|
2 | 2 |
|
3 | 3 | Create `content/changelog/<slug>/index.mdx`. The slugs `archive` and `preview` are reserved. The website, archive, individual pages, sitemap, and RSS all read published entries from this directory through the shared content registry. |
4 | 4 |
|
5 | | -The index shows up to 12 updates in chronological rows. When more exist, an Older updates link opens the archive at the first omitted entry. The archive links to every published update. Both lists read metadata only; the full article and its recordings are loaded on the individual update page. Rows use the local `ogImage` and `ogAlt`, so a reviewed demo poster also becomes the update's cover and share image. The page frame, typography, media radius, and row spacing use Sim's shared landing styles. |
| 5 | +The index shows up to 12 updates as dates and linked titles, without repeated summaries, thumbnails, or card borders. When more exist, an Older updates link opens the archive at the first omitted entry. The archive links to every published update. Both lists read metadata only; the full article and its media are loaded on the individual update page. Keep the concise description for search, sharing, and RSS. The reviewed poster supplies the share image; it does not need to repeat on the index. Article headers show the title and release date without a personal byline or another summary. The page frame, typography, media radius, and row spacing use Sim's shared landing styles. |
6 | 6 |
|
7 | 7 | Write for a team using Sim: what changed, what they can do with it, and where to find it. Group related PRs into one product update. Keep maintenance, refactors, and routine content edits in the technical GitHub release history. |
8 | 8 |
|
@@ -44,13 +44,15 @@ Give readers the next useful step in the sentence that introduces a capability. |
44 | 44 |
|
45 | 45 | Use the guide for the exact surface being announced. Live Search connectors have their own `/search/<source>` guides; an ordinary workflow integration page can describe a different connection and permission model. Feature announcements should link to the feature guide, and API changes should link to the current generated operation page. Related Sim product pages and explanatory blog posts are useful when they help a reader use or understand the update. |
46 | 46 |
|
47 | | -Choose a few relevant links instead of repeating every destination in a separate resource list. Keep link text descriptive. The content audit checks integration destinations against the generated catalog and docs destinations against the MDX source and generated OpenAPI pages. Use current canonical URLs, including the operation ID's exact case. Confirm the linked page's instructions and any deep anchor in the deployment preview; a source file alone does not establish rollout or anchor correctness. |
| 47 | +For a model update, link the specific `/models/<provider>/<model>` page, or the provider catalog when discussing several models. Use the actual generated route, not a slug guessed from the model's display name. Keep the workflow setup link alongside it when useful. |
| 48 | + |
| 49 | +Choose a few relevant links instead of repeating every destination in a separate resource list. Keep link text descriptive. The content audit checks integration and model destinations against their generated catalogs and docs destinations against the MDX source and generated OpenAPI pages. Use current canonical URLs, including the operation ID's exact case. Confirm the linked page's instructions and any deep anchor in the deployment preview; a source file alone does not establish rollout or anchor correctness. |
48 | 50 |
|
49 | 51 | ## Search and answer-engine visibility |
50 | 52 |
|
51 | 53 | Keep each summary independently useful: name Sim and the feature, explain the new capability, and state material availability limits in the article. Put setup steps and demonstration explanations in server-rendered text alongside the media. Do not leave important facts only inside a screenshot or video. |
52 | 54 |
|
53 | | -The index and archive use real links, semantic lists, and CollectionPage/ItemList data matching the visible entries. Articles keep their own canonical URLs, BlogPosting data, author bylines, original publication dates, and substantive correction dates. Public articles appear in the sitemap and RSS; previews stay noindex and drafts stay out of public routes. Preserve these properties when changing the layout. If the archive eventually needs pagination, use server-rendered pages with real next/previous links and a self-canonical URL for each page; do not make older entries accessible only through a JavaScript button. |
| 55 | +The index and archive use real links, semantic lists, and CollectionPage/ItemList data matching the visible entries. Articles keep their own canonical URLs, BlogPosting data, authorship metadata, original publication dates, and substantive correction dates. Public articles appear in the sitemap and RSS; previews stay noindex and drafts stay out of public routes. Preserve these properties when changing the layout. If the archive eventually needs pagination, use server-rendered pages with real next/previous links and a self-canonical URL for each page; do not make older entries accessible only through a JavaScript button. |
54 | 56 |
|
55 | 57 | These are standard search foundations, not a guarantee of indexing or AI citations. [Google's AI search guidance](https://developers.google.com/search/docs/appearance/ai-features) calls for crawlable, useful text, relevant media, internal links, and structured data matching the visible content; it does not require special AI-only markup. Avoid hidden keyword blocks, invented FAQs, and schema for information that the page does not show. |
56 | 58 |
|
|
0 commit comments