qdrant_client.local.distances module

class ContextPair(positive: list[float], negative: list[float])[source]

Bases: object

class ContextQuery(context_pairs: list[ContextPair])[source]

Bases: object

class DiscoveryQuery(target: list[float], context: list[ContextPair])[source]

Bases: object

class DistanceOrder(value)[source]

Bases: str, Enum

An enumeration.

BIGGER_IS_BETTER = 'bigger_is_better'

SMALLER_IS_BETTER = 'smaller_is_better'

class RecoQuery(positive: Optional[list[list[float]]] = None, negative: Optional[list[list[float]]] = None)[source]

Bases: object

calculate_context_scores(query: ContextQuery, vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], distance_type: Distance) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

calculate_discovery_ranks(context: list[ContextPair], vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], distance_type: Distance) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

calculate_discovery_scores(query: DiscoveryQuery, vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], distance_type: Distance) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

calculate_distance(query: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], distance_type: Distance) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

calculate_distance_core(query: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], distance_type: Distance) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

Calculate same internal distances as in core, rather than the final displayed distance

calculate_recommend_best_scores(query: RecoQuery, vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], distance_type: Distance) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

cosine_similarity(query: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]]) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

Calculate cosine distance between query and vectors :param query: query vector :param vectors: vectors to calculate distance with

Returns:

distances

distance_to_order(distance: Distance) → DistanceOrder[source]

Convert distance to order :param distance: distance to convert

Returns:

order

dot_product(query: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]]) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

Calculate dot product between query and vectors :param query: query vector. :param vectors: vectors to calculate distance with

Returns:

distances

euclidean_distance(query: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]]) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

Calculate euclidean distance between query and vectors :param query: query vector. :param vectors: vectors to calculate distance with

Returns:

distances

fast_sigmoid(x: float32) → float32[source]

manhattan_distance(query: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]], vectors: ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]]) → ndarray[tuple[int, ...], dtype[Union[bool, int8, int16, int32, int64, uint8, uint16, uint32, uint64, float16, float32, float64, longdouble]]][source]

Calculate manhattan distance between query and vectors :param query: query vector. :param vectors: vectors to calculate distance with

Returns:

distances

scaled_fast_sigmoid(x: float32) → float32[source]