feat: add delete dataset index

api/core/index/base.py

@@ -37,6 +37,9 @@ class BaseIndex(ABC):
     ) -> List[Document]:
         raise NotImplementedError
 
+    def delete(self) -> None:
+        raise NotImplementedError
+
     def _filter_duplicate_texts(self, texts: list[Document]) -> list[Document]:
         for text in texts:
             doc_id = text.metadata['doc_id']

api/core/index/keyword_table_index/keyword_table_index.py

@@ -114,6 +114,12 @@ class KeywordTableIndex(BaseIndex):
 
         return documents
 
+    def delete(self) -> None:
+        dataset_keyword_table = self._dataset.dataset_keyword_table
+        if dataset_keyword_table:
+            db.session.delete(dataset_keyword_table)
+            db.session.commit()
+
     def _save_dataset_keyword_table(self, keyword_table):
         keyword_table_dict = {
             '__type__': 'keyword_table',

api/core/index/vector_index/base.py

@@ -90,3 +90,9 @@ class BaseVectorIndex(BaseIndex):
 
         for node_id in ids:
             vector_store.del_text(node_id)
+
+    def delete(self) -> None:
+        vector_store = self._get_vector_store()
+        vector_store = cast(self._get_vector_store_class(), vector_store)
+
+        vector_store.delete()

api/core/vector_store/qdrant_vector_store.py

@@ -39,6 +39,11 @@ class QdrantVectorStore(Qdrant):
 
         return len(response) > 0
 
+    def delete(self):
+        self._reload_if_needed()
+
+        self.client.delete_collection(collection_name=self.collection_name)
+
     def _reload_if_needed(self):
         if isinstance(self.client, QdrantLocal):
             self.client = cast(QdrantLocal, self.client)

api/core/vector_store/weaviate_vector_store.py

@@ -33,3 +33,6 @@ class WeaviateVectorStore(Weaviate):
             return False
 
         return True
+
+    def delete(self):
+        self._client.schema.delete_class(self._index_name)

api/tasks/clean_dataset_task.py

@@ -33,26 +33,21 @@ def clean_dataset_task(dataset_id: str, tenant_id: str, indexing_technique: str,
         )
 
         documents = db.session.query(DocumentSegment).filter(DocumentSegment.dataset_id == dataset_id).all()
-        index_doc_ids = [document.id for document in documents]
         segments = db.session.query(DocumentSegment).filter(DocumentSegment.dataset_id == dataset_id).all()
-        index_node_ids = [segment.index_node_id for segment in segments]
 
         vector_index = IndexBuilder.get_index(dataset, 'high_quality')
         kw_index = IndexBuilder.get_index(dataset, 'economy')
 
         # delete from vector index
         if vector_index:
-            for index_doc_id in index_doc_ids:
             try:
-                    vector_index.delete_by_document_id(index_doc_id)
+                vector_index.delete()
             except Exception:
                 logging.exception("Delete doc index failed when dataset deleted.")
-                    continue
 
         # delete from keyword index
-        if index_node_ids:
         try:
-                kw_index.delete_by_ids(index_node_ids)
+            kw_index.delete()
         except Exception:
             logging.exception("Delete nodes index failed when dataset deleted.")
 
@@ -62,7 +57,6 @@ def clean_dataset_task(dataset_id: str, tenant_id: str, indexing_technique: str,
         for segment in segments:
             db.session.delete(segment)
 
-        db.session.query(DatasetKeywordTable).filter(DatasetKeywordTable.dataset_id == dataset_id).delete()
         db.session.query(DatasetProcessRule).filter(DatasetProcessRule.dataset_id == dataset_id).delete()
         db.session.query(DatasetQuery).filter(DatasetQuery.dataset_id == dataset_id).delete()
         db.session.query(AppDatasetJoin).filter(AppDatasetJoin.dataset_id == dataset_id).delete()