docs: release notes for v3.3.0-beta

docs/release-notes.md

@@ -15,43 +15,6 @@ Body, in this order. Omit a section only when it would be empty, except Upgradin
 
 Rules: bullets are full sentences ending with a period; user-facing wording, no commit prefixes; sentence-case headings; no horizontal rules, no metadata block, no "Pre-1.0" boilerplate (the prerelease flag says it); every release ships all 6 binaries from `task release`.
 
-## Next release (draft)
-
-Articles and sites now store the data a website needs for search and sharing, and every article gets an address that is unique on its site.
-
-### New
-
-**Article SEO fields**
-
-- An article has a slug, an excerpt, a search title and description, a cover image with its alt text, a social title, description and image, a canonical address, a noindex flag and a list of dismissed suggestions.
-- The slug is generated from the title, is unique per site and can be changed at any time; an empty slug regenerates it from the title.
-- A site has a site name, a title template and a default description.
-
-### API
-
-- `PATCH /v1/article/{id}` also takes the SEO fields. `cover_image_id` and `social_image_id` accept `null`. A slug already used on the target site answers `409` `slug_taken` with the header `X-Slug-Suggestion` holding the first free slug.
-- `GET /v1/article/{id}`, `POST /v1/article/`, `PATCH /v1/article/{id}`, `PATCH /v1/article/{id}/publish` and `PATCH /v1/article/{id}/unpublish` return the twelve SEO fields; the items of `GET /v1/article/` return `slug` and `excerpt`.
-- `POST /v1/site/` and `PATCH /v1/site/{id}` take `site_name`, `title_template` and `default_description`; a template without `%title%` is `400` `invalid_title_template`.
-- Changed: `GET /v1/media/{id}/usage` stays an array, and every item gains `type` and `fields` without losing `id`, `title` and `published`. Authors and sites that use the image are now listed, without a `published` key.
-- Changed: a cover image, a social image, an author avatar and a site default social image all count in `used_in` and in the `unused` filter. Deleting an image clears the covers and social images that use it, and merging or purging repoints them.
-
-### New: SEO output
-
-- `GET /v1/article/{id}/seo` returns the resolved search and sharing data of an article (title with the site template, description, canonical address, robots, Open Graph, Twitter, image, authors and ready JSON-LD), so a frontend can render its meta tags without reimplementing the fallback chain. Drafts answer verified callers only.
-- `POST /v1/article/{id}/seo/check` evaluates an article without saving, with optional overrides, and reports the suggestions, the status and whether the slug is free.
-- `GET /v1/article/` items carry `seo_status` and `seo_suggestions`, and the new `seo` parameter filters on the status.
-- `GET /v1/site/{key}/sitemap` lists the published URLs of a site, paged, with the total in `X-Total-Count`, which is now exposed to browsers through CORS.
-- `GET /v1/article/` counts gain `suggestions`, the number of visible articles that still have SEO suggestions.
-- `GET /v1/site/` takes `q`, `page` and `size` and sends the number of matching sites in `X-Total-Count`; without `page` and `size` every site is still returned, as a JSON array.
-- `GET /v1/author/` also sends its total in `X-Total-Count`.
-- `GET /v1/media/{id}` reads one image, with the fields of a media list item; a missing image answers `404` and a malformed id `400`.
-
-### Upgrading
-
-- The first start after the upgrade adds the new columns, gives every existing article a slug generated from its title (in id order within each site, with `-2`, `-3` on conflicts) and creates the unique `(site_id, slug)` index. Nothing else changes, and a second start changes nothing.
-- A deleted article now frees its slug.
-- The slug backfill writes in batches of 500 rows, so a large database is migrated without one long transaction, and a start that was interrupted resumes where it stopped.
-
 ## Published releases
 
 Each body lives in `docs/releases/<tag>.md`. Publish with `tea releases create --tag <tag> --title "<tag> - <theme>" --note-file docs/releases/<tag>.md --prerelease --asset <file>` after `task release`, with one `--asset` per file: the six binaries, `bin/checksums.txt` and `bin/checksums.txt.sig`. `task release` produces all eight and fails before building when the signing key is missing or wrong (see [deployment.md](deployment.md#release-signing-key-maintainers-once)). Never publish a release without the two checksum files.
@@ -76,3 +39,4 @@ Each body lives in `docs/releases/<tag>.md`. Publish with `tea releases create -
 | `v3.1.0-beta` | Image variants |
 | `v3.2.0-beta` | Article lists, editor links and activity export |
 | `v3.2.1-beta` | Paged media library and remembered page sizes |
+| `v3.3.0-beta` | Sites, authors and search and sharing |