Vectorstores

flowtask.interfaces.vectorstores

abstract

AbstractStore

AbstractStore(*args, **kwargs)

Bases: ABC

AbstractStore class.

Parameters:

Name	Type	Description	Default
embedding_model	dict	Embeddings.	required

milvus

MilvusStore

MilvusStore(*args, **kwargs)

Bases: AbstractStore, CredentialsInterface

Interface for managing document storage in Milvus using LangChain.

connect async

connect(alias=None)

Connects to the Milvus database.

create_collection async

create_collection(collection_name, origin=None, dimension=768, index_type='HNSW', metric_type='L2', **kwargs)

Create a Milvus collection based on an origin.

Parameters:

Name	Type	Description	Default
collection_name	str	Name of the Milvus collection.	required
origin	Union[str, Path, Any]	None for minimal schema, Path for Avro file, or dataclass for schema.	None
dimension	int	Dimension of the vector field. Defaults to 768.	768
index_type	str	Index type for vector field. Defaults to "HNSW".	'HNSW'
metric_type	str	Metric type for vector index. Defaults to "L2".	'L2'

Returns:

Name	Type	Description
dict	dict	Result of the collection creation.

create_default_collection async

create_default_collection(collection_name, document=None, dimension=768, index_type=None, metric_type=None, schema_type='default', database=None, metadata_field=None, **kwargs)

create_collection.

Create a Schema (Milvus Collection) on the Current Database.

Parameters:

Name	Type	Description	Default
collection_name	str	Collection Name.	required
document	Any	List of Documents.	None
dimension	int	Vector Dimension. Defaults to 768.	768
index_type	str	Default index type of Vector Field. Defaults to "HNSW".	None
metric_type	str	Default Metric for Vector Index. Defaults to "L2".	None
schema_type	str	Description of Model. Defaults to 'default'.	'default'

Returns:

Name	Type	Description
dict	dict	description

delete_documents_by_attr async

delete_documents_by_attr(collection_name, attribute_name, attribute_value)

Deletes documents in the Milvus collection that match a specific attribute.

This asynchronous method removes documents from a specified Milvus collection where the given attribute matches the provided value.

Parameters:

Name	Type	Description	Default
collection_name	str	The name of the Milvus collection to delete from.	required
attribute_name	str	The name of the attribute to filter on.	required
attribute_value	str	The value of the attribute to match for deletion.	required

Raises:

Type	Description
Exception	If the deletion operation fails, the error is logged and re-raised.

Returns:

Type	Description
	None

Note

The method logs a notice with the number of deleted documents upon successful deletion.

load_documents async

load_documents(documents, upsert=True, collection=None, pk='source_type', dimension=768, index_type='HNSW', metric_type='L2', **kwargs)

Loads LangChain documents into the Milvus collection.

Parameters:

Name	Type	Description	Default
documents	List[Document]	List of LangChain Document objects.	required
upsert	bool	If True, delete existing documents with matching attributes before inserting.	True
pk	str	str: If upsert True, Key to be used for deleting documents before inserting.	'source_type'
collection	str	Name of the collection.	None