Vectors are available from MariaDB 11.7.
MariaDB has a dedicated VECTOR(N) data type with a built-in data validation. N is the number of dimensions that all vector values in the column have.
Vector indexes are dimensionality-specific.
All vectors inserted into an indexed column must match the index's target dimensionality.
Inserting vectors with different dimensionalities will result in an error.
Consider the following table:
CREATETABLEembeddings (
doc_id BIGINT UNSIGNED PRIMARY KEY,
embedding VECTOR(1536)
);To have a fast vector search, you have to index the vector column, creating a VECTOR index:
CREATETABLEembeddings (
doc_id BIGINT UNSIGNED PRIMARY KEY,
embedding VECTOR(1536) NOT NULL,
VECTORINDEX (embedding)
);Note that there can be only one vector index in the table, and the indexed vector column must be NOT NULL.
There are two options that can be used to configure the vector index:
M— Larger values mean slowerSELECTandINSERTstatements, larger index size and higher memory consumption, but more accurate results. The valid range is from3to200.DISTANCE— Distance function to build the vector index for. Searches using a different distance function will not be able to use a vector index. Valid values arecosineandeuclidean(the default).
Declare DISTANCE explicitly. The default is euclidean, and a query using a different distance function than the one the index was built for cannot use the index, it falls back to a full table scan. Match the distance function to the metric your embedding model recommends.
This page is licensed: CC BY-SA / Gnu FDL
Last updated
Was this helpful?