0

Deploy on Vercel & observability

Productionused in ~78% of apps

Shipping is half the job; knowing what broke at 2am is the other half. The interview question that separates 'I deployed once' from 'I run things in production' is: what's the difference between a preview and a production deployment, and how do you find the request that 500'd?

Core idea. Every push to Vercel builds an immutable deployment with its own URL. Branch pushes become preview deployments (their own env vars, their own URL); merging to your production branch promotes one to production. Observability — Sentry for errors, Web Vitals for UX, build/runtime logs — is what turns 'it works on my machine' into 'I can prove it works for users and find it fast when it doesn't.'
Playground, this isn't a screenshot
1Push
2Build
3Preview
4Review
5Production

Repo connected to Vercel

You imported the GitHub repo once. Vercel auto-detected the framework (Vite -> build: vite build, output: dist) and now watches every branch. No CI to wire up.

deploymentsevery branch -> its own URL
mainProduction
pulse.acme.com
Live (previous build)
feat/likesPreview
pulse-git-feat-likes-acme.vercel.app
No deployment yet
feat/dark-modePreview
pulse-git-feat-dark-mode-acme.vercel.app
Ready (separate URL)

Each branch deploys in isolation to its own preview URL. Only main is wired to the production domain.

Connect the repo once, then git is the deploy button: push -> auto-build -> preview URL per branch -> merge to main -> production.

↑ play a little first, then dig in