Crate vectx_core 

vectX Core

Core library for the vectX vector database.

This crate provides the fundamental data structures and algorithms:

• Vector - Dense vector representation with SIMD operations
• Point - A vector with ID and optional payload
• Collection - Container for points with indexing
• HnswIndex - HNSW approximate nearest neighbor index
• BM25Index - Full-text search with BM25 ranking

Example

use vectx_core::{Vector, Point, PointId, Collection, CollectionConfig, Distance};

// Create a collection
let config = CollectionConfig {
    name: "test".to_string(),
    vector_dim: 3,
    distance: Distance::Cosine,
    use_hnsw: true,
    enable_bm25: false,
};
let collection = Collection::new(config);

// Insert a point
let vector = Vector::new(vec![1.0, 0.0, 0.0]);
let point = Point::new(PointId::String("p1".to_string()), vector, None);
collection.upsert(point).unwrap();

// Search
let query = Vector::new(vec![1.0, 0.0, 0.0]);
let results = collection.search(&query, 10, None);

Re-exports

pub use collection::Collection;

pub use collection::CollectionConfig;

pub use collection::Distance;

pub use collection::PayloadIndexType;

pub use vector::Vector;

pub use error::Error;

pub use error::Result;

pub use point::Point;

pub use point::PointId;

pub use point::VectorData;

pub use point::SparseVector;

pub use hnsw::HnswIndex;

pub use graph::Node;

pub use graph::Edge;

pub use graph::NodeId;

pub use graph::EdgeId;

pub use bm25::BM25Index;

pub use filter::Filter;

pub use filter::PayloadFilter;

pub use filter::FilterCondition;

pub use multivector::MultiVector;

pub use multivector::MultiVectorConfig;

pub use multivector::MultiVectorComparator;

Modules

background

bm25

collection

error

filter

graph

hnsw

multivector

MultiVector support for ColBERT-style late interaction retrieval

point

simd

SIMD-optimized vector operations

vector