Installation

npm install @neural-tools/vector-db

Install provider-specific dependencies:

# For Pinecone
npm install @pinecone-database/pinecone

# For Qdrant
npm install @qdrant/js-client-rest

# For ChromaDB
npm install chromadb

Usage

Pinecone Example

import { VectorDB } from '@neural-tools/vector-db';

const db = new VectorDB({
  provider: 'pinecone',
  apiKey: process.env.PINECONE_API_KEY,
  index: 'my-index'
});

// Store embeddings
await db.upsert([
  {
    id: '1',
    values: [0.1, 0.2, 0.3, ...],
    metadata: { text: 'Sample document' }
  }
]);

// Query similar vectors
const results = await db.query({
  vector: [0.1, 0.2, 0.3, ...],
  topK: 5
});

Local Storage Example

const db = new VectorDB({
  provider: 'local',
  path: './vectors.json'
});

// Same API, works offline
await db.upsert([...]);
const results = await db.query({...});

Supported Providers

Pinecone

• Serverless vector database
• High performance at scale
• Automatic scaling

Qdrant

• Open-source vector search engine
• Self-hosted or cloud
• Advanced filtering

ChromaDB

• Developer-friendly API
• Built for LLM applications
• Easy local development

Local Storage

• JSON-based persistence
• No external dependencies
• Perfect for development and testing

API Reference

VectorDB

Main class for vector operations:

const db = new VectorDB(config);

Methods

• upsert(vectors) - Insert or update vectors
• query(params) - Search for similar vectors
• delete(ids) - Remove vectors by ID
• fetch(ids) - Retrieve vectors by ID

Related Packages

• @neural-tools/semantic-cache - Semantic caching
• @neural-tools/core - Core utilities