vectors

High-performance vector database for storing and querying embeddings with support for similarity search, hybrid search, and real-time indexing.

Overview

The vectors primitive provides a specialized database optimized for vector embeddings, enabling fast similarity search, recommendation systems, and semantic retrieval at scale.

Parent Primitive: llm - Universal LLM interface

SDK Object Mapping

This primitive maps to the db and ai SDK objects for vector operations:

import { db, ai, vectors } from 'sdk.do'

// Create collection
const products = await vectors.collection('products', {
  dimensions: 1536,
  metric: 'cosine',
})

// Generate embedding and insert
const embedding = await ai.embed({
  text: 'High-performance laptop for developers',
  model: 'text-embedding-3-large',
})

await products.insert({
  id: 'product-123',
  vector: embedding,
  metadata: {
    name: 'Laptop',
    price: 999,
    category: 'Electronics',
  },
})

// Search by similarity
const queryEmbedding = await ai.embed({
  text: 'affordable laptop',
  model: 'text-embedding-3-large',
})

const similar = await products.search({
  vector: queryEmbedding,
  limit: 10,
  filter: { price: { lte: 1000 } },
})

Quick Example

import { vectors } from 'sdk.do'

// Create collection
const products = await vectors.collection('products', {
  dimensions: 1536,
  metric: 'cosine',
})

// Insert vectors
await products.insert({
  id: 'product-123',
  vector: embedding,
  metadata: {
    name: 'Laptop',
    price: 999,
    category: 'Electronics',
  },
})

// Search by similarity
const similar = await products.search({
  vector: queryEmbedding,
  limit: 10,
  filter: { price: { lte: 1000 } },
})

Core Capabilities

Fast Similarity Search - Cosine, Euclidean, and dot product metrics
Hybrid Search - Combine vector and keyword search
Metadata Filtering - Filter results by metadata fields
Real-Time Indexing - Immediate availability of new vectors
Scalable Storage - Billions of vectors with low latency

Access Methods

SDK

TypeScript/JavaScript library for vector operations

await vectors.collection('products').search({ vector: queryEmbedding, limit: 10 })

→ SDK Documentation

CLI

Command-line tool for vector database management

do vectors search products --vector-file query.json --limit 10

→ CLI Documentation

API

REST/RPC endpoints for vector operations

curl -X POST https://api.do/v1/vectors/products/search -d '{"vector":[...],"limit":10}'

→ API Documentation

MCP

Model Context Protocol for AI-driven vector search

Search the products collection for 10 similar items to the query vector

→ MCP Documentation

Parent Primitive

llm - Universal LLM interface

Sibling Primitives

models - AI model management
embeddings - Vector embeddings generation

ai - AI operations (SDK object mapping)
db - Database operations (SDK object mapping)
database - General data storage

vectors

On this page