@smartpointstech/core

High-performance vector database with HNSW indexing - built in Rust for AI/ML similarity search and semantic search applications

Overview

This package provides native Node.js bindings to the spt-core Rust crate. It includes a vector database with HNSW (Hierarchical Navigable Small World) indexing for approximate nearest-neighbor search, plus a graph database for labeled node/edge operations. A stub fallback is available when native bindings are not installed.

Installation

npm install @smartpointstech/core

Key Features

• HNSW Vector Search -- fast approximate nearest-neighbor queries with configurable ef parameters
• Multiple Distance Metrics -- Cosine, Euclidean, DotProduct, Manhattan
• Quantization Support -- scalar, product, and binary quantization for memory efficiency
• Graph Database -- labeled nodes and typed edges with property storage
• Batch Operations -- bulk insert for high-throughput ingestion

Dependencies

No internal npm dependencies. Uses platform-specific native binaries via optional dependencies (e.g., spt-core-darwin-arm64). Built with @napi-rs/cli.

Quick Start

import { VectorDB } from '@smartpointstech/core';

const db = VectorDB.withDimensions(384);
await db.insert({ vector: embedding, metadata: JSON.stringify({ text: 'hello' }) });
const results = await db.search({ vector: queryVec, k: 10 });

API Reference

Exports

Export	Type	Description
VectorDB	class	Vector database with HNSW indexing
GraphDB	class	Labeled property graph database
VectorEntry	interface	Entry with vector, optional id and metadata
SearchQuery	interface	Query with vector, k, efSearch, filter
SearchResult	interface	Result with id, score, optional vector/metadata
DbOptions	interface	Config: dimensions, distance metric, HNSW params
HnswConfig	interface	HNSW tuning: m, efConstruction, efSearch
QuantizationConfig	interface	Quantization type and parameters
getImplementationType	function	Returns 'native' or 'stub'