                self.save_application_mapping(application_dataset_id_list, application_id)
             chat_cache.clear_by_application_id(application_id)
             return self.one(with_valid=False)
 
+        @staticmethod
+        def save_application_mapping(dataset_id_list, application_id):
+            # 删除已经关联的id
+            QuerySet(ApplicationDatasetMapping).filter(dataset_id__in=dataset_id_list,
+                                                       application_id=application_id).delete()
+            # 插入
+            QuerySet(ApplicationDatasetMapping).bulk_create(
+                [ApplicationDatasetMapping(application_id=application_id, dataset_id=dataset_id) for dataset_id in
+                 dataset_id_list]) if len(dataset_id_list) > 0 else None
+
         def list_dataset(self, with_valid=True):
             if with_valid:
                 self.is_valid(raise_exception=True)