Struct VectorConfig 

pub struct VectorConfig {
    pub dimensions: u32,
    pub index_type: VectorIndexType,
    pub distance_metric: DistanceMetric,
}

Configuration for a vector field (pgvector).

This represents the configuration for a vector embedding field, including dimensions, index type, and distance metric.

The same config describes both vector field types: for FieldType::Vector dimensions counts float components, for FieldType::BitVector it counts bits. Which distance metrics are legal follows from the field type — see VectorConfig::validate_for.

Example

use fraiseql_core::schema::{VectorConfig, VectorIndexType, DistanceMetric};

let config = VectorConfig {
    dimensions: 1536,
    index_type: VectorIndexType::Hnsw,
    distance_metric: DistanceMetric::Cosine,
};

Fields

dimensions: u32

Number of dimensions in the vector: float components for FieldType::Vector, bits for FieldType::BitVector (e.g., 1536 for OpenAI embeddings).

index_type: VectorIndexType

Type of index to use for similarity search.

distance_metric: DistanceMetric

Distance metric for similarity calculations.

Implementations

impl VectorConfig

pub fn new(dimensions: u32) -> Self

Create a new vector config with default index and distance metric.

pub const fn openai() -> Self

Create a vector config for OpenAI embeddings (1536 dimensions, cosine).

pub const fn openai_small() -> Self

Create a vector config for OpenAI small embeddings (512 dimensions, cosine).

pub const fn with_index(self, index_type: VectorIndexType) -> Self

Set the index type.

pub const fn with_distance(self, distance_metric: DistanceMetric) -> Self

Set the distance metric.

pub const fn binary(bits: u32) -> Self

Create a binary-vector config (bit(N), hamming distance, HNSW).

bits is the bit width of the column, e.g. 768 for a binary_quantize(embedding)::bit(768) fingerprint.

pub fn validate_for(&self, field_type: &FieldType) -> Result<()>

Check that this config is coherent with the field type carrying it (#959).

The three refusals are all cases pgvector cannot execute, and each one fails at a different moment when it is not caught here: a float metric on a bit(N) column is an operator that does not exist, a binary metric on a vector(N) column is the same in reverse, and ivfflat + jaccard is DDL CREATE INDEX rejects — pgvector 0.8 ships bit_jaccard_ops for hnsw only.

Errors

Returns FraiseQLError::Validation when dimensions is zero, when the metric and the field type disagree about float versus binary vectors, or when the index type has no operator class for the metric.

Trait Implementations

impl Clone for VectorConfig

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for VectorConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for VectorConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for VectorConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Eq for VectorConfig

impl PartialEq for VectorConfig

fn eq(&self, other: &Self) -> bool

Equality operator ==.

fn ne(&self, other: &Rhs) -> bool

Inequality operator !=.

impl Serialize for VectorConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.

impl StructuralPartialEq for VectorConfig