# Feeds & Machine Output RSS, llms.txt, sitemaps, and Markdown content negotiation for anything reading the site programmatically. Every long-form surface on this site is readable without a browser. There are three ways in, in rough order of how much you care about presentation. ## RSS | Feed | Path | Contents | Rendering | | --- | --- | --- | --- | | Blog | `/blog/rss.xml` | Full posts from 無人之境, newest first. | Prerendered at build time | | Mood | `/mood/rss.xml` | The short-form feed, newest 50 items. | Rendered per request | Both are `application/rss+xml`, but they are not built the same way and it matters if you poll them. `/blog/rss.xml` is a static asset baked at build time. Its contents change only on deploy, so polling it more often than the site ships is pure waste. `/mood/rss.xml` is server-rendered on every request against the D1 archive and capped at 50 items. It answers `Cache-Control: public, max-age=0, s-maxage=300` — no browser caching, five minutes at the Cloudflare edge — so a new mood shows up within about five minutes without a deploy. A failure renders as a plain-text `500` rather than an empty feed, so a reader keeps the last good copy instead of silently emptying your subscription. ## Markdown pages Append `/index.md` to a supported page URL and it returns Markdown directly, without requiring a special request header. Pages advertise this explicit URL in their ``, so agents and crawlers can discover it: ```bash curl https://buxx.me/docs/writing/poem/index.md curl https://buxx.me/blog/index.md ``` The shorter `.md` form permanently redirects to the explicit alternate: `/docs/writing/authors.md` becomes `/docs/writing/authors/index.md`. Content negotiation remains available. Send `Accept: text/markdown` to the canonical page URL and it returns the same Markdown instead of HTML: ```bash curl -H 'Accept: text/markdown' https://buxx.me/blog curl -H 'Accept: text/markdown' https://buxx.me/mood ``` Pages that carry a Markdown representation advertise it in their ``: ```html ``` Supported pages: | Path | Returns | | --- | --- | | `/` | Profile, projects, recent posts. | | `/blog` | The post index. | | `/blog/{slug}` | One post, in full. | | `/blog/tags`, `/blog/tag/{slug}` | Tag directory and tag archives. | | `/mood` | The feed, paginated by cursor. | | `/mood/{id}` | One mood post. | | `/privacy` | The privacy policy. | | `/docs`, `/docs/{path}` | The documentation index and source content. | Every page under `/docs` also carries a **Copy page** control beside its breadcrumb. It fetches that page's `/index.md` and writes it to the clipboard; the menu beside it opens the same URL in a tab for reading. Responses carry an `x-markdown-tokens` header with an approximate token count of the body, so a client can budget before it reads. Anything else falls through to HTML. Header negotiation is strict about quality values: `Accept: text/html, text/markdown;q=0.9` gets you HTML, as it should. ## llms.txt ``` GET /llms.txt ``` A Markdown map of the site following the [llms.txt](https://llmstxt.org/) convention — a title, a one-line summary, then link sections with a short note on why a model would open each one. It is generated from the same site data the pages render from, so it does not drift. Use it as the entry point: read `llms.txt`, pick a URL, then append `/index.md` or fetch the canonical URL with `Accept: text/markdown`. ## Sitemap ``` GET /sitemap.xml ``` Standard sitemap index. Deliberately narrow — it lists the canonical public pages and skips preview routes, embeds, and API endpoints.