Crate vector 

OpenData Vector Database

A vector database built on SlateDB with SPANN-style indexing for efficient approximate nearest neighbor search.

Example

let config = Config {
    dimensions: 384,
    distance_metric: DistanceMetric::L2,
    flush_interval: Duration::from_secs(60),
    ..Default::default()
};
let db = VectorDb::open(config).await?;

let vectors = vec![
    Vector::builder("product-001", vec![0.1; 384])
        .attribute("category", "electronics")
        .attribute("price", 99i64)
        .build(),
];

db.write(vectors).await?;
db.flush().await?;

Re-exports

pub use db::VectorDb;

pub use db::VectorDbRead;

pub use model::Attribute;

pub use model::AttributeValue;

pub use model::Config;

pub use model::DistanceMetric;

pub use model::FieldSelection;

pub use model::FieldType;

pub use model::Filter;

pub use model::MetadataFieldSpec;

pub use model::Query;

pub use model::ReaderConfig;

pub use model::SearchResult;

pub use model::Vector;

pub use model::VectorBuilder;

pub use reader::VectorDbReader;

Modules

db

Vector database implementation with atomic flush semantics.

delta

In-memory delta for buffering vector writes before flush.

distance

Distance computation functions for vector similarity.

flusher

Flusher implementation for VectorDb.

hnsw

HNSW graph implementations for centroid search.

model

Public API types for the vector database (RFC 0002).

reader

Read-only client for the vector database.

serde

Serialization/deserialization for vector database storage.

Enums

Error

Error type for OpenData Vector operations.

Type Aliases

Result

Result type alias for OpenData Vector operations.