Vector Types on Vchord

Half-Precision Vector: `vecf16`

Mon, 01 Jan 0001 00:00:00 +0000

Stored as a half-precision number format, vecf16 takes advantage of the 16-bit float, which uses half the memory and bandwidth compared to vector.

Here’s an example of creating a table with a vecf16 column and inserting values:

CREATE TABLE items (
  id bigserial PRIMARY KEY,
  embedding vecf16(3) NOT NULL
);

INSERT INTO items (embedding) VALUES ('[0.1, 0.2, 0]'), ('[0, 0.1, 0.2]');

Index can be created on vecf16 type as well.

8-Bit Integer Vector: `veci8`

Mon, 01 Jan 0001 00:00:00 +0000

since v0.3.0

Int8 is a data type used to represent signed integers. It stands for “integer 8-bit” and typically uses 8 bits of memory to store a value. This allows it to represent integer values ranging from -128 to 127.

Since each element in the vector only requires 8 bits (1 byte) of memory, veci8 can help reduce the overall memory footprint of your program.

Here’s an example of creating a table with a veci8 column and inserting values:

Sparse Vector: `svector`

Mon, 01 Jan 0001 00:00:00 +0000

since v0.3.0

Unlike dense vectors, sparse vectors are very high-dimensional but contain few non-zero values.

Typically, sparse vectors can be created from:

Word co-occurrence matrices
Term frequency-inverse document frequency (TF-IDF) vectors
User-item interaction matrices
Network adjacency matrices

Sparse vectors in pgvecto.rs are called svector.

Here’s an example of creating a table with a svector column and inserting values:

CREATE TABLE items (
  id bigserial PRIMARY KEY,
  embedding svector(10) NOT NULL
);

INSERT INTO items (embedding) VALUES ('[0.1,0,0,0,0,0,0,0,0,0]'), ('[0,0,0,0,0,0,0,0,0,0.5]');

Index can be created on svector type as well.

Binary Vector: `bvector`

Mon, 01 Jan 0001 00:00:00 +0000

since v0.3.0

The bvector type is a binary vector type in pgvecto.rs. It represents a binary vector, which is a vector where each component can take on two possible values, typically 0 and 1.

Here’s an example of creating a table with a bvector column and inserting values:

CREATE TABLE items (
  id bigserial PRIMARY KEY,
  embedding bvector(3) NOT NULL
);

INSERT INTO items (embedding) VALUES ('[1,0,1]'), ('[0,1,0]');

Index can be created on bvector type as well.