Skip to content

armgit / kiln

Public Imported from GitHub
0

Build and revise procedural 3D assets with your coding agent. Local MCP server, CLI and TypeScript engine with rendered review, editable source and GLB export.

8.2 KB
Raw Blame

Search indexing — September 6, 2026

Historical record. play.kilnstudio.tools has since been retired and no longer resolves, so the redirect rows for it below describe September 2026 only.

The owner reported Search Console's “Duplicate without user-selected canonical” warning for kilnstudio.tools. The domain property's report was inspected in the owner's browser. Its six affected URLs were last crawled August 28–September 1, before the current OSS site snapshot. Google's individual selected canonicals were not inspected; the current live responses were checked for all six.

Reported URL Current first response Current destination
https://www.kilnstudio.tools/ 301 Canonical HTTPS apex
http://www.kilnstudio.tools/ 301 Canonical HTTPS apex
https://play.kilnstudio.tools/ 302 GitHub Pages, then canonical HTTPS apex
https://play.kilnstudio.tools/pocket-settlement 302 Same redirect chain
https://play.kilnstudio.tools/tower-defense 302 Same redirect chain
https://play.kilnstudio.tools/chess-board 302 Same redirect chain

All six end at a 200 HTML document declaring https://kilnstudio.tools/ canonical. Search Console accepted Validate Fix and showed Validation started on September 6. This requests a fresh evaluation; it does not mean Google has finished validation or indexed every duplicate. Analytics configuration was not changed.

Live checks before this change

  • https://kilnstudio.tools/ returns 200 with exactly one static HTML canonical: https://kilnstudio.tools/.
  • HTTP, https://www.kilnstudio.tools/, and the GitHub Pages project address redirect to that HTTPS apex URL.
  • /index.html and a query-string variant return 200 and declare the same canonical root. They should consolidate into the root, not be indexed separately.
  • /robots.txt and /sitemap.xml returned 404.

This change adds a crawl-allowing robots file referencing a sitemap that contains only the canonical root. The existing canonical remains in the original HTML. The account already had /sitemap-index.xml submitted from the old site, with an error and zero discovered pages. That path is restored as a sitemap index pointing to the new root sitemap, so Google's existing submission remains usable. Hash routes such as #/gallery are views within one document; they are not listed as separate crawlable pages or given fragment canonicals. Individually indexed gallery pages would require real page URLs and distinct page content.

Google treats redirects and canonical links as strong signals and sitemap entries as a weaker, complementary signal. A missing sitemap alone does not establish the cause of this warning. Do not block duplicate URLs with robots.txt or add noindex to the main page to try to solve canonical selection.

Reference: Google's canonicalization guidance.

Remaining Search Console work

  1. Allow Google to recrawl. Deployment cannot guarantee indexing or immediately clear the report. A duplicate excluded in favor of the intended root is normal.
  2. If the report persists after a fresh crawl, inspect Google's selected canonical for the remaining examples. The retired play host could be simplified to a direct permanent redirect at its hosting configuration; that infrastructure is not owned by this GitHub Pages repository and was not changed here.

After deployment, the homepage, robots file, sitemap, and sitemap index all returned 200. The existing sitemap index was resubmitted in Search Console, which confirmed Sitemap submitted successfully and then Success, with its last read on September 6. Discovered pages initially remained zero; processing and indexing are separate from successful retrieval. No account permissions changed.

Sitemap dates and pages.dev hosts — October 26, 2026

Live checks against build 238d1aa8 (the main head): all 329 sitemap URLs returned 200 with a self-referencing canonical on https://kilnstudio.tools and no noindex; every share image returned 200 at 1200×630. The sitemap carried no <lastmod>, and kilnstudio.pages.dev served the full site with no X-Robots-Tag. Preview aliases already answered with X-Robots-Tag: noindex.

  • finalize-site now writes a <lastmod> for every sitemap URL (site/scripts/sitemap-lastmod.mjs): the newest commit date among the files the page is built from (its Astro page, the local modules and data records it imports, and for a guide its Markdown). The shared base layout and site constants are not followed, so a navigation or version-label change does not redate every page. validate-static fails a sitemap URL without exactly one W3C Datetime <lastmod>.
  • _headers sends X-Robots-Tag: noindex on kilnstudio.pages.dev and every *.kilnstudio.pages.dev alias, following Cloudflare's documented absolute-URL rule. The production host is unaffected; validate-static checks both.

Robots rules, canonicals and the noindex pages are unchanged.

Retired scene-pack releases — October 26, 2026

Search Console reported 404s for files under scene-pack releases that production had since replaced, for example /scene-packs/golden-gate/g9/THIRD-PARTY-NOTICES.txt and /scene-packs/foundry-floor/ff3-review2/licenses/ASSET-LICENSE.txt. Earlier Pages deployments show five retired releases were served: Golden Gate g9 and g9-code5, Foundry Floor ff3-review2, and Troy troy-20261005-04 and troy-20261005-06.

site/public/_redirects moves each retired directory to its current release with a 301 splat rule, for example /scene-packs/golden-gate/g9/* to /scene-packs/golden-gate/g9-code6/:splat. Every Golden Gate and Foundry Floor file keeps its name in the current release. Most Troy files do too; those that do not redirect and then return 404: from troy-20261005-04, 24 per-revision source files and 8 qualification pages (32 of 1,912), and from troy-20261005-06, 36 hashed bundle chunks (of 1,937). validate-static checks that each rule is a 301 splat move, that no built file sits under a retired directory, that each target directory has files (in a packs build; the reduced KILN_SITE_PACKS=0 build has no scene packs), and that no target is itself redirected. Asset JSON and GLB 404s and the retired play. hosts are left alone.

Raw data files — November 5, 2026

Search Console showed Google crawling raw files as documents, for example /asset-demos/**/manifest.json, /asset-demos/**/*.kiln-metadata.json and /scene-packs/**/THIRD-PARTY-NOTICES.txt. _headers now sends X-Robots-Tag: noindex on .json, .txt, .md and SHA256SUMS under /asset-demos/ and /scene-packs/, and on the same kind of file elsewhere: /assets/*.json, /scene-runtime/*.json, /sources/*.md, /fonts/*.txt, licence and lock files beside the raw data, and the root build receipts. The header does not stop a fetch: scenes and agents load these files as before. /llms.txt is deliberately left indexable so agents and AI tools can find it. robots.txt stays allow-all, because a crawler cannot see a noindex on a blocked URL. validate-static fails a build where a raw sample loses the header, where a rule reaches a built .html file, a sitemap URL, llms.txt, the media a page shows or .well-known, or where robots.txt blocks a raw file. GLBs, images and the scene-source code files are unchanged. To check production after a deploy:

curl -sI https://kilnstudio.tools/scene-packs/golden-gate/g9-code6/SHA256SUMS | grep -i x-robots   # noindex
curl -sI https://kilnstudio.tools/assets/abyssal-surveyor.poster.json | grep -i x-robots           # noindex
curl -sI https://kilnstudio.tools/llms.txt | grep -i x-robots                                      # nothing
curl -sI https://kilnstudio.tools/ | grep -i x-robots                                              # nothing
curl -s https://kilnstudio.tools/ | grep -c 'cdn-cgi/l/email-protection'                           # 0

The footer's Contact link sits inside <!--email_off-->, because Cloudflare's Email Address Obfuscation is on for this zone outside /docs/. It would otherwise rewrite the link to a /cdn-cgi/ URL that answers 404 to crawlers and to visitors without JavaScript.