Qdrant base

QdrantVectorType = tuple[str, list[float]] | qdrant_models.NamedSparseVector module-attribute

logger = logging.getLogger(__name__) module-attribute

IncompatibleIndexParamsError

Bases: BackendError

Raised when the index parameters are incompatible with the index type.

QdrantAdapter

My implementation of a qdrant client tied to a specific collection.

Everything here should be general to any collection in qdrant.

collection_name property writable

dense_code_embedder = dense_code_embedder instance-attribute

dense_embedder = dense_embedder instance-attribute

qdrant_client = qdrant_client instance-attribute

sparse_embedder property

__init__(qdrant_client, dense_embedder=default_dense_embedder, dense_code_embedder=default_dense_embedder, sparse_embedder=default_sparse_embedder, collection_name=None)

add_index(metadata_key, params) async

Add an index to the collection metadata.

PARAMETER	DESCRIPTION
metadata_key	The key to add the index to TYPE: str
params	The parameters for the index TYPE: PayloadSchemaParams

add_items(upsertables, wait=True) async

Adds qdrant points to the collection.

Points contain the QdrantPayload information (id, content, metadata). Note: The content is either what was embedded, or where to find it.

PARAMETER	DESCRIPTION
upsertables	List of Upsertables to add to the collection TYPE: Sequence[Upsertable]
wait	Whether to wait for the operation to complete (defaults to True). TYPE: bool DEFAULT: True

RETURNS	DESCRIPTION
UpdateResult	The qdrant result of the operation

batch_update(update_operations, wait=False) async

delete_index(metadata_key) async

Delete an index from the collection metadata.

delete_items(points_selector, wait=True) async

get_collection_info() async

get_count(count_filter) async

get_index_infos() async

scroll(scroll_filter=None, limit=100, latest_point_id=None) async

Scroll through points in collection.

Note: Use scroll_all to iterate all records.

PARAMETER	DESCRIPTION
scroll_filter	Additional filter to apply to the scroll TYPE: Filter | None DEFAULT: None
limit	How many results to return in one go TYPE: int DEFAULT: 100
latest_point_id	The previous latest point id to start from TYPE: PointId | None DEFAULT: None

RETURNS	DESCRIPTION
tuple[list[QdrantPayload], PointId | None]	Tuple of records and the last point id (last point will be None if no more records)

scroll_all(scroll_filter=None, num_per_iteration=100, limit=None) async

Scroll through all points in collection.

PARAMETER	DESCRIPTION
scroll_filter	Additional filter to apply to the scroll TYPE: Filter | None DEFAULT: None
num_per_iteration	How many results to return per iteration TYPE: int DEFAULT: 100
limit	Maximum number of records to return TYPE: int | None DEFAULT: None

YIELDS	DESCRIPTION
AsyncIterator[list[QdrantPayload]]	List of records

search(query_text, query_vector_name=None, k=5, query_filter=None, score_threshold=None, use_sparse=False, offset=0, group_by=None, group_by_count=None) async

Search collection for items related to query_text.

PARAMETER	DESCRIPTION
query_text	Text that will be embedded to produce a query vector TYPE: str
query_vector_name	Name of the vector to use for the query TYPE: str | None DEFAULT: None
k	Number of results to return TYPE: int DEFAULT: 5
query_filter	Additional qdrant filter to apply to the search TYPE: Filter | None DEFAULT: None
score_threshold	Minimum score to return TYPE: float | None DEFAULT: None
use_sparse	Whether to use the sparse or dense embeddings TYPE: bool DEFAULT: False
offset	Offset for pagination TYPE: int DEFAULT: 0
group_by	Metadata field to group by TYPE: str | None DEFAULT: None
group_by_count	Number of results to return per group TYPE: int | None DEFAULT: None

RETURNS	DESCRIPTION
list[QdrantSearchResult]	List of search results

ensure_metadata_prefix(key)

Ensure the key has the metadata prefix.

lc_to_qdrant_sparse_vectors(sparse_vectors)

Ensure the correct SparseVector type is used for qdrant.

Necessary because langchain_qdrant defines an identical SparseVector class that prevents subclass checks passing within qdrant. Re-validating from the qdrant version ensures that the correct class is used.

make_payload_schema_params(type_, is_tenant=False, is_principal=False, on_disk=False)

Create a PayloadSchemaParams object from the given args.

PARAMETER	DESCRIPTION
type_	The type of the index (e.g., qdrant_models.PayloadSchemaType.KEYWORD) TYPE: PayloadSchemaType
is_tenant	Whether to use as a tenant index (e.g., user_id is generally a good tenant index) TYPE: bool DEFAULT: False
is_principal	Whether to use as a primary index (e.g., timestamp is good for time series) TYPE: bool DEFAULT: False
on_disk	Whether to store the index on disk (otherwise in-memory) TYPE: bool DEFAULT: False

Checks if the additional args are compatible with the type.

strip_metadata_prefix(key)

Strip the metadata prefix from the key.