Content

Everything you publish is MDX, and it lives in one place: web/next/content/. Under it, docs/ and blog/ are two Fumadocs collections that split on a single axis: who owns each page's metadata.

• Docs declare their structure and metadata centrally, in web/next/docs.config.ts. You write only the body.
• Blog posts carry their own metadata in frontmatter, and the listing builds itself from the posts.

Docs: the config is the source of truth

web/next/docs.config.ts is the single source for docs structure and metadata. Each collection maps to ordered sidebar groups, and every entry is keyed by the page's URL:

Manage: [
  {
    "/docs/manage/content": { title: "Content", label: "Content (Blog & Docs)", description: "…" },
  },
]

title and description are the page metadata; label overrides the sidebar text when it should differ from the title. A generator (.github/scripts/docs.ts) runs inside the web/next build and dev and, from that config:

• writes each page's frontmatter (slug, title, description, and label when set), so you only author the body;
• generates content/docs/meta.json: the reading order and prev/next;
• scaffolds a stub MDX in dev for any page declared in the config but missing on disk;
• fails a --strict build if the config and the files on disk have drifted.

So adding a page is three steps: add the entry to docs.config.ts, run the dev server (or bun .github/scripts/docs.ts) to scaffold the file, then write the body. Leave the frontmatter alone; it is regenerated on every build.

Blog: one file per post

A blog post is a single MDX file in web/next/content/blog/, and it owns its metadata in frontmatter. There is no list and no meta.json to maintain by hand; the /blog listing derives itself from the posts, newest first.

Two timestamps decide visibility: createdAt is bookkeeping; publishedAt is consent to publish. A post is public only when it is not a draft and publishedAt is set and not in the future.

---
title: "My Post Title"
description: "A short summary of the post."
createdAt: 2026-06-11
publishedAt: 2026-06-11
---

Set draft: true to hide a post regardless of its timestamps, and omit publishedAt for a draft with no release time yet. Timestamps are either YYYY-MM-DD (the start of that UTC day) or an ISO datetime with an offset ("2026-06-18T09:00:00+05:30") for a post that should go live at an exact time. Scheduled posts need no rebuild: every public blog surface re-evaluates publishing on render, so a post appears within about 60 seconds of its publishedAt as ISR regenerates. Post images are committed under web/next/public/blog/<slug>/images/ and referenced by absolute path.

Each post page renders its own publish and update dates and emits article Open Graph metadata from that same frontmatter. Running bun .github/scripts/docs.ts stamps a missing createdAt with today's date on a new post file and prints what it wrote, so validation always has one. Review and commit that line.

The publishing rule lives in one place (web/next/src/lib/blog-policy.ts) and the listing, sitemap, llms.txt, and OG routes all read it, so they can never disagree about what is public.

Search

Full-text search is powered by Orama through Fumadocs, built by the route handler at web/next/src/app/api/search/route.ts. Press Cmd+K (or Ctrl+K) on any docs page to open it. Only the docs collection is indexed; the blog and the private console docs are not.

The private console docs

web/next/source.config.ts defines a third collection, consoleDocs (content/console/docs/), served only under the auth-gated /console area. Its source is deliberately never referenced by a public route, so it is excluded from search, the sitemap, and llms.txt by construction: a home for internal docs that ship with the repo but stay behind the console.

Next

• SEO & Metadata: the OG images, sitemap, and robots that cover everything you publish.
• llms.txt: how the docs collection becomes agent-readable context.