barrel_vectordb_distance (barrel_vectordb v2.1.1)

Vector Distance Computation Module

Provides SIMD-accelerated distance computations via NIF. Falls back to pure Erlang implementation if NIF is not available.

Supports: - Cosine distance - Euclidean distance - Dot product

Vectors can be represented as: - Lists of floats (Erlang native, slower) - Binaries of 32-bit floats (NIF optimized, faster)

Summary

vector/0

Compute cosine distance: 1 - cos(a, b)

Compute cosine distance for pre-normalized vectors (faster)

Compute dot product of two vectors

Compute Euclidean distance

from_binary(Bin)

Convert binary vector to list of floats

nif_available()

Check if NIF is available

simd_info()

Return SIMD backend information

to_binary(List)

Convert list of floats to binary vector (32-bit floats)

-type vector() :: [float()] | binary().

-spec cosine_distance(vector(), vector()) -> float().