Cloud / Architecture / Build note
A technical portfolio does not need an application server.
The boundary between building a site and serving it—and why this portfolio keeps those responsibilities separate.
A description of this Astro and Cloudflare Static Assets setup. Hosting plans and platform limits can change.
Start with what happens on a request.
A visitor to this portfolio needs a page, a few locally hosted images and fonts, and links to work. There is no login, personalised data, or server-side conversation. The AI animation is an illustration, not an inference endpoint. Those requirements do not need an application process behind every request.
Astro builds the routes ahead of time. Cloudflare serves the resulting static assets. The configuration deliberately has no runtime Worker entry point, database binding, or on-demand rendering adapter. This is a workload choice, not a rule that every site should avoid a backend.
Build and serve are different responsibilities.
The source repository is not the published site. A clean build must produce the output directory before the deployment command can upload it. In this setup, the hosted build runs the tests and Astro build; the deploy step then publishes the static output.
That explicit boundary avoids relying on generated output checked into Git or on a local development folder that does not exist in a fresh build environment. The important verification is the produced HTML, routes, headers, and assets—not simply whether the source compiled.
{
"assets": {
"directory": "./dist",
"not_found_handling": "404-page"
}
}Static does not mean motionless.
Theme switching, a keyboard command palette, native dialogs, and an animated SVG can all run in the browser. The page still has real navigation and readable content without those enhancements. Images and fonts are shipped locally rather than fetched from an unrelated service during a visit.
The distinction matters when deciding what to add next. A notes page is another generated document. A contact form with storage or an AI chatbot would introduce a different data flow, operating responsibility, and cost model. They should not be added merely because the portfolio is about technology.
Verify the boundaries you chose.
The test suite checks internal links, one main heading per page, canonical URLs, the sitemap, and the custom 404. It also checks that excluded project routes remain absent and that raw source or original camera files are not part of the published output. A nonpublishing Cloudflare dry run validates the asset package before a release.
After a push, the remaining question is whether the public domain actually serves the new release. A successful Git operation is not evidence that the hosted build has finished. Checking the live HTML and asset responses closes that gap without inventing a deployment status.
References & further reading
Have a counterexample or a better approach? That is a good reason to talk.
Discuss this noteExplore the engineering approach