Getting Started with NorthStar SDK: Your First AI-Native App

NorthStar SDK makes it easy to add AI capabilities to your applications. In this tutorial, we'll build a semantic search feature for a documentation site.

Installation

npm install @strug-city/northstar
# or
pip install northstar-sdk

Core Concepts

NorthStar provides three main abstractions:

Embeddings - Convert text to vectors

VectorStore - Store and search vectors

Agents - Orchestrate AI workflows

Building Semantic Search

search.ts

import { NorthStar, VectorStore } from '@strug-city/northstar';

const ns = new NorthStar({
  apiKey: process.env.NORTHSTAR_API_KEY,
});

// Create embeddings from documents
const docs = await ns.embeddings.create({
  input: documentTexts,
  model: 'text-embedding-3',
});

// Store in vector database
const store = new VectorStore('docs');
await store.add(docs);

// Search semantically
const results = await store.search({
  query: 'How do I deploy my app?',
  limit: 5,
});

console.log(results);

Adding Hybrid Search

Combine semantic and keyword search for better results:

const results = await store.search({
  query: 'deployment configuration',
  limit: 5,
  hybrid: {
    semantic: 0.7,  // 70% semantic
    keyword: 0.3,   // 30% keyword
  },
});

That's it! You've built a semantic search feature in less than 30 lines of code. Check out our full documentation for advanced features like multi-language support, custom embeddings, and agent orchestration.