WeaviateEx.Types.Vector (WeaviateEx v0.7.4)

Vector type with multi-dimensional support for ColBERT and other advanced models.

Supports both standard 1D vectors and multi-dimensional vectors (2D arrays) used by models like ColBERT for late interaction retrieval.

Standard Vectors

Most embeddings are 1D vectors:

vector = [0.1, 0.2, 0.3, 0.4]
Vector.validate(vector)
# => :ok

Vector.shape(vector)
# => {4}

Multi-Dimensional Vectors

Some models like ColBERT produce multi-dimensional vectors:

multi_vector = [[0.1, 0.2], [0.3, 0.4], [0.5, 0.6]]
Vector.validate(multi_vector)
# => :ok

Vector.shape(multi_vector)
# => {3, 2}

Summary

Types

t()

Functions

cosine_similarity(v1, v2)

Computes the cosine similarity between two vectors.

dimensionality(vector)

Returns the dimensionality of a vector.

dot_product(v1, v2)

Computes the dot product of two vectors.

multi_dimensional?(arg1)

Checks if a vector is multi-dimensional.

normalize(vector)

Normalizes a vector to unit length (L2 normalization).

one_dimensional?(arg1)

Checks if a vector is 1-dimensional.

shape(vector)

Returns the shape of a vector.

validate(vector)

Validates a vector or multi-dimensional vector.

Types

t()

@type t() :: [float()] | [[float()]]

Functions

cosine_similarity(v1, v2)

@spec cosine_similarity([float()], [float()]) :: float()

Computes the cosine similarity between two vectors.

Examples

Vector.cosine_similarity([1.0, 0.0], [0.0, 1.0])
# => 0.0

Vector.cosine_similarity([1.0, 0.0], [1.0, 0.0])
# => 1.0

dimensionality(vector)

@spec dimensionality(t()) :: non_neg_integer()

Returns the dimensionality of a vector.

For 1D vectors, returns the length. For multi-dimensional vectors, returns the inner dimension.

Examples

Vector.dimensionality([0.1, 0.2, 0.3])
# => 3

Vector.dimensionality([[0.1, 0.2], [0.3, 0.4], [0.5, 0.6]])
# => 2

dot_product(v1, v2)

@spec dot_product([float()], [float()]) :: float()

Computes the dot product of two vectors.

Examples

Vector.dot_product([1.0, 2.0], [3.0, 4.0])
# => 11.0

multi_dimensional?(arg1)

@spec multi_dimensional?(t()) :: boolean()

Checks if a vector is multi-dimensional.

Examples

Vector.multi_dimensional?([[0.1, 0.2], [0.3, 0.4]])
# => true

Vector.multi_dimensional?([0.1, 0.2, 0.3])
# => false

normalize(vector)

@spec normalize([float()]) :: [float()]

Normalizes a vector to unit length (L2 normalization).

Examples

Vector.normalize([3.0, 4.0])
# => [0.6, 0.8]

one_dimensional?(arg1)

@spec one_dimensional?(t()) :: boolean()

Checks if a vector is 1-dimensional.

Examples

Vector.one_dimensional?([0.1, 0.2, 0.3])
# => true

Vector.one_dimensional?([[0.1, 0.2], [0.3, 0.4]])
# => false

shape(vector)

@spec shape(t()) :: {non_neg_integer()} | {non_neg_integer(), non_neg_integer()}

Returns the shape of a vector.

Examples

Vector.shape([0.1, 0.2, 0.3, 0.4])
# => {4}

Vector.shape([[0.1, 0.2], [0.3, 0.4], [0.5, 0.6]])
# => {3, 2}

validate(vector)

@spec validate(term()) :: :ok | {:error, String.t()}

Validates a vector or multi-dimensional vector.

Examples

Vector.validate([0.1, 0.2, 0.3])
# => :ok

Vector.validate([[0.1, 0.2], [0.3, 0.4]])
# => :ok

Vector.validate([])
# => {:error, "Vector cannot be empty"}

Vector.validate([[0.1, 0.2], [0.3]])
# => {:error, "Multi-dimensional vectors must have consistent inner dimensions"}