During development, the index.html is not a result of server-side rendering, but rather the Qwik app is built using client-side JavaScript only. This is ideal for development with Vite and its ability to reload modules quickly and on-demand. However, this mode is only for development and does not showcase "how" Qwik works since JavaScript is required to execute, and Vite imports many development modules for the app to work.

pnpm dev

Server-side rendered index.html, with client-side modules prefetched and loaded by the browser. This can be used to test out server-side rendered content during development, but will be slower than the client-only development builds.

pnpm dev.ssr

A production build should generate the client and server modules by running both client and server build commands.

pnpm build

The docs build also generates these public LLM-facing files:

• dist/llms.txt
• dist/llms-ctx.txt
• dist/llms-ctx-full.txt
• curated markdown mirrors, such as dist/docs/getting-started.md

You can regenerate them directly with:

pnpm generate.llms

To emit links for a different host or subdomain, set QWIK_LLMS_BASE_URL when generating:

QWIK_LLMS_BASE_URL=https://v2.qwik.dev pnpm generate.llms

On Cloudflare Pages, the generator also falls back to CF_PAGES_URL, so preview deployments can emit links for their deployment URL automatically. Set QWIK_LLMS_BASE_URL in the Pages production environment if you want production builds to emit a custom domain instead of the default *.pages.dev URL.

pnpm build runs the site build first and then writes the generated LLM files into dist/. The generation is chained to the build.client script (vite build && pnpm generate.llms), so the files exist in dist/ before the server build runs. This ensures the Cloudflare Pages adapter discovers them as static paths and the Worker serves them correctly.

If you add or want to expose a new docs or API page in the LLM outputs, update the curated manifest in scripts/generate-llms.ts.

For each new entry, add:

• section
• title
• pathname
• sourcePath
• description
• optional optional flag

After updating the manifest, run:

pnpm generate.llms

or just:

pnpm build

Notes:

• Editing an existing page that is already in the manifest is automatic; rebuilding regenerates the LLM files from the latest source.
• Adding a brand new page to the LLM surface is not automatic yet; it must be added to the manifest on purpose.

Production build that creates only the client-side modules that are dynamically imported by the browser.

pnpm build.client

Production build that creates the server-side render (SSR) module that is used by the server to render the HTML.

pnpm build.ssr

Cloudflare's wrangler CLI can be used to preview a production build locally. To start a local server, run:

pnpm serve

Then visit http://localhost:8787/

Cloudflare Pages are deployable through their Git provider integrations.

If you don't already have an account, then create a Cloudflare account here. Next go to your dashboard and follow the Cloudflare Pages deployment guide.

Within the projects "Settings" for "Build and deployments", the "Build command" should be pnpm build, and the "Build output directory" should be set to dist.

Cloudflare Page's function-invocation-routes config can be used to include, or exclude, certain paths to be used by the worker functions. Having a _routes.json file gives developers more granular control over when your Function is invoked. This is useful to determine if a page response should be Server-Side Rendered (SSR) or if the response should use a static-site generated (SSG) index.html file.

By default, the Cloudflare pages adaptor does not include a public/_routes.json config, but rather it is auto-generated from the build by the Cloudflare adaptor. An example of an auto-generate dist/_routes.json would be:

{
  "include": [
    "/*"
  ],
  "exclude": [
    "/_headers",
    "/_redirects",
    "/build/*",
    "/favicon.ico",
    "/manifest.json",
    "/service-worker.js",
    "/about"
  ],
  "version": 1
}

In the above example, it's saying all pages should be SSR'd. However, the root static files such as /favicon.ico and any static assets in /build/* should be excluded from the Functions, and instead treated as a static file.

In most cases the generated dist/_routes.json file is ideal. However, if you need more granular control over each path, you can instead provide you're own public/_routes.json file. When the project provides its own public/_routes.json file, then the Cloudflare adaptor will not auto-generate the routes config and instead use the one committed within the public directory.

STILL WIP

resource: https://docsearch.algolia.com/

Setup in https://crawler.algolia.com/

To crawl localhost site for testing index settings for content hierarchy. use this docker command

# create apiKey via https://www.algolia.com/account/api-keys
touch .env
# APPLICATION_ID=APPLICATION_ID
# API_KEY=API_KEY
docker run -it --rm --env-file=.env -e "CONFIG=$(cat ./packages/docs/algolia.json | jq -r tostring)" algolia/docsearch-scraper

see guide of DocSearch-legacy docker command

In mac machines, docker containers can't access host's network directly, a workaround is to use host.docker.internal`