The Northwind Coffee series builds a real site from an empty organization and ends with a static bundle: Vite, React, and the Content Delivery API called from the browser. No server of our own anywhere in it.
This is the same site with a server in front of it. Same content model, same seed bundles, same renderers, on Next.js. Five parts, because the content model is identical and restating it would waste your time — if you want to know how the model is built, that is the other series.
What this one is for is the question the static series deliberately left open: what actually changes, and is it worth it?
Start with what a static site cannot do
A static site has to publish its delivery key. There is no alternative — the code that reads content runs in the browser, so the credential it uses is in the bundle, and anyone can read it.
That is not a flaw we are working around. It is why ebitex has browser-safe keys at all: a key restricted to a set of origins and IP ranges, scoped to read published content and nothing else. The static sample ships one on purpose.
A server has a choice. Its key never leaves the machine, so it need not be restricted at all:
# northwind-coffee (static) — this key reaches the browser, so it must be browser-safe
VITE_CONTENT_DELIVERY_KEY=...
# northwind-coffee-ssr — this key never leaves the server, so it need not be restricted
CONTENT_DELIVERY_KEY=...
CONTENT_SITE_ID=...
In the sample that boundary is enforced rather than documented: lib/content.ts begins with import 'server-only', so a client component importing it is a build error instead of a quietly shipped credential. That is the shape to copy. A convention that only holds while everyone remembers it is not a boundary.
What else a server buys
The first paint is real HTML. A static site sends a shell and fills it in; a crawler that does not run JavaScript sees the shell. With a server, the content is in the document. Whether that matters depends entirely on who is reading your site — for a marketing site with organic search traffic it usually does, and for an internal tool it usually does not.
Freshness stops being bounded by deploys. On a static site, a page changes when you rebuild. With a server, it changes when a cache lets it. That is better, and it is also a new thing you have to own — enough of a subject that it is part 3.
Personalization can decide the first paint. The static sample keeps the visitor type in localStorage, reads it after hydration, and corrects the page. A server reads a cookie that arrived with the request and resolves the right content the first time. The difference is visible: no flash of the wrong words, and a crawler sees a coherent page rather than a half-personalized one.
What it costs
An instance that can cold-start. A static bundle on a CDN has no warm-up and essentially cannot fall over. A server has both properties in reverse, and you are now running something.
Headers and routing stop being declarative. On a static host, cache policy and redirects are a config file. Behind a server they are code, which is more flexible and also more places for them to be wrong.
A browser that cannot hold a key cannot query the Delivery API. This is the one that surprised us, and it is the subject of part 4. It sounds fatal and turns out to be much smaller than it looks — but the moment the key moves to the server, anything the browser was fetching for itself becomes something you serve.
The honest recommendation
Neither model is the real one. That is a rule we hold ourselves to rather than a diplomatic sentence: the SDK ships the capability and lets you own the plumbing, precisely so that a static build never has to wait for an adapter written for somebody else's stack.
The site you are reading this on is static. ebitex.io resolves its content in the browser, exactly like the Northwind static sample, and we have no plans to change that. If this series read as "and now do it properly", it would be selling against our own architecture.
Choose a server when you need one of the things above. Choose a static build otherwise, and do not feel behind.
Part 2 puts the resolve on the server.
