cameri/nostream nostream-website@1.3.0

on GitHub

Minor Changes

• #884 e654852 Thanks @chappie-daemon! - Add a social preview card generated from source.

  website/site/scripts/build-social-card.mjs renders one composition at two sizes:
  public/media/social-card.png (1200×630, the site's og:image) and
  public/media/github-social-preview.png (1280×640, the GitHub repository preview).
  Base.astro now emits the full preview set — og:image:width, og:image:height,
  og:image:alt, twitter:image, twitter:title, twitter:description,
  og:site_name, og:locale — and takes an ogImage prop that defaults to the card;
  hero-poster.webp stays the film's poster.

  The GitHub preview is uploaded by hand: Settings → Social preview in the
  repository, which no workflow can set.

Patch Changes

• #882 9146fd3 Thanks @Ferryx349! - Replace the home page admin screenshot gallery with an operator console film (explain beats on the hero backdrop, full-frame UI shots) and a pause control; add Framefields sources and ship admin.mp4 in site media.

• #889 eb68475 Thanks @chappie-daemon! - The site now lives at https://nostream.dev, and every other host redirects to it over https.

  The router worker answers on nostream.dev and www.nostream.dev as well as the preview
  domain, and CANONICAL_HOST decides which one serves content: any other host the routes
  catch — www, or ema.tnsor.network for a non-preview path — gets a 301 to the canonical
  host, and http gets a 301 to the same URL over https. One origin serving means one
  canonical URL, so www and the apex cannot split search ranking by both answering 200.

  Previews deliberately stay on ema.tnsor.network/preview/<n>/: that is why the domain
  remains bound. A preview build must never sit on the production domain, and a preview's own
  SITE_URL stays the preview origin so its absolute links and canonical tags cannot claim
  nostream.dev.

  The domain is an input now, not a constant: SITE_URL for the build, CANONICAL_HOST for
  the worker, both set in deploy-website.yml. Moving again is two lines rather than a search
  through the source, the social card and the README.

  Cloudflare-side setup this needs before it takes effect: the nostream.dev zone must
  exist, and both nostream.dev and www.nostream.dev resolve through the worker's
  declared custom domains on the next deploy. Turning on the zone's Always Use HTTPS
  setting makes the upgrade happen at the edge, before the worker; the worker's own 301
  covers the case where that setting is off.

• #888 07dc215 Thanks @chappie-daemon! - A pull request's site preview now deploys even when the comment cannot be
  written, and the CDN cache is purged after a deploy so a fresh preview is not
  served the previous build.

  A comment can no longer prevent a deploy. The preview workflow's comment
  steps ran before the deploy, so a pull request opened from a fork — whose
  GITHUB_TOKEN is read-only whatever the permissions: block grants — failed at
  GraphQL: Resource not accessible by integration (addComment) and skipped the
  deploy entirely (run 39134047965).
  The deploy now runs first, every comment step is continue-on-error, and the
  job grants issues: write (PR conversation comments are issue comments;
  pull-requests: write alone does not cover them).

  The cache is purged after deploying, scoped to https://ema.tnsor.network/preview/<n>/
  via the prefixes purge, so a stale window cannot be served — on 2026-10-18
  /preview/885/ returned production HTML for about three minutes after the
  build landed. The production workflow purges the whole zone after its router
  deploy. Both purges skip with a notice when CLOUDFLARE_ZONE_ID is not set for
  their environment, so an unconfigured environment deploys normally instead of
  failing.

• #885 f88a563 Thanks @Priyanshubhartistm! - Add maintainer bio and social links on the community page for Priyanshu Bharti.