Backend Terms for Vibe Coding

I want people to be able to type an easy-to-remember name to open my website.

I bought a domain, so why do I still need to set up DNS? Help me point it to my server.

What's that long string in the address bar? I want every page to have its own permanent address.

When I click a button, what actually happens between the browser and the server?

Login succeeds, but a refresh signs the user out. Help me check whether the session cookie is set and sent back with requests.

My browser says the site isn't secure. How do I get that lock icon in the address bar?

People far away say the images load slowly. Can they download them from somewhere closer?

The local app runs on port 3000 and the API on 4000. Help me confirm which one the browser reaches.

The old event link should take people to the new page instead of showing a 404.

My website needs data from the backend. People keep saying I need an API—what is that?

AI keeps returning data with curly braces inside curly braces. How do I read and write it?

The browser blocked this page's API request. Check whether CORS is not allowing this site.

People can submit complete nonsense. How do I stop bad input before it gets saved?

After upload, do not keep the page waiting. Enqueue the transcoding task, return a task ID, and let me check its status later.

A payment callback may arrive more than once, but the same order must count as paid only once.

The message was sent, but the queue keeps growing. Check whether a consumer is receiving, processing, and acknowledging the messages.

Keep this draft after refresh, but do not treat it as a database shared by every user.

The content people submit disappears after a refresh. We need somewhere to actually save it.

I need to add a phone column to our user table, but production already has tens of thousands of rows. How do we update the schema without losing data or causing downtime?

How do I write a query for users who placed an order in the last seven days?

Creating an order and deducting stock must succeed together; if any step fails, it must not leave half an order behind.

The API says saved, but the nickname returns to its old value after refresh. Confirm the write really committed.

AI suggested using Supabase and PostgreSQL for my users and orders; how is that different from saving files?

Where does the part users can't see run, and who handles things like login and saving data?

Is there a ready-made framework for building the backend, so I don't have to start from scratch?

When people visit different URLs, how do I send each one to the right code?

When two users buy the last item at once, there cannot be two successful orders; checking and deducting stock must act as one whole operation.

The image worker crashes and takes down the whole site. Put that work in a separate process and verify the web service stays available.

Two requests update the same cache. Confirm whether they are threads in one process and make the shared-state order explicit.

Do not block the execution path while an endpoint waits on external services. Suspend the coroutine and let other work proceed.

Don't make the user stare at a spinner while generating this huge report. Put the job into a message queue and let background workers process it.

A user is only previewing a price. Do not silently write an order or send email; do that only after explicit submission.

Two requests decrement the last item concurrently. Lock the shared record so one proceeds and the other waits and rechecks.

Coupons occasionally oversell while one test run looks normal. Reproduce the race with concurrent timing, fix it, and verify repeatedly.

When downstream processing cannot keep up, do not let the queue and memory grow forever. Slow or pause the upstream and resume after recovery.

Two scheduler instances must not run the same summary. Let one hold time-limited ownership, let it expire after failure, and let the other take over.

Orders, inventory, and payment run in different services. When a service or network fails, show exactly how far the order got.

How do I verify that someone is who they say they are? I need to add sign-in.

After people sign in, I still need different access levels for admins and regular users.

An administrator changed a permission. Show who changed what, when, and whether the action succeeded or was denied.

A user bought a Pro subscription; how should the backend check if they can access this export feature?

The new styles are deployed, but some people still see the old version. Find out which cache layer has not updated.

Help me clarify this product article's title, body hierarchy, and links so both readers and search engines can understand it.

Keep the product's campaign URLs accessible, but declare the parameter-free URL as canonical and use it in the sitemap.

The new article is live but has few internal links. Generate a sitemap.xml containing only valid canonical URLs.

Limit crawler access to duplicate filter paths, but do not use robots.txt as permission for the admin area.

This article URL no longer exists. Make a real 404 page that still lets people return to the article list.

I know a secret key shouldn't go in the code. Where should I put it?

I finished the site. How do I make it available at a URL that other people can open?

Confirm that this project has hosting that can continuously serve the site, then deploy it publicly. Do not treat localhost as live.

Send this CI-passing version to testing first; production needs human approval unless we explicitly choose continuous deployment.

Deploy the candidate to staging with production-like configuration and de-identified order data; do not run destructive verification in production.

After release, monitor login error rate and checkout success rate; alert above a threshold and show which version was running then.

When a new version sends login error rate sharply up, roll back to the last known-good version; then confirm both running version and metrics recover.

Open the new recommendation feature to employees first; if error rate rises, turn off the new path immediately without redeploying.

Send 5% of users to the new release first; expand only while error rate stays healthy, and return to the stable release on anomaly.

Keep blue serving while deploying and accepting the new release on green; switch production traffic only after it passes.

I only have a small feature. Is there a way to run code on demand without buying or managing a server?

There's an error in production. Where can I see what happened on the server?

The payment dependency keeps timing out. Stop making every request wait, fail fast or degrade, and probe it again after a pause.

Run this service as one instance and explain the short outage during restart, while separating in-memory task risk from durable data.

Run several instances to share requests, but keep users logged in when their next request reaches another instance.