It's 5:47am and your phone buzzes. You want to keep sleeping, but you know that sound means only one thing: six inches of fresh powder at Grand Targhee overnight, temperature sitting at 18°F. Exactly the conditions you told the app to let you know about. You roll out of bed, pour coffee directly into your throat, and drive straight to the mountain.

In this course, you'll build Ski Alerts, the SvelteKit 3 app behind that notification. It streams AI chat responses and parses natural language into structured alert rules. A background workflow evaluates those rules against live weather data, and Vercel hosts the app.

What you'll build

You'll take Ski Alerts from its first deployment through these production features:

Deployment

Configure and deploy a SvelteKit app to Vercel
Set up environment variables across development, preview, and production
Implement preview deployments for team collaboration

AI features

Build streaming chat interfaces with AI SDK v6
Create tools and multi-step agents
Extract structured data with Valibot schemas

Background processing

Build durable workflows with the Workflow SDK
Run parallel steps and schedule re-checks with sleep
Handle errors with FatalError, RetryableError, and exponential backoff

Production

Configure ISR to cache pages
Set up observability and logging
Optimize performance for real users

Prerequisites

Familiarity with SvelteKit basics and the official tutorial
Node.js 24 and npm (or pnpm) installed
SvelteKit 3 and Svelte 5; the starter includes the compatible Vite 8, TypeScript 6, and Vercel adapter 7 dependencies
A Vercel account (free tier works)

Svelte on Vercel

What you'll build

Deployment

AI features

Background processing

Production

Prerequisites

Course sections

Section 1: Deployment Foundations

Section 2: AI Gateway

Section 3: Workflows

Section 4: Production