velesdb-core

High-performance vector database engine written in Rust.

Features

• Blazing Fast: HNSW index with explicit SIMD (4x faster than auto-vectorized)
• Hybrid Search: Combine vector similarity + BM25 full-text search with RRF fusion
• Persistent Storage: Memory-mapped files for efficient disk access
• Multiple Distance Metrics: Cosine, Euclidean, Dot Product, Hamming, Jaccard
• ColumnStore Filtering: 122x faster than JSON filtering at scale
• VelesQL: SQL-like query language with MATCH support for full-text search
• Bulk Operations: Optimized batch insert with parallel HNSW indexing

Installation

cargo add velesdb-core

Quick Start

use velesdb_core::{Database, DistanceMetric, Point};
use serde_json::json;

fn main() -> Result<(), Box<dyn std::error::Error>> {
    // Create a new database
    let db = Database::open("./my_vectors")?;

    // Create a collection with 384-dimensional vectors
    let collection = db.create_collection("documents", 384, DistanceMetric::Cosine)?;

    // Insert vectors with metadata
    let points = vec![
        Point::new(1, vec![0.1; 384], Some(json!({"title": "Hello World", "category": "greeting"}))),
        Point::new(2, vec![0.2; 384], Some(json!({"title": "Rust Programming", "category": "tech"}))),
    ];
    collection.upsert(&points)?;

    // Vector similarity search
    let query = vec![0.15; 384];
    let results = collection.search(&query, 5)?;

    for result in results {
        println!("ID: {}, Score: {:.4}", result.point.id, result.score);
    }

    // Hybrid search (vector + full-text)
    let hybrid_results = collection.hybrid_search(
        &query,
        "rust programming",
        5,
        Some(0.7) // 70% vector, 30% text
    )?;

    Ok(())
}

Distance Metrics

Performance

Vector Operations (768D)

End-to-End Benchmark (10k vectors, 768D)

Key Performance Features

• Search latency: < 5ms for 10k vectors
• Bulk import: 3,300 vectors/sec with upsert_bulk()
• ColumnStore filtering: 122x faster than JSON at 100k items
• Memory efficient with SQ8 quantization (4x reduction)

📊 Benchmark kit: See benchmarks/ for reproducible tests.

License

Elastic License 2.0 (ELv2)

See LICENSE for details.