Skip to content

armgit / dify

Public Imported from GitHub
0

Build Agentic workflows, RAG pipelines, with rich AI model and tool support on one collaborative workspace. Deploy on cloud, VPC, or self-hosted, so teams move from prototype to production without rebuilding the stack.

fix: use caller session for Qdrant annotation bindings (#43543)

Byron.wang8023189566da
diff --git a/api/core/rag/datasource/vdb/vector_factory.py b/api/core/rag/datasource/vdb/vector_factory.py
index 9bd9eed6245..4ff29a7d980 100644
--- a/api/core/rag/datasource/vdb/vector_factory.py
+++ b/api/core/rag/datasource/vdb/vector_factory.py
@@ -29,7 +29,16 @@ logger = logging.getLogger(__name__)
class AbstractVectorFactory(ABC):
@abstractmethod
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> BaseVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> BaseVector:
+ """Initialize the backend using the caller's session for metadata reads.
+
+ ``None`` is used by callers that initialize a resolved vector backend
+ outside their database transaction; backend-specific legacy reads may
+ still use their own session in that case. A supplied session belongs to
+ the caller and must not be committed, rolled back, or closed here.
+ """
raise NotImplementedError
@staticmethod
@@ -142,7 +151,7 @@ class Vector:
self._session = session
if vector_type is not None:
self._vector_processor = self.get_vector_factory(vector_type)().init_vector(
- dataset, self._attributes, self._embeddings
+ dataset, self._attributes, self._embeddings, session=session
)
else:
if session is None:
@@ -172,7 +181,7 @@ class Vector:
def _init_vector(self, *, session: Session) -> BaseVector:
vector_type = self.resolve_vector_type(self._dataset, session=session)
vector_factory_cls = self.get_vector_factory(vector_type)
- return vector_factory_cls().init_vector(self._dataset, self._attributes, self._embeddings)
+ return vector_factory_cls().init_vector(self._dataset, self._attributes, self._embeddings, session=session)
@staticmethod
def get_vector_factory(vector_type: str) -> type[AbstractVectorFactory]:
diff --git a/api/providers/vdb/vdb-alibabacloud-mysql/src/dify_vdb_alibabacloud_mysql/alibabacloud_mysql_vector.py b/api/providers/vdb/vdb-alibabacloud-mysql/src/dify_vdb_alibabacloud_mysql/alibabacloud_mysql_vector.py
index c6fe8988d13..ddce5abc620 100644
--- a/api/providers/vdb/vdb-alibabacloud-mysql/src/dify_vdb_alibabacloud_mysql/alibabacloud_mysql_vector.py
+++ b/api/providers/vdb/vdb-alibabacloud-mysql/src/dify_vdb_alibabacloud_mysql/alibabacloud_mysql_vector.py
@@ -8,6 +8,7 @@ from typing import Any, Literal, cast, override
import mysql.connector
from mysql.connector import Error as MySQLError
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.field import parse_metadata_json
@@ -365,7 +366,9 @@ class AlibabaCloudMySQLVectorFactory(AbstractVectorFactory):
return cast(Literal["cosine", "euclidean"], distance_function)
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> AlibabaCloudMySQLVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> AlibabaCloudMySQLVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-alibabacloud-mysql/tests/unit_tests/test_alibabacloud_mysql_factory.py b/api/providers/vdb/vdb-alibabacloud-mysql/tests/unit_tests/test_alibabacloud_mysql_factory.py
index ba18801f82b..51902fd7445 100644
--- a/api/providers/vdb/vdb-alibabacloud-mysql/tests/unit_tests/test_alibabacloud_mysql_factory.py
+++ b/api/providers/vdb/vdb-alibabacloud-mysql/tests/unit_tests/test_alibabacloud_mysql_factory.py
@@ -39,7 +39,7 @@ def test_factory_init_vector_uses_existing_index_struct_class_prefix(monkeypatch
monkeypatch.setattr(alibaba_module.dify_config, "ALIBABACLOUD_MYSQL_HNSW_M", 6)
with patch.object(alibaba_module, "AlibabaCloudMySQLVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
assert result == "vector"
assert vector_cls.call_args.kwargs["collection_name"] == "existing_collection"
@@ -61,7 +61,7 @@ def test_factory_init_vector_generates_collection_name_when_index_struct_is_miss
monkeypatch.setattr(alibaba_module.dify_config, "ALIBABACLOUD_MYSQL_HNSW_M", 12)
with patch.object(alibaba_module, "AlibabaCloudMySQLVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
assert result == "vector"
vector_cls.assert_called_once()
diff --git a/api/providers/vdb/vdb-analyticdb/src/dify_vdb_analyticdb/analyticdb_vector.py b/api/providers/vdb/vdb-analyticdb/src/dify_vdb_analyticdb/analyticdb_vector.py
index 62d8819fe60..0f8cf538940 100644
--- a/api/providers/vdb/vdb-analyticdb/src/dify_vdb_analyticdb/analyticdb_vector.py
+++ b/api/providers/vdb/vdb-analyticdb/src/dify_vdb_analyticdb/analyticdb_vector.py
@@ -1,6 +1,8 @@
import json
from typing import Any, override
+from sqlalchemy.orm import Session
+
from configs import dify_config
from core.rag.datasource.vdb.vector_base import BaseVector
from core.rag.datasource.vdb.vector_factory import AbstractVectorFactory
@@ -74,7 +76,9 @@ class AnalyticdbVector(BaseVector):
class AnalyticdbVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> AnalyticdbVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> AnalyticdbVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-analyticdb/tests/unit_tests/test_analyticdb_vector.py b/api/providers/vdb/vdb-analyticdb/tests/unit_tests/test_analyticdb_vector.py
index 57ca12f4082..a78c2dcc6a5 100644
--- a/api/providers/vdb/vdb-analyticdb/tests/unit_tests/test_analyticdb_vector.py
+++ b/api/providers/vdb/vdb-analyticdb/tests/unit_tests/test_analyticdb_vector.py
@@ -100,7 +100,7 @@ def test_factory_builds_openapi_config_when_host_is_missing(monkeypatch: pytest.
monkeypatch.setattr(analyticdb_module.dify_config, "ANALYTICDB_NAMESPACE_PASSWORD", "ns-password")
with patch.object(analyticdb_module, "AnalyticdbVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
assert result == "vector"
args = vector_cls.call_args.args
@@ -123,7 +123,7 @@ def test_factory_builds_sql_config_when_host_is_present(monkeypatch: pytest.Monk
monkeypatch.setattr(analyticdb_module.dify_config, "ANALYTICDB_NAMESPACE", "dify")
with patch.object(analyticdb_module, "AnalyticdbVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
assert result == "vector"
args = vector_cls.call_args.args
diff --git a/api/providers/vdb/vdb-baidu/src/dify_vdb_baidu/baidu_vector.py b/api/providers/vdb/vdb-baidu/src/dify_vdb_baidu/baidu_vector.py
index 2d8caccc9b6..3a1a638b979 100644
--- a/api/providers/vdb/vdb-baidu/src/dify_vdb_baidu/baidu_vector.py
+++ b/api/providers/vdb/vdb-baidu/src/dify_vdb_baidu/baidu_vector.py
@@ -26,6 +26,7 @@ from pymochow.model.schema import (
VectorIndex,
) # type: ignore
from pymochow.model.table import AnnSearch, BM25SearchRequest, HNSWSearchParams, Partition, Row # type: ignore
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.field import Field as VDBField
@@ -378,7 +379,9 @@ class BaiduVector(BaseVector):
class BaiduVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> BaiduVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> BaiduVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-baidu/tests/unit_tests/test_baidu_vector.py b/api/providers/vdb/vdb-baidu/tests/unit_tests/test_baidu_vector.py
index 796fda94b2e..ae97e0180ee 100644
--- a/api/providers/vdb/vdb-baidu/tests/unit_tests/test_baidu_vector.py
+++ b/api/providers/vdb/vdb-baidu/tests/unit_tests/test_baidu_vector.py
@@ -274,7 +274,7 @@ def test_factory_initializes_collection_name_and_index_struct(baidu_module, monk
monkeypatch.setattr(baidu_module.dify_config, "BAIDU_VECTOR_DB_REBUILD_INDEX_TIMEOUT_IN_SECONDS", 300)
with patch.object(baidu_module, "BaiduVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
assert result == "vector"
assert vector_cls.call_args.kwargs["collection_name"] == "auto_collection"
@@ -547,7 +547,7 @@ def test_factory_uses_existing_collection_prefix_when_index_struct_exists(
monkeypatch.setattr(baidu_module.dify_config, "BAIDU_VECTOR_DB_REBUILD_INDEX_TIMEOUT_IN_SECONDS", 300)
with patch.object(baidu_module, "BaiduVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
assert result == "vector"
assert vector_cls.call_args.kwargs["collection_name"] == "existing_collection"
diff --git a/api/providers/vdb/vdb-chroma/src/dify_vdb_chroma/chroma_vector.py b/api/providers/vdb/vdb-chroma/src/dify_vdb_chroma/chroma_vector.py
index e92ac94afa8..869add5991c 100644
--- a/api/providers/vdb/vdb-chroma/src/dify_vdb_chroma/chroma_vector.py
+++ b/api/providers/vdb/vdb-chroma/src/dify_vdb_chroma/chroma_vector.py
@@ -4,6 +4,7 @@ from typing import Any, TypedDict, override
import chromadb
from chromadb import QueryResult, Settings
from pydantic import BaseModel
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.vector_base import BaseVector, VectorIndexStructDict
@@ -158,7 +159,9 @@ class ChromaVector(BaseVector):
class ChromaVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> BaseVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> BaseVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-chroma/tests/unit_tests/test_chroma_vector.py b/api/providers/vdb/vdb-chroma/tests/unit_tests/test_chroma_vector.py
index b58002795c0..d8a3852acad 100644
--- a/api/providers/vdb/vdb-chroma/tests/unit_tests/test_chroma_vector.py
+++ b/api/providers/vdb/vdb-chroma/tests/unit_tests/test_chroma_vector.py
@@ -190,8 +190,8 @@ def test_factory_init_vector_uses_existing_or_generated_collection(chroma_module
monkeypatch.setattr(chroma_module.dify_config, "CHROMA_AUTH_CREDENTIALS", None)
with patch.object(chroma_module, "ChromaVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-clickzetta/src/dify_vdb_clickzetta/clickzetta_vector.py b/api/providers/vdb/vdb-clickzetta/src/dify_vdb_clickzetta/clickzetta_vector.py
index 6e80c6efa2c..3882fa25624 100644
--- a/api/providers/vdb/vdb-clickzetta/src/dify_vdb_clickzetta/clickzetta_vector.py
+++ b/api/providers/vdb/vdb-clickzetta/src/dify_vdb_clickzetta/clickzetta_vector.py
@@ -12,6 +12,7 @@ from typing import TYPE_CHECKING, Any, override
import clickzetta # type: ignore
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
if TYPE_CHECKING:
from clickzetta.connector.v0.connection import Connection # type: ignore
@@ -1069,7 +1070,9 @@ class ClickzettaVectorFactory(AbstractVectorFactory):
"""Factory for creating Clickzetta vector instances."""
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> BaseVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> BaseVector:
"""Initialize a Clickzetta vector instance."""
# Get configuration from environment variables or dataset config
config = ClickzettaConfig(
diff --git a/api/providers/vdb/vdb-clickzetta/tests/unit_tests/test_clickzetta_vector.py b/api/providers/vdb/vdb-clickzetta/tests/unit_tests/test_clickzetta_vector.py
index be476d9175e..60c58789dc8 100644
--- a/api/providers/vdb/vdb-clickzetta/tests/unit_tests/test_clickzetta_vector.py
+++ b/api/providers/vdb/vdb-clickzetta/tests/unit_tests/test_clickzetta_vector.py
@@ -238,7 +238,7 @@ def test_factory_initializes_clickzetta_vector(clickzetta_module, monkeypatch: p
monkeypatch.setattr(clickzetta_module.dify_config, "CLICKZETTA_VECTOR_DISTANCE_FUNCTION", "cosine_distance")
with patch.object(clickzetta_module, "ClickzettaVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
assert result == "vector"
assert vector_cls.call_args.kwargs["collection_name"] == "collection"
diff --git a/api/providers/vdb/vdb-couchbase/src/dify_vdb_couchbase/couchbase_vector.py b/api/providers/vdb/vdb-couchbase/src/dify_vdb_couchbase/couchbase_vector.py
index 7ac7edc6a0c..c4818af3386 100644
--- a/api/providers/vdb/vdb-couchbase/src/dify_vdb_couchbase/couchbase_vector.py
+++ b/api/providers/vdb/vdb-couchbase/src/dify_vdb_couchbase/couchbase_vector.py
@@ -15,6 +15,7 @@ from couchbase.options import ClusterOptions, SearchOptions # type: ignore
from couchbase.vector_search import VectorQuery, VectorSearch # type: ignore
from flask import current_app
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from core.rag.datasource.vdb.vector_base import BaseVector
from core.rag.datasource.vdb.vector_factory import AbstractVectorFactory
@@ -366,7 +367,9 @@ class CouchbaseVector(BaseVector):
class CouchbaseVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> CouchbaseVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> CouchbaseVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-couchbase/tests/unit_tests/test_couchbase_vector.py b/api/providers/vdb/vdb-couchbase/tests/unit_tests/test_couchbase_vector.py
index f05cabc129f..e91004a6fab 100644
--- a/api/providers/vdb/vdb-couchbase/tests/unit_tests/test_couchbase_vector.py
+++ b/api/providers/vdb/vdb-couchbase/tests/unit_tests/test_couchbase_vector.py
@@ -357,8 +357,8 @@ def test_delete_collection_and_factory(couchbase_module, monkeypatch: pytest.Mon
)
with patch.object(couchbase_module, "CouchbaseVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_ja_vector.py b/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_ja_vector.py
index 49da9db987a..001a30237b0 100644
--- a/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_ja_vector.py
+++ b/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_ja_vector.py
@@ -3,6 +3,7 @@ import logging
from typing import Any, override
from flask import current_app
+from sqlalchemy.orm import Session
from core.rag.datasource.vdb.field import Field
from core.rag.datasource.vdb.vector_type import VectorType
@@ -87,7 +88,9 @@ class ElasticSearchJaVector(ElasticSearchVector):
class ElasticSearchJaVectorFactory(ElasticSearchVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> ElasticSearchJaVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> ElasticSearchJaVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_vector.py b/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_vector.py
index b83cddb057a..8ca4bf52eed 100644
--- a/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_vector.py
+++ b/api/providers/vdb/vdb-elasticsearch/src/dify_vdb_elasticsearch/elasticsearch_vector.py
@@ -9,6 +9,7 @@ from elasticsearch import Elasticsearch
from flask import current_app
from packaging.version import parse as parse_version
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from core.rag.datasource.vdb.field import Field
from core.rag.datasource.vdb.vector_base import BaseVector
@@ -307,7 +308,9 @@ class ElasticSearchVector(BaseVector):
class ElasticSearchVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> ElasticSearchVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> ElasticSearchVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_ja_vector.py b/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_ja_vector.py
index 16a4e363bfb..7e0c9850925 100644
--- a/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_ja_vector.py
+++ b/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_ja_vector.py
@@ -112,8 +112,8 @@ def test_ja_factory_uses_existing_or_generated_collection(elasticsearch_ja_modul
)
with patch.object(elasticsearch_ja_module, "ElasticSearchJaVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_vector.py b/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_vector.py
index 8953928a5c9..80e85f75b71 100644
--- a/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_vector.py
+++ b/api/providers/vdb/vdb-elasticsearch/tests/unit_tests/test_elasticsearch_vector.py
@@ -358,7 +358,7 @@ def test_elasticsearch_factory_branches(elasticsearch_module, monkeypatch: pytes
)
with patch.object(elasticsearch_module, "ElasticSearchVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
cfg = vector_cls.call_args.kwargs["config"]
assert cfg.use_cloud is False
@@ -377,7 +377,7 @@ def test_elasticsearch_factory_branches(elasticsearch_module, monkeypatch: pytes
),
)
with patch.object(elasticsearch_module, "ElasticSearchVector", return_value="vector") as vector_cls:
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_2 == "vector"
cfg = vector_cls.call_args.kwargs["config"]
assert cfg.use_cloud is True
@@ -399,7 +399,7 @@ def test_elasticsearch_factory_branches(elasticsearch_module, monkeypatch: pytes
),
)
with patch.object(elasticsearch_module, "ElasticSearchVector", return_value="vector") as vector_cls:
- factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
cfg = vector_cls.call_args.kwargs["config"]
assert cfg.use_cloud is False
assert cfg.host == "fallback-host"
diff --git a/api/providers/vdb/vdb-hologres/src/dify_vdb_hologres/hologres_vector.py b/api/providers/vdb/vdb-hologres/src/dify_vdb_hologres/hologres_vector.py
index abaacec35ad..3929259965e 100644
--- a/api/providers/vdb/vdb-hologres/src/dify_vdb_hologres/hologres_vector.py
+++ b/api/providers/vdb/vdb-hologres/src/dify_vdb_hologres/hologres_vector.py
@@ -7,6 +7,7 @@ import holo_search_sdk as holo # type: ignore
from holo_search_sdk.types import BaseQuantizationType, DistanceType, TokenizerType
from psycopg import sql as psql
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.field import parse_metadata_json
@@ -344,7 +345,9 @@ class HologresVectorFactory(AbstractVectorFactory):
"""Factory class for creating HologresVector instances."""
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> HologresVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> HologresVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-hologres/tests/unit_tests/test_hologres_vector.py b/api/providers/vdb/vdb-hologres/tests/unit_tests/test_hologres_vector.py
index eff8b4184f2..b899e269749 100644
--- a/api/providers/vdb/vdb-hologres/tests/unit_tests/test_hologres_vector.py
+++ b/api/providers/vdb/vdb-hologres/tests/unit_tests/test_hologres_vector.py
@@ -353,8 +353,8 @@ def test_hologres_factory_uses_existing_or_generated_collection(hologres_module,
monkeypatch.setattr(hologres_module.dify_config, "HOLOGRES_EF_CONSTRUCTION", 400)
with patch.object(hologres_module, "HologresVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-huawei-cloud/src/dify_vdb_huawei_cloud/huawei_cloud_vector.py b/api/providers/vdb/vdb-huawei-cloud/src/dify_vdb_huawei_cloud/huawei_cloud_vector.py
index 6f42ce1dbe6..89053e40064 100644
--- a/api/providers/vdb/vdb-huawei-cloud/src/dify_vdb_huawei_cloud/huawei_cloud_vector.py
+++ b/api/providers/vdb/vdb-huawei-cloud/src/dify_vdb_huawei_cloud/huawei_cloud_vector.py
@@ -5,6 +5,7 @@ from typing import Any, override
from elasticsearch import Elasticsearch
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from typing_extensions import TypedDict
from configs import dify_config
@@ -217,7 +218,9 @@ class HuaweiCloudVector(BaseVector):
class HuaweiCloudVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> HuaweiCloudVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> HuaweiCloudVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-huawei-cloud/tests/unit_tests/test_huawei_cloud_vector.py b/api/providers/vdb/vdb-huawei-cloud/tests/unit_tests/test_huawei_cloud_vector.py
index b057a6cb348..ed55d924487 100644
--- a/api/providers/vdb/vdb-huawei-cloud/tests/unit_tests/test_huawei_cloud_vector.py
+++ b/api/providers/vdb/vdb-huawei-cloud/tests/unit_tests/test_huawei_cloud_vector.py
@@ -233,8 +233,8 @@ def test_huawei_factory_branches(huawei_module, monkeypatch: pytest.MonkeyPatch)
monkeypatch.setattr(huawei_module.dify_config, "HUAWEI_CLOUD_PASSWORD", "pass")
with patch.object(huawei_module, "HuaweiCloudVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-iris/src/dify_vdb_iris/iris_vector.py b/api/providers/vdb/vdb-iris/src/dify_vdb_iris/iris_vector.py
index 0326cc646a4..19a41260336 100644
--- a/api/providers/vdb/vdb-iris/src/dify_vdb_iris/iris_vector.py
+++ b/api/providers/vdb/vdb-iris/src/dify_vdb_iris/iris_vector.py
@@ -13,6 +13,8 @@ import uuid
from contextlib import contextmanager
from typing import TYPE_CHECKING, Any, override
+from sqlalchemy.orm import Session
+
from configs import dify_config
from configs.middleware.vdb.iris_config import IrisVectorConfig
from core.rag.datasource.vdb.field import parse_metadata_json
@@ -491,7 +493,9 @@ class IrisVectorFactory(AbstractVectorFactory):
"""Factory for creating IrisVector instances."""
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> IrisVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> IrisVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-iris/tests/unit_tests/test_iris_vector.py b/api/providers/vdb/vdb-iris/tests/unit_tests/test_iris_vector.py
index 2364626d5eb..e3dd1f70f11 100644
--- a/api/providers/vdb/vdb-iris/tests/unit_tests/test_iris_vector.py
+++ b/api/providers/vdb/vdb-iris/tests/unit_tests/test_iris_vector.py
@@ -402,8 +402,8 @@ def test_iris_vector_delete_create_collection_and_factory(iris_module, monkeypat
monkeypatch.setattr(iris_module.dify_config, "IRIS_TEXT_INDEX_LANGUAGE", "en")
with patch.object(iris_module, "IrisVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-lindorm/src/dify_vdb_lindorm/lindorm_vector.py b/api/providers/vdb/vdb-lindorm/src/dify_vdb_lindorm/lindorm_vector.py
index 8c3292c6b1f..ab191c9b1e8 100644
--- a/api/providers/vdb/vdb-lindorm/src/dify_vdb_lindorm/lindorm_vector.py
+++ b/api/providers/vdb/vdb-lindorm/src/dify_vdb_lindorm/lindorm_vector.py
@@ -6,6 +6,7 @@ from typing import Any, override
from opensearchpy import OpenSearch, helpers
from opensearchpy.helpers import BulkIndexError
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from tenacity import retry, stop_after_attempt, wait_exponential
from typing_extensions import TypedDict
@@ -388,7 +389,9 @@ class LindormVectorStore(BaseVector):
class LindormVectorStoreFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> LindormVectorStore:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> LindormVectorStore:
lindorm_config = LindormVectorStoreConfig(
hosts=dify_config.LINDORM_URL,
username=dify_config.LINDORM_USERNAME,
diff --git a/api/providers/vdb/vdb-lindorm/tests/unit_tests/test_lindorm_vector.py b/api/providers/vdb/vdb-lindorm/tests/unit_tests/test_lindorm_vector.py
index 3ec045036da..261f0668097 100644
--- a/api/providers/vdb/vdb-lindorm/tests/unit_tests/test_lindorm_vector.py
+++ b/api/providers/vdb/vdb-lindorm/tests/unit_tests/test_lindorm_vector.py
@@ -361,7 +361,7 @@ def test_lindorm_factory_branches(lindorm_module, monkeypatch: pytest.MonkeyPatc
monkeypatch.setattr(lindorm_module.dify_config, "LINDORM_USING_UGC", None)
with pytest.raises(ValueError, match="LINDORM_USING_UGC is not set"):
- factory.init_vector(dataset, attributes=[], embeddings=embeddings)
+ factory.init_vector(dataset, attributes=[], embeddings=embeddings, session=None)
monkeypatch.setattr(lindorm_module.dify_config, "LINDORM_USING_UGC", False)
@@ -369,7 +369,7 @@ def test_lindorm_factory_branches(lindorm_module, monkeypatch: pytest.MonkeyPatc
id="dataset-1", index_struct=json.dumps({"vector_store": {"class_prefix": "EXISTING"}, "using_ugc": False})
)
with patch.object(lindorm_module, "LindormVectorStore", return_value="vector") as store_cls:
- result = factory.init_vector(dataset_existing_plain, attributes=[], embeddings=embeddings)
+ result = factory.init_vector(dataset_existing_plain, attributes=[], embeddings=embeddings, session=None)
assert result == "vector"
assert store_cls.call_args.args[0] == "existing"
@@ -386,7 +386,7 @@ def test_lindorm_factory_branches(lindorm_module, monkeypatch: pytest.MonkeyPatc
),
)
with patch.object(lindorm_module, "LindormVectorStore", return_value="vector") as store_cls:
- factory.init_vector(dataset_existing_ugc, attributes=[], embeddings=embeddings)
+ factory.init_vector(dataset_existing_ugc, attributes=[], embeddings=embeddings, session=None)
assert store_cls.call_args.args[0] == "ugc_index_1536_hnsw_l2"
assert store_cls.call_args.kwargs["routing_value"] == "ROUTING"
@@ -394,7 +394,7 @@ def test_lindorm_factory_branches(lindorm_module, monkeypatch: pytest.MonkeyPatc
monkeypatch.setattr(lindorm_module.dify_config, "LINDORM_USING_UGC", True)
with patch.object(lindorm_module, "LindormVectorStore", return_value="vector") as store_cls:
- factory.init_vector(dataset_new, attributes=[], embeddings=embeddings)
+ factory.init_vector(dataset_new, attributes=[], embeddings=embeddings, session=None)
assert store_cls.call_args.args[0] == "ugc_index_3_hnsw_l2"
assert store_cls.call_args.kwargs["routing_value"] == "auto_collection"
assert dataset_new.index_struct is not None
@@ -402,6 +402,6 @@ def test_lindorm_factory_branches(lindorm_module, monkeypatch: pytest.MonkeyPatc
dataset_new_plain = Dataset(id="dataset-3")
monkeypatch.setattr(lindorm_module.dify_config, "LINDORM_USING_UGC", False)
with patch.object(lindorm_module, "LindormVectorStore", return_value="vector") as store_cls:
- factory.init_vector(dataset_new_plain, attributes=[], embeddings=embeddings)
+ factory.init_vector(dataset_new_plain, attributes=[], embeddings=embeddings, session=None)
assert store_cls.call_args.args[0] == "auto_collection"
assert store_cls.call_args.kwargs["routing_value"] is None
diff --git a/api/providers/vdb/vdb-matrixone/src/dify_vdb_matrixone/matrixone_vector.py b/api/providers/vdb/vdb-matrixone/src/dify_vdb_matrixone/matrixone_vector.py
index 7a51776ad26..37ae212aef8 100644
--- a/api/providers/vdb/vdb-matrixone/src/dify_vdb_matrixone/matrixone_vector.py
+++ b/api/providers/vdb/vdb-matrixone/src/dify_vdb_matrixone/matrixone_vector.py
@@ -7,6 +7,7 @@ from typing import Any, Concatenate, override
from mo_vector.client import MoVectorClient # type: ignore
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.field import parse_metadata_json
@@ -228,7 +229,9 @@ class MatrixoneVector(BaseVector):
class MatrixoneVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> MatrixoneVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> MatrixoneVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-matrixone/tests/unit_tests/test_matrixone_vector.py b/api/providers/vdb/vdb-matrixone/tests/unit_tests/test_matrixone_vector.py
index 40ef49387bd..059aabbe176 100644
--- a/api/providers/vdb/vdb-matrixone/tests/unit_tests/test_matrixone_vector.py
+++ b/api/providers/vdb/vdb-matrixone/tests/unit_tests/test_matrixone_vector.py
@@ -246,8 +246,8 @@ def test_matrixone_factory_uses_existing_or_generated_collection(matrixone_modul
monkeypatch.setattr(matrixone_module.dify_config, "MATRIXONE_METRIC", "l2")
with patch.object(matrixone_module, "MatrixoneVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-milvus/src/dify_vdb_milvus/milvus_vector.py b/api/providers/vdb/vdb-milvus/src/dify_vdb_milvus/milvus_vector.py
index 18113c6831b..488e331028d 100644
--- a/api/providers/vdb/vdb-milvus/src/dify_vdb_milvus/milvus_vector.py
+++ b/api/providers/vdb/vdb-milvus/src/dify_vdb_milvus/milvus_vector.py
@@ -6,6 +6,7 @@ from packaging import version
from pydantic import BaseModel, model_validator
from pymilvus import MilvusClient, MilvusException # type: ignore
from pymilvus.milvus_client import IndexParams # type: ignore
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.field import Field
@@ -422,7 +423,9 @@ class MilvusVectorFactory(AbstractVectorFactory):
"""
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> MilvusVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> MilvusVector:
"""
Initialize a MilvusVector instance for the given dataset.
"""
diff --git a/api/providers/vdb/vdb-milvus/tests/unit_tests/test_milvus.py b/api/providers/vdb/vdb-milvus/tests/unit_tests/test_milvus.py
index 6dc05ca8b80..467e27e7245 100644
--- a/api/providers/vdb/vdb-milvus/tests/unit_tests/test_milvus.py
+++ b/api/providers/vdb/vdb-milvus/tests/unit_tests/test_milvus.py
@@ -436,8 +436,8 @@ def test_factory_initializes_milvus_vector(milvus_module, monkeypatch: pytest.Mo
monkeypatch.setattr(milvus_module.dify_config, "MILVUS_ANALYZER_PARAMS", '{"tokenizer":"standard"}')
with patch.object(milvus_module, "MilvusVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-myscale/src/dify_vdb_myscale/myscale_vector.py b/api/providers/vdb/vdb-myscale/src/dify_vdb_myscale/myscale_vector.py
index c43459ea44f..4f2cfa871bd 100644
--- a/api/providers/vdb/vdb-myscale/src/dify_vdb_myscale/myscale_vector.py
+++ b/api/providers/vdb/vdb-myscale/src/dify_vdb_myscale/myscale_vector.py
@@ -7,6 +7,7 @@ from typing import Any, override
from clickhouse_connect import get_client # type: ignore[import-untyped]
from pydantic import BaseModel
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.vector_base import BaseVector
@@ -196,7 +197,9 @@ class MyScaleVector(BaseVector):
class MyScaleVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> MyScaleVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> MyScaleVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-myscale/tests/unit_tests/test_myscale_vector.py b/api/providers/vdb/vdb-myscale/tests/unit_tests/test_myscale_vector.py
index 28acb640073..081a0ddca91 100644
--- a/api/providers/vdb/vdb-myscale/tests/unit_tests/test_myscale_vector.py
+++ b/api/providers/vdb/vdb-myscale/tests/unit_tests/test_myscale_vector.py
@@ -108,8 +108,8 @@ def test_factory_initializes_lower_case_collection_name(myscale_module, monkeypa
monkeypatch.setattr(myscale_module.dify_config, "MYSCALE_FTS_PARAMS", "")
with patch.object(myscale_module, "MyScaleVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-oceanbase/src/dify_vdb_oceanbase/oceanbase_vector.py b/api/providers/vdb/vdb-oceanbase/src/dify_vdb_oceanbase/oceanbase_vector.py
index e986951308a..7c041e74251 100644
--- a/api/providers/vdb/vdb-oceanbase/src/dify_vdb_oceanbase/oceanbase_vector.py
+++ b/api/providers/vdb/vdb-oceanbase/src/dify_vdb_oceanbase/oceanbase_vector.py
@@ -8,6 +8,7 @@ from pyobvector import VECTOR, ObVecClient, cosine_distance, inner_product, l2_d
from sqlalchemy import JSON, Column, String
from sqlalchemy.dialects.mysql import LONGTEXT
from sqlalchemy.exc import SQLAlchemyError
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.field import parse_metadata_json
@@ -564,6 +565,8 @@ class OceanBaseVectorFactory(AbstractVectorFactory):
dataset: Dataset,
attributes: list,
embeddings: Embeddings,
+ *,
+ session: Session | None,
) -> BaseVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
diff --git a/api/providers/vdb/vdb-oceanbase/tests/unit_tests/test_oceanbase_vector.py b/api/providers/vdb/vdb-oceanbase/tests/unit_tests/test_oceanbase_vector.py
index c2c41ad9f72..dd648bd65e2 100644
--- a/api/providers/vdb/vdb-oceanbase/tests/unit_tests/test_oceanbase_vector.py
+++ b/api/providers/vdb/vdb-oceanbase/tests/unit_tests/test_oceanbase_vector.py
@@ -591,8 +591,8 @@ def test_oceanbase_factory_uses_existing_or_generated_collection(oceanbase_modul
monkeypatch.setattr(oceanbase_module.dify_config, "OCEANBASE_HNSW_REFRESH_THRESHOLD", 1000)
with patch.object(oceanbase_module, "OceanBaseVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-opengauss/src/dify_vdb_opengauss/opengauss.py b/api/providers/vdb/vdb-opengauss/src/dify_vdb_opengauss/opengauss.py
index 70bf5f2a62b..0f206bcd0ff 100644
--- a/api/providers/vdb/vdb-opengauss/src/dify_vdb_opengauss/opengauss.py
+++ b/api/providers/vdb/vdb-opengauss/src/dify_vdb_opengauss/opengauss.py
@@ -6,6 +6,7 @@ from typing import Any, override
import psycopg2.extras
import psycopg2.pool
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.vector_base import BaseVector
@@ -250,7 +251,9 @@ class OpenGauss(BaseVector):
class OpenGaussFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> OpenGauss:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> OpenGauss:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-opengauss/tests/unit_tests/test_opengauss.py b/api/providers/vdb/vdb-opengauss/tests/unit_tests/test_opengauss.py
index 8836232040b..85d967152c7 100644
--- a/api/providers/vdb/vdb-opengauss/tests/unit_tests/test_opengauss.py
+++ b/api/providers/vdb/vdb-opengauss/tests/unit_tests/test_opengauss.py
@@ -390,8 +390,8 @@ def test_opengauss_factory_uses_existing_or_generated_collection(opengauss_modul
monkeypatch.setattr(opengauss_module.dify_config, "OPENGAUSS_ENABLE_PQ", False)
with patch.object(opengauss_module, "OpenGauss", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-opensearch/src/dify_vdb_opensearch/opensearch_vector.py b/api/providers/vdb/vdb-opensearch/src/dify_vdb_opensearch/opensearch_vector.py
index c145e5479a2..8f1f06d6e99 100644
--- a/api/providers/vdb/vdb-opensearch/src/dify_vdb_opensearch/opensearch_vector.py
+++ b/api/providers/vdb/vdb-opensearch/src/dify_vdb_opensearch/opensearch_vector.py
@@ -6,6 +6,7 @@ from uuid import uuid4
from opensearchpy import OpenSearch, Urllib3AWSV4SignerAuth, Urllib3HttpConnection, helpers
from opensearchpy.helpers import BulkIndexError
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from typing_extensions import TypedDict
from configs import dify_config
@@ -310,7 +311,9 @@ class OpenSearchVector(BaseVector):
class OpenSearchVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> OpenSearchVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> OpenSearchVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-opensearch/tests/unit_tests/test_opensearch_vector.py b/api/providers/vdb/vdb-opensearch/tests/unit_tests/test_opensearch_vector.py
index 3622bf77639..f7ebb39c6ac 100644
--- a/api/providers/vdb/vdb-opensearch/tests/unit_tests/test_opensearch_vector.py
+++ b/api/providers/vdb/vdb-opensearch/tests/unit_tests/test_opensearch_vector.py
@@ -352,8 +352,8 @@ def test_opensearch_factory_initializes_expected_collection_name(opensearch_modu
monkeypatch.setattr(opensearch_module.dify_config, "OPENSEARCH_AWS_SERVICE", None)
with patch.object(opensearch_module, "OpenSearchVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-oracle/src/dify_vdb_oracle/oraclevector.py b/api/providers/vdb/vdb-oracle/src/dify_vdb_oracle/oraclevector.py
index 831ccd32009..38f3f74f839 100644
--- a/api/providers/vdb/vdb-oracle/src/dify_vdb_oracle/oraclevector.py
+++ b/api/providers/vdb/vdb-oracle/src/dify_vdb_oracle/oraclevector.py
@@ -10,6 +10,7 @@ import numpy
import oracledb
from oracledb.connection import Connection
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.vector_base import BaseVector
@@ -383,7 +384,9 @@ class OracleVector(BaseVector):
class OracleVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> OracleVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> OracleVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-oracle/tests/unit_tests/test_oraclevector.py b/api/providers/vdb/vdb-oracle/tests/unit_tests/test_oraclevector.py
index 97af40b646e..cb5a7e40850 100644
--- a/api/providers/vdb/vdb-oracle/tests/unit_tests/test_oraclevector.py
+++ b/api/providers/vdb/vdb-oracle/tests/unit_tests/test_oraclevector.py
@@ -367,8 +367,8 @@ def test_oracle_factory_init_vector_uses_existing_or_generated_collection(
monkeypatch.setattr(oracle_module.dify_config, "ORACLE_IS_AUTONOMOUS", False)
with patch.object(oracle_module, "OracleVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-pgvecto-rs/src/dify_vdb_pgvecto_rs/pgvecto_rs.py b/api/providers/vdb/vdb-pgvecto-rs/src/dify_vdb_pgvecto_rs/pgvecto_rs.py
index 006dcd11851..31adc2aa658 100644
--- a/api/providers/vdb/vdb-pgvecto-rs/src/dify_vdb_pgvecto_rs/pgvecto_rs.py
+++ b/api/providers/vdb/vdb-pgvecto-rs/src/dify_vdb_pgvecto_rs/pgvecto_rs.py
@@ -217,7 +217,9 @@ class PGVectoRS(BaseVector):
class PGVectoRSFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> PGVectoRS:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> PGVectoRS:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-pgvecto-rs/tests/unit_tests/test_pgvecto_rs.py b/api/providers/vdb/vdb-pgvecto-rs/tests/unit_tests/test_pgvecto_rs.py
index 2f04f24e7fc..a20ed5b6528 100644
--- a/api/providers/vdb/vdb-pgvecto-rs/tests/unit_tests/test_pgvecto_rs.py
+++ b/api/providers/vdb/vdb-pgvecto-rs/tests/unit_tests/test_pgvecto_rs.py
@@ -334,8 +334,8 @@ def test_factory_uses_existing_or_generated_collection(pgvecto_module, monkeypat
embeddings.embed_query.return_value = [0.1, 0.2, 0.3]
with patch.object(module, "PGVectoRS", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=embeddings)
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=embeddings)
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=embeddings, session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=embeddings, session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-pgvector/src/dify_vdb_pgvector/pgvector.py b/api/providers/vdb/vdb-pgvector/src/dify_vdb_pgvector/pgvector.py
index 244b288f918..f779dba90d7 100644
--- a/api/providers/vdb/vdb-pgvector/src/dify_vdb_pgvector/pgvector.py
+++ b/api/providers/vdb/vdb-pgvector/src/dify_vdb_pgvector/pgvector.py
@@ -9,6 +9,7 @@ import psycopg2.errors
import psycopg2.extras
import psycopg2.pool
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.vector_base import BaseVector
@@ -280,7 +281,9 @@ class PGVector(BaseVector):
class PGVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> PGVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> PGVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-pgvector/tests/unit_tests/test_pgvector.py b/api/providers/vdb/vdb-pgvector/tests/unit_tests/test_pgvector.py
index b6505285398..eb3a6580888 100644
--- a/api/providers/vdb/vdb-pgvector/tests/unit_tests/test_pgvector.py
+++ b/api/providers/vdb/vdb-pgvector/tests/unit_tests/test_pgvector.py
@@ -484,8 +484,8 @@ def test_pgvector_factory_initializes_expected_collection_name(monkeypatch: pyte
monkeypatch.setattr(pgvector_module.dify_config, "PGVECTOR_PG_BIGM", False)
with patch.object(pgvector_module, "PGVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-qdrant/src/dify_vdb_qdrant/qdrant_vector.py b/api/providers/vdb/vdb-qdrant/src/dify_vdb_qdrant/qdrant_vector.py
index d52a3d15544..305e06970b6 100644
--- a/api/providers/vdb/vdb-qdrant/src/dify_vdb_qdrant/qdrant_vector.py
+++ b/api/providers/vdb/vdb-qdrant/src/dify_vdb_qdrant/qdrant_vector.py
@@ -19,6 +19,7 @@ from qdrant_client.http.models import (
)
from qdrant_client.local.qdrant_local import QdrantLocal
from sqlalchemy import select
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.field import Field
@@ -497,10 +498,15 @@ class QdrantVector(BaseVector):
class QdrantVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> QdrantVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> QdrantVector:
if dataset.collection_binding_id:
stmt = select(DatasetCollectionBinding).where(DatasetCollectionBinding.id == dataset.collection_binding_id)
- dataset_collection_binding = db.session.scalars(stmt).one_or_none()
+ # Annotation bindings may only be flushed in the caller's transaction.
+ # Keep the legacy lookup for workflows initialized with session=None.
+ binding_session = session if session is not None else db.session
+ dataset_collection_binding = binding_session.scalars(stmt).one_or_none()
if dataset_collection_binding:
collection_name = dataset_collection_binding.collection_name
else:
diff --git a/api/providers/vdb/vdb-qdrant/tests/unit_tests/test_qdrant_vector.py b/api/providers/vdb/vdb-qdrant/tests/unit_tests/test_qdrant_vector.py
index 3f88219e6fc..e7b3c68b24b 100644
--- a/api/providers/vdb/vdb-qdrant/tests/unit_tests/test_qdrant_vector.py
+++ b/api/providers/vdb/vdb-qdrant/tests/unit_tests/test_qdrant_vector.py
@@ -4,13 +4,16 @@ import os
import sys
import types
from collections import UserDict
+from pathlib import Path
from types import SimpleNamespace
from typing import override
from unittest.mock import MagicMock, patch
import pytest
+from sqlalchemy import create_engine, select
from sqlalchemy.orm import Session
+from core.rag.datasource.vdb.vector_factory import Vector
from core.rag.models.document import Document
from models.dataset import Dataset, DatasetCollectionBinding
from models.enums import CollectionBindingType
@@ -295,12 +298,15 @@ def test_search_and_helper_methods(qdrant_module):
@pytest.mark.parametrize("sqlite3_session", [(DatasetCollectionBinding,)], indirect=True)
[email protected]("use_caller_session", [True, False])
def test_qdrant_factory_paths(
qdrant_module,
monkeypatch: pytest.MonkeyPatch,
sqlite3_session: Session,
+ use_caller_session: bool,
):
factory = qdrant_module.QdrantVectorFactory()
+ session = sqlite3_session if use_caller_session else None
dataset = Dataset(id="dataset-1", tenant_id="tenant-1", collection_binding_id=None)
monkeypatch.setattr(qdrant_module.Dataset, "gen_collection_name_by_id", lambda _id: "AUTO_COLLECTION")
monkeypatch.setattr(qdrant_module, "current_app", SimpleNamespace(config=SimpleNamespace(root_path="/root")))
@@ -312,7 +318,7 @@ def test_qdrant_factory_paths(
monkeypatch.setattr(qdrant_module.dify_config, "QDRANT_REPLICATION_FACTOR", 1)
with patch.object(qdrant_module, "QdrantVector", return_value="vector") as vector_cls:
- result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ result = factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=session)
assert result == "vector"
assert vector_cls.call_args.kwargs["collection_name"] == "AUTO_COLLECTION"
assert dataset.index_struct is not None
@@ -331,10 +337,53 @@ def test_qdrant_factory_paths(
monkeypatch.setattr(qdrant_module.db, "session", sqlite3_session)
with patch.object(qdrant_module, "QdrantVector", return_value="vector") as vector_cls:
- factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=session)
assert vector_cls.call_args.kwargs["collection_name"] == "BOUND_COLLECTION"
sqlite3_session.delete(binding)
sqlite3_session.commit()
with pytest.raises(ValueError, match="Dataset Collection Bindings does not exist"):
- factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=session)
+
+
[email protected]("vector_type", [None, "qdrant"])
+def test_vector_reads_uncommitted_annotation_binding(
+ qdrant_module, monkeypatch: pytest.MonkeyPatch, tmp_path: Path, vector_type: str | None
+):
+ # Separate connections must not see each other's uncommitted rows. An
+ # in-memory SQLite engine would share a connection and hide this regression.
+ engine = create_engine(f"sqlite:///{tmp_path / 'bindings.db'}")
+ DatasetCollectionBinding.__table__.create(engine)
+ monkeypatch.setattr(Vector, "get_vector_factory", staticmethod(lambda _: qdrant_module.QdrantVectorFactory))
+ monkeypatch.setattr(qdrant_module, "current_app", SimpleNamespace(config=SimpleNamespace(root_path="/root")))
+ monkeypatch.setattr(qdrant_module.dify_config, "VECTOR_STORE", "qdrant")
+ monkeypatch.setattr(qdrant_module.dify_config, "VECTOR_STORE_WHITELIST_ENABLE", False)
+ monkeypatch.setattr(qdrant_module.dify_config, "QDRANT_URL", "http://localhost:6333")
+
+ try:
+ with Session(engine) as session, Session(engine) as other_session:
+ monkeypatch.setattr(qdrant_module.db, "session", other_session)
+ binding = DatasetCollectionBinding(
+ provider_name="provider",
+ model_name="model",
+ type=CollectionBindingType.ANNOTATION,
+ collection_name="ANNOTATION_COLLECTION",
+ )
+ session.add(binding)
+ session.flush()
+ stmt = select(DatasetCollectionBinding).where(DatasetCollectionBinding.id == binding.id)
+ assert other_session.scalar(stmt) is None
+ dataset = Dataset(id="app-1", tenant_id="tenant-1", collection_binding_id=binding.id)
+
+ vector = Vector(
+ dataset, attributes=["doc_id", "annotation_id", "app_id"], session=session, vector_type=vector_type
+ )
+
+ assert vector._vector_processor.collection_name == "ANNOTATION_COLLECTION"
+ assert dataset.index_struct_dict is not None
+ assert dataset.index_struct_dict["vector_store"]["class_prefix"] == "ANNOTATION_COLLECTION"
+ assert other_session.scalar(stmt) is None
+ session.rollback()
+ assert session.scalar(stmt) is None
+ finally:
+ engine.dispose()
diff --git a/api/providers/vdb/vdb-relyt/src/dify_vdb_relyt/relyt_vector.py b/api/providers/vdb/vdb-relyt/src/dify_vdb_relyt/relyt_vector.py
index 423f893f57e..37baced67c3 100644
--- a/api/providers/vdb/vdb-relyt/src/dify_vdb_relyt/relyt_vector.py
+++ b/api/providers/vdb/vdb-relyt/src/dify_vdb_relyt/relyt_vector.py
@@ -302,7 +302,9 @@ class RelytVector(BaseVector):
class RelytVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> RelytVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> RelytVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-relyt/tests/unit_tests/test_relyt_vector.py b/api/providers/vdb/vdb-relyt/tests/unit_tests/test_relyt_vector.py
index 95e63910b3e..57fbb4ec1a9 100644
--- a/api/providers/vdb/vdb-relyt/tests/unit_tests/test_relyt_vector.py
+++ b/api/providers/vdb/vdb-relyt/tests/unit_tests/test_relyt_vector.py
@@ -312,8 +312,8 @@ def test_relyt_factory_existing_and_generated_collection(relyt_module, monkeypat
monkeypatch.setattr(relyt_module.dify_config, "RELYT_DATABASE", "relyt")
with patch.object(relyt_module, "RelytVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-tablestore/src/dify_vdb_tablestore/tablestore_vector.py b/api/providers/vdb/vdb-tablestore/src/dify_vdb_tablestore/tablestore_vector.py
index 9f1371fdd10..395da73d710 100644
--- a/api/providers/vdb/vdb-tablestore/src/dify_vdb_tablestore/tablestore_vector.py
+++ b/api/providers/vdb/vdb-tablestore/src/dify_vdb_tablestore/tablestore_vector.py
@@ -6,6 +6,7 @@ from typing import Any, override
import tablestore # type: ignore
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from tablestore import BatchGetRowRequest, TableInBatchGetRowItem
from configs import dify_config
@@ -404,7 +405,9 @@ class TableStoreVector(BaseVector):
class TableStoreVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> TableStoreVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> TableStoreVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-tablestore/tests/unit_tests/test_tablestore_vector.py b/api/providers/vdb/vdb-tablestore/tests/unit_tests/test_tablestore_vector.py
index 2acd214472c..8f915f03023 100644
--- a/api/providers/vdb/vdb-tablestore/tests/unit_tests/test_tablestore_vector.py
+++ b/api/providers/vdb/vdb-tablestore/tests/unit_tests/test_tablestore_vector.py
@@ -305,8 +305,8 @@ def test_tablestore_factory_uses_existing_or_generated_collection(tablestore_mod
monkeypatch.setattr(tablestore_module.dify_config, "TABLESTORE_NORMALIZE_FULLTEXT_BM25_SCORE", True)
with patch.object(tablestore_module, "TableStoreVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-tencent/src/dify_vdb_tencent/tencent_vector.py b/api/providers/vdb/vdb-tencent/src/dify_vdb_tencent/tencent_vector.py
index 9d1cc8ff132..cbacb33f9db 100644
--- a/api/providers/vdb/vdb-tencent/src/dify_vdb_tencent/tencent_vector.py
+++ b/api/providers/vdb/vdb-tencent/src/dify_vdb_tencent/tencent_vector.py
@@ -4,6 +4,7 @@ import math
from typing import Any, TypedDict, override
from pydantic import BaseModel
+from sqlalchemy.orm import Session
from tcvdb_text.encoder import BM25Encoder # type: ignore
from tcvectordb import RPCVectorDBClient, VectorDBException # type: ignore
from tcvectordb.model import document, enum # type: ignore
@@ -337,7 +338,9 @@ class TencentVector(BaseVector):
class TencentVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> TencentVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> TencentVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-tencent/tests/unit_tests/test_tencent_vector.py b/api/providers/vdb/vdb-tencent/tests/unit_tests/test_tencent_vector.py
index 41db20db61a..5d4d05aec28 100644
--- a/api/providers/vdb/vdb-tencent/tests/unit_tests/test_tencent_vector.py
+++ b/api/providers/vdb/vdb-tencent/tests/unit_tests/test_tencent_vector.py
@@ -316,8 +316,8 @@ def test_tencent_factory_existing_and_generated_collection(tencent_module, monke
monkeypatch.setattr(tencent_module.dify_config, "TENCENT_VECTOR_DB_ENABLE_HYBRID_SEARCH", True)
with patch.object(tencent_module, "TencentVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-tidb-on-qdrant/src/dify_vdb_tidb_on_qdrant/tidb_on_qdrant_vector.py b/api/providers/vdb/vdb-tidb-on-qdrant/src/dify_vdb_tidb_on_qdrant/tidb_on_qdrant_vector.py
index be6df15cae1..9211d2e0a4f 100644
--- a/api/providers/vdb/vdb-tidb-on-qdrant/src/dify_vdb_tidb_on_qdrant/tidb_on_qdrant_vector.py
+++ b/api/providers/vdb/vdb-tidb-on-qdrant/src/dify_vdb_tidb_on_qdrant/tidb_on_qdrant_vector.py
@@ -8,6 +8,7 @@ from typing import TYPE_CHECKING, Any, override
import httpx
import qdrant_client
+from sqlalchemy.orm import Session
logger = logging.getLogger(__name__)
from flask import current_app
@@ -437,7 +438,9 @@ class TidbOnQdrantVector(BaseVector):
class TidbOnQdrantVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> TidbOnQdrantVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> TidbOnQdrantVector:
logger.info("init_vector: tenant_id=%s, dataset_id=%s", dataset.tenant_id, dataset.id)
stmt = select(TidbAuthBinding).where(TidbAuthBinding.tenant_id == dataset.tenant_id)
tidb_auth_binding = db.session.scalars(stmt).one_or_none()
diff --git a/api/providers/vdb/vdb-tidb-vector/src/dify_vdb_tidb_vector/tidb_vector.py b/api/providers/vdb/vdb-tidb-vector/src/dify_vdb_tidb_vector/tidb_vector.py
index 1fadb5aef58..17807479a27 100644
--- a/api/providers/vdb/vdb-tidb-vector/src/dify_vdb_tidb_vector/tidb_vector.py
+++ b/api/providers/vdb/vdb-tidb-vector/src/dify_vdb_tidb_vector/tidb_vector.py
@@ -356,7 +356,9 @@ class TiDBVector(BaseVector):
class TiDBVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> TiDBVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> TiDBVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-tidb-vector/tests/unit_tests/test_tidb_vector.py b/api/providers/vdb/vdb-tidb-vector/tests/unit_tests/test_tidb_vector.py
index 52ec6fc9999..0cf49357f93 100644
--- a/api/providers/vdb/vdb-tidb-vector/tests/unit_tests/test_tidb_vector.py
+++ b/api/providers/vdb/vdb-tidb-vector/tests/unit_tests/test_tidb_vector.py
@@ -548,8 +548,8 @@ def test_tidb_factory_uses_existing_or_generated_collection(tidb_module, monkeyp
monkeypatch.setattr(tidb_module.dify_config, "APPLICATION_NAME", "dify-app")
with patch.object(tidb_module, "TiDBVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-upstash/src/dify_vdb_upstash/upstash_vector.py b/api/providers/vdb/vdb-upstash/src/dify_vdb_upstash/upstash_vector.py
index ddf51d7ca42..799021eb4c9 100644
--- a/api/providers/vdb/vdb-upstash/src/dify_vdb_upstash/upstash_vector.py
+++ b/api/providers/vdb/vdb-upstash/src/dify_vdb_upstash/upstash_vector.py
@@ -3,6 +3,7 @@ from typing import Any, override
from uuid import uuid4
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from upstash_vector import Index, Vector
from configs import dify_config
@@ -136,7 +137,9 @@ class UpstashVector(BaseVector):
class UpstashVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> UpstashVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> UpstashVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-upstash/tests/unit_tests/test_upstash_vector.py b/api/providers/vdb/vdb-upstash/tests/unit_tests/test_upstash_vector.py
index 5412a4bce33..4062a873f3a 100644
--- a/api/providers/vdb/vdb-upstash/tests/unit_tests/test_upstash_vector.py
+++ b/api/providers/vdb/vdb-upstash/tests/unit_tests/test_upstash_vector.py
@@ -176,8 +176,8 @@ def test_upstash_factory_uses_existing_or_generated_collection(upstash_module, m
monkeypatch.setattr(upstash_module.dify_config, "UPSTASH_VECTOR_TOKEN", "token-123")
with patch.object(upstash_module, "UpstashVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-vastbase/src/dify_vdb_vastbase/vastbase_vector.py b/api/providers/vdb/vdb-vastbase/src/dify_vdb_vastbase/vastbase_vector.py
index 0e4c36a3923..5f090c628ed 100644
--- a/api/providers/vdb/vdb-vastbase/src/dify_vdb_vastbase/vastbase_vector.py
+++ b/api/providers/vdb/vdb-vastbase/src/dify_vdb_vastbase/vastbase_vector.py
@@ -9,6 +9,7 @@ import psycopg2.errors
import psycopg2.extras
import psycopg2.pool
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from configs import dify_config
from core.rag.datasource.vdb.vector_base import BaseVector
@@ -288,7 +289,9 @@ class VastbaseVector(BaseVector):
class VastbaseVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> VastbaseVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> VastbaseVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix
diff --git a/api/providers/vdb/vdb-vastbase/tests/unit_tests/test_vastbase_vector.py b/api/providers/vdb/vdb-vastbase/tests/unit_tests/test_vastbase_vector.py
index 97efad8dad0..0853cbfbc41 100644
--- a/api/providers/vdb/vdb-vastbase/tests/unit_tests/test_vastbase_vector.py
+++ b/api/providers/vdb/vdb-vastbase/tests/unit_tests/test_vastbase_vector.py
@@ -382,8 +382,8 @@ def test_vastbase_factory_uses_existing_or_generated_collection(vastbase_module,
monkeypatch.setattr(vastbase_module.dify_config, "VASTBASE_MAX_CONNECTION", 5)
with patch.object(vastbase_module, "VastbaseVector", return_value="vector") as vector_cls:
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
diff --git a/api/providers/vdb/vdb-vikingdb/src/dify_vdb_vikingdb/vikingdb_vector.py b/api/providers/vdb/vdb-vikingdb/src/dify_vdb_vikingdb/vikingdb_vector.py
index bc5e542cf0a..96eecf77ffd 100644
--- a/api/providers/vdb/vdb-vikingdb/src/dify_vdb_vikingdb/vikingdb_vector.py
+++ b/api/providers/vdb/vdb-vikingdb/src/dify_vdb_vikingdb/vikingdb_vector.py
@@ -2,6 +2,7 @@ import json
from typing import Any, override
from pydantic import BaseModel
+from sqlalchemy.orm import Session
from volcengine.viking_db import ( # type: ignore
Data,
DistanceType,
@@ -220,7 +221,9 @@ class VikingDBVector(BaseVector):
class VikingDBVectorFactory(AbstractVectorFactory):
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> VikingDBVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> VikingDBVector:
if dataset.index_struct_dict:
class_prefix: str = dataset.index_struct_dict["vector_store"]["class_prefix"]
collection_name = class_prefix.lower()
diff --git a/api/providers/vdb/vdb-vikingdb/tests/unit_tests/test_vikingdb_vector.py b/api/providers/vdb/vdb-vikingdb/tests/unit_tests/test_vikingdb_vector.py
index efa02bc4ac3..aad93aa3886 100644
--- a/api/providers/vdb/vdb-vikingdb/tests/unit_tests/test_vikingdb_vector.py
+++ b/api/providers/vdb/vdb-vikingdb/tests/unit_tests/test_vikingdb_vector.py
@@ -272,8 +272,8 @@ def test_vikingdb_factory_validates_config_and_builds_vector(vikingdb_module, mo
monkeypatch.setattr(vikingdb_module.dify_config, "VIKINGDB_CONNECTION_TIMEOUT", 10)
monkeypatch.setattr(vikingdb_module.dify_config, "VIKINGDB_SOCKET_TIMEOUT", 20)
- result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock())
- result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock())
+ result_1 = factory.init_vector(dataset_with_index, attributes=[], embeddings=MagicMock(), session=None)
+ result_2 = factory.init_vector(dataset_without_index, attributes=[], embeddings=MagicMock(), session=None)
assert result_1 == "vector"
assert result_2 == "vector"
@@ -306,4 +306,4 @@ def test_vikingdb_factory_raises_when_required_config_missing(
monkeypatch.setattr(vikingdb_module.dify_config, field, None)
with pytest.raises(ValueError, match=message):
- factory.init_vector(dataset, attributes=[], embeddings=MagicMock())
+ factory.init_vector(dataset, attributes=[], embeddings=MagicMock(), session=None)
diff --git a/api/providers/vdb/vdb-weaviate/src/dify_vdb_weaviate/weaviate_vector.py b/api/providers/vdb/vdb-weaviate/src/dify_vdb_weaviate/weaviate_vector.py
index 5735b6a018a..73d31a1a802 100644
--- a/api/providers/vdb/vdb-weaviate/src/dify_vdb_weaviate/weaviate_vector.py
+++ b/api/providers/vdb/vdb-weaviate/src/dify_vdb_weaviate/weaviate_vector.py
@@ -18,6 +18,7 @@ from urllib.parse import urlparse
import weaviate
import weaviate.classes.config as wc
from pydantic import BaseModel, model_validator
+from sqlalchemy.orm import Session
from weaviate.classes.data import DataObject
from weaviate.classes.init import Auth
from weaviate.classes.query import Filter, MetadataQuery
@@ -606,7 +607,9 @@ class WeaviateVectorFactory(AbstractVectorFactory):
"""Factory class for creating WeaviateVector instances."""
@override
- def init_vector(self, dataset: Dataset, attributes: list, embeddings: Embeddings) -> WeaviateVector:
+ def init_vector(
+ self, dataset: Dataset, attributes: list, embeddings: Embeddings, *, session: Session | None
+ ) -> WeaviateVector:
"""
Initializes a WeaviateVector instance for the given dataset.
diff --git a/api/providers/vdb/vdb-weaviate/tests/unit_tests/test_weaviate_vector.py b/api/providers/vdb/vdb-weaviate/tests/unit_tests/test_weaviate_vector.py
index fa1a70e18d8..bec25a0121a 100644
--- a/api/providers/vdb/vdb-weaviate/tests/unit_tests/test_weaviate_vector.py
+++ b/api/providers/vdb/vdb-weaviate/tests/unit_tests/test_weaviate_vector.py
@@ -893,7 +893,7 @@ class TestWeaviateVectorFactory(unittest.TestCase):
patch("dify_vdb_weaviate.weaviate_vector.WeaviateVector", return_value="vector") as mock_vector,
):
factory = weaviate_vector_module.WeaviateVectorFactory()
- result = factory.init_vector(dataset, attributes, MagicMock())
+ result = factory.init_vector(dataset, attributes, MagicMock(), session=None)
assert result == "vector"
config = mock_vector.call_args.kwargs["config"]
@@ -922,7 +922,7 @@ class TestWeaviateVectorFactory(unittest.TestCase):
patch("dify_vdb_weaviate.weaviate_vector.WeaviateVector", return_value="vector") as mock_vector,
):
factory = weaviate_vector_module.WeaviateVectorFactory()
- result = factory.init_vector(dataset, attributes, MagicMock())
+ result = factory.init_vector(dataset, attributes, MagicMock(), session=None)
assert result == "vector"
assert mock_vector.call_args.kwargs["collection_name"] == "GeneratedCollection_Node"
diff --git a/api/tests/unit_tests/core/rag/datasource/vdb/test_vector_factory.py b/api/tests/unit_tests/core/rag/datasource/vdb/test_vector_factory.py
index 9c4dc266ad1..b69512daef2 100644
--- a/api/tests/unit_tests/core/rag/datasource/vdb/test_vector_factory.py
+++ b/api/tests/unit_tests/core/rag/datasource/vdb/test_vector_factory.py
@@ -203,6 +203,24 @@ def test_vector_init_uses_default_and_custom_attributes(vector_factory_module, u
assert [call.kwargs["session"] for call in init_vector.call_args_list] == [unbound_session, unbound_session]
[email protected]("with_session", [True, False])
+def test_vector_with_resolved_type_passes_session(vector_factory_module, unbound_session: Session, with_session: bool):
+ dataset = _dataset()
+ session = unbound_session if with_session else None
+ factory = MagicMock()
+ with (
+ patch.object(vector_factory_module.Vector, "get_vector_factory", return_value=factory),
+ patch.object(vector_factory_module.Vector, "resolve_vector_type") as resolve_vector_type,
+ ):
+ vector = vector_factory_module.Vector(dataset, session=session, vector_type="qdrant")
+
+ resolve_vector_type.assert_not_called()
+ factory.return_value.init_vector.assert_called_once_with(
+ dataset, vector._attributes, vector._embeddings, session=session
+ )
+ assert vector._vector_processor is factory.return_value.init_vector.return_value
+
+
def test_lazy_embeddings_defer_real_load_until_first_embed_call(vector_factory_module, monkeypatch: pytest.MonkeyPatch):
"""``Vector(dataset, session=...)`` must not transitively call ``ModelManager`` during
construction. The real embedding model should only be materialized on the
@@ -265,8 +283,8 @@ def test_init_vector_prefers_dataset_index_struct(
calls = {"vector_type": None, "init_args": None}
class _Factory:
- def init_vector(self, dataset, attributes, embeddings):
- calls["init_args"] = (dataset, attributes, embeddings)
+ def init_vector(self, dataset, attributes, embeddings, *, session):
+ calls["init_args"] = (dataset, attributes, embeddings, session)
return "vector-processor"
monkeypatch.setattr(
@@ -284,7 +302,7 @@ def test_init_vector_prefers_dataset_index_struct(
assert result == "vector-processor"
assert calls["vector_type"] == vector_factory_module.VectorType.UPSTASH
- assert calls["init_args"] == (vector._dataset, ["doc_id"], "embeddings")
+ assert calls["init_args"] == (vector._dataset, ["doc_id"], "embeddings", unbound_session)
def test_init_vector_uses_whitelist_override(
@@ -293,7 +311,7 @@ def test_init_vector_uses_whitelist_override(
calls = {"vector_type": None}
class _Factory:
- def init_vector(self, dataset, attributes, embeddings):
+ def init_vector(self, dataset, attributes, embeddings, *, session):
return "vector-processor"
tenant_id = str(uuid4())