Merge 23c36453c3 into bd43ca6275

api/core/rag/datasource/vdb/vector_factory.py

@@ -28,152 +28,151 @@ class AbstractVectorFactory(ABC):
         index_struct_dict = {"type": vector_type, "vector_store": {"class_prefix": collection_name}}
         return index_struct_dict
 
+VECTOR_FACTORY_CLS: type[AbstractVectorFactory]
 
-class Vector:
+match dify_config.VECTOR_STORE:
-    def __init__(self, dataset: Dataset, attributes: Optional[list] = None):
-        if attributes is None:
-            attributes = ["doc_id", "dataset_id", "document_id", "doc_hash"]
-        self._dataset = dataset
-        self._embeddings = self._get_embeddings()
-        self._attributes = attributes
-        self._vector_processor = self._init_vector()
-
-    def _init_vector(self) -> BaseVector:
-        vector_type = dify_config.VECTOR_STORE
-
-        if self._dataset.index_struct_dict:
-            vector_type = self._dataset.index_struct_dict["type"]
-        else:
-            if dify_config.VECTOR_STORE_WHITELIST_ENABLE:
-                whitelist = (
-                    db.session.query(Whitelist)
-                    .where(Whitelist.tenant_id == self._dataset.tenant_id, Whitelist.category == "vector_db")
-                    .one_or_none()
-                )
-                if whitelist:
-                    vector_type = VectorType.TIDB_ON_QDRANT
-
-        if not vector_type:
-            raise ValueError("Vector store must be specified.")
-
-        vector_factory_cls = self.get_vector_factory(vector_type)
-        return vector_factory_cls().init_vector(self._dataset, self._attributes, self._embeddings)
-
-    @staticmethod
-    def get_vector_factory(vector_type: str) -> type[AbstractVectorFactory]:
-        match vector_type:
     case VectorType.CHROMA:
         from core.rag.datasource.vdb.chroma.chroma_vector import ChromaVectorFactory
 
-                return ChromaVectorFactory
+        VECTOR_FACTORY_CLS = ChromaVectorFactory
     case VectorType.MILVUS:
         from core.rag.datasource.vdb.milvus.milvus_vector import MilvusVectorFactory
 
-                return MilvusVectorFactory
+        VECTOR_FACTORY_CLS = MilvusVectorFactory
     case VectorType.MYSCALE:
         from core.rag.datasource.vdb.myscale.myscale_vector import MyScaleVectorFactory
 
-                return MyScaleVectorFactory
+        VECTOR_FACTORY_CLS = MyScaleVectorFactory
     case VectorType.PGVECTOR:
         from core.rag.datasource.vdb.pgvector.pgvector import PGVectorFactory
 
-                return PGVectorFactory
+        VECTOR_FACTORY_CLS = PGVectorFactory
     case VectorType.VASTBASE:
         from core.rag.datasource.vdb.pyvastbase.vastbase_vector import VastbaseVectorFactory
 
-                return VastbaseVectorFactory
+        VECTOR_FACTORY_CLS = VastbaseVectorFactory
     case VectorType.PGVECTO_RS:
         from core.rag.datasource.vdb.pgvecto_rs.pgvecto_rs import PGVectoRSFactory
 
-                return PGVectoRSFactory
+        VECTOR_FACTORY_CLS = PGVectoRSFactory
     case VectorType.QDRANT:
         from core.rag.datasource.vdb.qdrant.qdrant_vector import QdrantVectorFactory
 
-                return QdrantVectorFactory
+        VECTOR_FACTORY_CLS = QdrantVectorFactory
     case VectorType.RELYT:
         from core.rag.datasource.vdb.relyt.relyt_vector import RelytVectorFactory
 
-                return RelytVectorFactory
+        VECTOR_FACTORY_CLS = RelytVectorFactory
     case VectorType.ELASTICSEARCH:
         from core.rag.datasource.vdb.elasticsearch.elasticsearch_vector import ElasticSearchVectorFactory
 
-                return ElasticSearchVectorFactory
+        VECTOR_FACTORY_CLS = ElasticSearchVectorFactory
     case VectorType.ELASTICSEARCH_JA:
         from core.rag.datasource.vdb.elasticsearch.elasticsearch_ja_vector import (
             ElasticSearchJaVectorFactory,
         )
 
-                return ElasticSearchJaVectorFactory
+        VECTOR_FACTORY_CLS = ElasticSearchJaVectorFactory
     case VectorType.TIDB_VECTOR:
         from core.rag.datasource.vdb.tidb_vector.tidb_vector import TiDBVectorFactory
 
-                return TiDBVectorFactory
+        VECTOR_FACTORY_CLS = TiDBVectorFactory
     case VectorType.WEAVIATE:
         from core.rag.datasource.vdb.weaviate.weaviate_vector import WeaviateVectorFactory
 
-                return WeaviateVectorFactory
+        VECTOR_FACTORY_CLS = WeaviateVectorFactory
     case VectorType.TENCENT:
         from core.rag.datasource.vdb.tencent.tencent_vector import TencentVectorFactory
 
-                return TencentVectorFactory
+        VECTOR_FACTORY_CLS = TencentVectorFactory
     case VectorType.ORACLE:
         from core.rag.datasource.vdb.oracle.oraclevector import OracleVectorFactory
 
-                return OracleVectorFactory
+        VECTOR_FACTORY_CLS = OracleVectorFactory
     case VectorType.OPENSEARCH:
         from core.rag.datasource.vdb.opensearch.opensearch_vector import OpenSearchVectorFactory
 
-                return OpenSearchVectorFactory
+        VECTOR_FACTORY_CLS = OpenSearchVectorFactory
     case VectorType.ANALYTICDB:
         from core.rag.datasource.vdb.analyticdb.analyticdb_vector import AnalyticdbVectorFactory
 
-                return AnalyticdbVectorFactory
+        VECTOR_FACTORY_CLS = AnalyticdbVectorFactory
     case VectorType.COUCHBASE:
         from core.rag.datasource.vdb.couchbase.couchbase_vector import CouchbaseVectorFactory
 
-                return CouchbaseVectorFactory
+        VECTOR_FACTORY_CLS = CouchbaseVectorFactory
     case VectorType.BAIDU:
         from core.rag.datasource.vdb.baidu.baidu_vector import BaiduVectorFactory
 
-                return BaiduVectorFactory
+        VECTOR_FACTORY_CLS = BaiduVectorFactory
     case VectorType.VIKINGDB:
         from core.rag.datasource.vdb.vikingdb.vikingdb_vector import VikingDBVectorFactory
 
-                return VikingDBVectorFactory
+        VECTOR_FACTORY_CLS = VikingDBVectorFactory
     case VectorType.UPSTASH:
         from core.rag.datasource.vdb.upstash.upstash_vector import UpstashVectorFactory
 
-                return UpstashVectorFactory
+        VECTOR_FACTORY_CLS = UpstashVectorFactory
     case VectorType.TIDB_ON_QDRANT:
         from core.rag.datasource.vdb.tidb_on_qdrant.tidb_on_qdrant_vector import TidbOnQdrantVectorFactory
 
-                return TidbOnQdrantVectorFactory
+        VECTOR_FACTORY_CLS = TidbOnQdrantVectorFactory
     case VectorType.LINDORM:
         from core.rag.datasource.vdb.lindorm.lindorm_vector import LindormVectorStoreFactory
 
-                return LindormVectorStoreFactory
+        VECTOR_FACTORY_CLS = LindormVectorStoreFactory
     case VectorType.OCEANBASE:
         from core.rag.datasource.vdb.oceanbase.oceanbase_vector import OceanBaseVectorFactory
 
-                return OceanBaseVectorFactory
+        VECTOR_FACTORY_CLS = OceanBaseVectorFactory
     case VectorType.OPENGAUSS:
         from core.rag.datasource.vdb.opengauss.opengauss import OpenGaussFactory
 
-                return OpenGaussFactory
+        VECTOR_FACTORY_CLS = OpenGaussFactory
     case VectorType.TABLESTORE:
         from core.rag.datasource.vdb.tablestore.tablestore_vector import TableStoreVectorFactory
 
-                return TableStoreVectorFactory
+        VECTOR_FACTORY_CLS = TableStoreVectorFactory
     case VectorType.HUAWEI_CLOUD:
         from core.rag.datasource.vdb.huawei.huawei_cloud_vector import HuaweiCloudVectorFactory
 
-                return HuaweiCloudVectorFactory
+        VECTOR_FACTORY_CLS = HuaweiCloudVectorFactory
     case VectorType.MATRIXONE:
         from core.rag.datasource.vdb.matrixone.matrixone_vector import MatrixoneVectorFactory
 
-                return MatrixoneVectorFactory
+        VECTOR_FACTORY_CLS = MatrixoneVectorFactory
     case _:
-                raise ValueError(f"Vector store {vector_type} is not supported.")
+        raise ValueError(f"Vector store {dify_config.VECTOR_STORE} is not supported.")
+
+
+class Vector:
+    def __init__(self, dataset: Dataset, attributes: Optional[list] = None):
+        if attributes is None:
+            attributes = ["doc_id", "dataset_id", "document_id", "doc_hash"]
+        self._dataset = dataset
+        self._embeddings = self._get_embeddings()
+        self._attributes = attributes
+        self._vector_processor = self._init_vector()
+
+    def _init_vector(self) -> BaseVector:
+        vector_type = dify_config.VECTOR_STORE
+
+        if self._dataset.index_struct_dict:
+            vector_type = self._dataset.index_struct_dict["type"]
+        else:
+            if dify_config.VECTOR_STORE_WHITELIST_ENABLE:
+                whitelist = (
+                    db.session.query(Whitelist)
+                    .where(Whitelist.tenant_id == self._dataset.tenant_id, Whitelist.category == "vector_db")
+                    .one_or_none()
+                )
+                if whitelist:
+                    vector_type = VectorType.TIDB_ON_QDRANT
+
+        if not vector_type:
+            raise ValueError("Vector store must be specified.")
+
+        return VECTOR_FACTORY_CLS().init_vector(self._dataset, self._attributes, self._embeddings)
 
     def create(self, texts: Optional[list] = None, **kwargs):
         if texts: