Surrealdb Vector

SurrealDB Vector Search

HNSW Index

Create a basic HNSW index:

DEFINE INDEX hnsw_idx ON pts FIELDS point HNSW DIMENSION 4;

With specific distance function and type:

DEFINE INDEX hnsw_idx ON pts FIELDS point HNSW DIMENSION 4 DIST EUCLIDEAN TYPE F64;

Available types: F64, F32, I64, I32, I16.

Full Table Example

DEFINE TABLE OVERWRITE document SCHEMALESS;
DEFINE FIELD OVERWRITE embedding ON document TYPE array<float>;
DEFINE INDEX OVERWRITE hnsw_idx_document ON document
    FIELDS embedding
    HNSW DIMENSION 384
    DIST COSINE
    TYPE F32
    EFC 150 M 12 M0 24;

HNSW Parameters

Querying Vectors

The <|K, EF|> operator performs KNN search. K is the number of results, EF is the search effort (higher = more accurate, slower).

Recommended effort values:

• 40 — default, good accuracy
• 17 — fast but may miss some results

Basic KNN Query

SELECT
    *,
    vector::distance::knn() AS dist
FROM document
WHERE embedding <|10, 40|> $vector;

vector::distance::knn() uses the distance function defined by the index.

Scored Results with Threshold

SELECT *, score
FROM (
    SELECT *, (1 - vector::distance::knn()) AS score
    FROM document
    WHERE embedding <|20, 40|> $vector
)
WHERE score >= $threshold
ORDER BY score DESC;