Zero Dependencies
In-Memory
TypeScript

Vector search for
modern applications

Lightning-fast in-memory vector database with automatic OpenAI embeddings. Build semantic search, recommendations, and AI features in minutes.

Quick Start
npm install neuradb

import { NeuraDB } from 'neuradb'

const db = new NeuraDB()
await db.addDocument({
  content: "AI will transform software",
  metadata: { category: "tech" }
})

const results = await db.search("artificial intelligence")

Documentation

Everything you need to integrate NeuraDB into your application

AI-Powered

Automatic OpenAI embeddings with multiple model support

Zero Dependencies

Lightweight library with no external dependencies

High Performance

Optimized in-memory operations for fast retrieval

Type Safe

Full TypeScript support with comprehensive types

Installation
Get started with NeuraDB in seconds
npm install neuradb
Quick Start
Basic usage example
import { NeuraDB } from 'neuradb'

// Initialize database
const db = new NeuraDB()

// Add documents
await db.addDocument({
  id: 'doc1',
  content: 'Machine learning transforms data analysis',
  metadata: { category: 'AI', difficulty: 'intermediate' }
})

// Search with natural language
const results = await db.search('artificial intelligence', {
  limit: 5,
  threshold: 0.7,
  similarityMethod: 'cosine'
})
Advanced Usage
Using with OpenAI integration
import { NeuraDB } from 'neuradb'
import OpenAI from 'openai'

const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY })
const db = new NeuraDB({
  openai,
  embeddingModel: 'text-embedding-3-small',
  defaultBatchSize: 100
})

// Batch add with automatic embeddings
await db.addDocuments(documents, {
  createEmbedding: true,
  batchSize: 50,
  onProgress: (processed, total) => {
    console.log(`Progress: ${processed}/${total}`)
  }
})
Similarity Methods
Choose the right similarity calculation for your use case

Cosine Similarity

Measures the cosine of the angle between vectors. Best for text similarity.

Recommended

Euclidean Distance

Measures straight-line distance between points. Good for numerical data.

Numerical

Dot Product

Measures vector alignment. Useful for recommendation systems.

Recommendations