Module vector 

First-class VECTOR value type.

LoraDB VECTOR values are fixed-dimension, typed numeric coordinate collections. A LoraVector can be stored directly as a node or relationship property, returned through every binding, compared for equality, and used as input to the built-in vector math functions (vector.similarity.cosine, vector.similarity.euclidean, vector_distance, vector_norm, vector_dimension_count, toIntegerList, toFloatList).

Vector indexes and approximate kNN are intentionally out of scope for this pass — exhaustive search via ORDER BY vector.similarity.*(…) LIMIT k works today; an index-backed variant is future work.

Structs

LoraVector

A first-class VECTOR value.

Enums

RawCoordinate

Raw numeric input for one coordinate before it has been coerced into the destination coordinate type. Executors / binding layers feed values through this enum so the coercion rules live in one place.

VectorBuildError

Error returned by LoraVector::try_new. Kept as a concrete enum so the executor can render a single-line error message without inspecting the underlying cause.

VectorCoordinateType

Canonical coordinate type for a vector.

VectorValues

Internal storage for a vector. One variant per supported coordinate type; dimension is implicit in the inner Vec’s length.

Constants

MAX_VECTOR_DIMENSION

Maximum dimension accepted by LoraDB’s vector(...) constructor.

Functions

cosine_similarity_bounded

Cosine similarity squashed into [0, 1]. Matches the documented vector.similarity.cosine behaviour.

cosine_similarity_raw

Raw cosine similarity in the range [-1, 1]. Returns None when either vector has zero norm, since cosine is undefined in that case.

dot_product

Dot product (f32 arithmetic, widened back to f64).

euclidean_distance

Euclidean (L2) distance.

euclidean_distance_squared

Squared Euclidean distance (sum of squared differences). Uses f32 arithmetic to match LoraDB’s vector function implementation.

euclidean_norm

Euclidean (L2) norm.

euclidean_similarity

Similarity score derived from squared Euclidean distance: 1 / (1 + d²). For the documented example where distance² == 22, this yields 1 / 23 ≈ 0.043478.

hamming_distance

Hamming distance: count of positions where the two vectors differ.

manhattan_distance

Manhattan (L1) distance.

manhattan_norm

Manhattan (L1) norm.

parse_string_values

Parse a string-form coordinate list, e.g. "[1.05e+00, 0.123, 5]". Used by vector() when vectorValue is a STRING.