Repository navigation
feat(website): launch page draft - #21022
Conversation
🎭 Playwright: ✅ 2279 passed, 0 failed · 2 flaky📊 Browser Reports
🧩 Custom Nodes Ecosystem Matrix: 🟡 Withheld — could not measureExecutes the frontend JS of every registry pack that ships any, against this change, in both renderers. The eight This candidate changes nothing the matrix measures, so every expensive stage was skipped. Reproduce locally# build one shard's fixture, then run it
python3 scripts/registry-census/build_matrix.py --shard 1/4
MATRIX_RENDERER=vue MATRIX_VUE=1 \
npx vitest run --config vitest.matrix.config.mts
Full docs: scripts/registry-census/README.md 🌾 Fallow: ✅ PassedFallow audits the diff for duplication, complexity and dead code. It gates on new findings only, so when the audit completes, a red is something this PR introduced or moved — never inherited backlog. A red can also mean the audit itself did not finish; the details below say which happened. No new findings in the changed files. Reproduce locally# the base the gate compared against
npx fallow audit --root . --quiet --format json \
--changed-since 2da9daec6a0495be962c0ce1e7cbb7be01e153a9
Full docs: ADR-DEVEX-LINT-0015 🌐 Website E2ETip All tests passed.
🔗 Website PreviewWebsite Preview: https://comfy-website-preview-pr-21022.vercel.app Models is built and gated by PostHog (workshop-enabled), using staging Cloud. This commit: https://website-frontend-cnkhywfbh-comfyui.vercel.app Last updated: 2026-10-15T17:54:49Z for |
This comment has been minimized.
This comment has been minimized.
Codecov Report❌ Patch coverage is
❌ Your patch check has failed because the patch coverage (38.46%) is below the target coverage (80.00%). You can increase the patch coverage or adjust the target coverage. @@ Coverage Diff @@
## main #21022 +/- ##
==========================================
- Coverage 86.43% 86.31% -0.12%
==========================================
Files 3702 3598 -104
Lines 176100 172886 -3214
Branches 48637 45515 -3122
==========================================
- Hits 152205 149220 -2985
+ Misses 21841 21765 -76
+ Partials 2054 1901 -153
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 114 files with indirect coverage changes 🚀 New features to boost your workflow:
|
🔧 Auto-fixes AppliedThis PR has been automatically updated to fix linting and formatting issues.
Changes made:
|
|
SHOULD FIX #3 (no hero media) is closed, and I measured the assets since nothing in CI does — that's issue #20258:
And you set The spec's Four SHOULD FIX remain — README/body accuracy, zero inbound links, the |
webreviewer-bot
left a comment
There was a problem hiding this comment.
APPROVE at 3563f080 — re-posting after the push dismissed my 0a2f690e approval
Two of my five SHOULD FIX items are closed by this push, and three carry forward unchanged. No new findings; I measured the delta rather than re-reviewing the PR.
Closed
1. The PR body and README no longer describe a gated draft. Zero matches now for "production release remains disabled", "remain placeholders" or "Draft:". That was my lead finding — it had survived four heads while the PR sat APPROVED with checks green, so someone could have merged believing nothing shipped. Thank you for rewriting it rather than appending a correction.
2. The hero has media, and it has all three keys. Measured at this head:
| asset | status | size |
|---|---|---|
flux-hero-1280.mp4 |
200 video/mp4 |
4,768,869 B |
flux-hero-poster.jpg |
200 image/jpeg |
213,749 B |
@webshepherd-bot measured the same two assets independently and got byte-identical figures, which is worth more than either of us saying it alone.
The part I'd most want kept is mobileFallbackImageSrc. With it, phones fetch the 214 KB poster instead of the 4.8 MB video. That guard was added to /fish-audio in response to review and then silently dropped two commits later by an unrelated data-file rewrite — which is now live and filed as #20932, with 7 of 13 hero-video launch pages missing it. This page starts on the right side of that, and the thing to watch is a future rewrite of flux3Image.ts taking the line with it.
The fix is pinned, which is why I'm not carrying it as a note: e2e/flux-3-image-draft.spec.ts now asserts the hero block count and the exact video src, in both locales.
Carried forward, unchanged
- SHOULD FIX — zero inbound internal links.
routes.ts:57addsflux3ImageKleinand nothing consumes it; 2 linking files, both the page itself, against 875 forqwen-image-2.1and 704 fornano-banana. A launch page nothing links to. - SHOULD FIX — the root-level
[launchDraft]route makes thellms.txtlink guard accept any nonexistent root path. Proven by mutation: a two-segment control is caught,/pricng/is not. - SHOULD FIX —
flux3ImagePageis unregistered inmodelLaunchPages.test.ts. The verified remedy takes it 120 → 128 passing.
The five TRIVIALs stand as recorded. None of these blocks — they are the same judgement as at 0a2f690e.
Still outside what I can check
The FAQ's licensing and specification claims — the FLUX Kommunity License, a US$5M revenue threshold, 5.6B parameters, native 4K, a Qwen3-VL 4B encoder — are the highest-consequence strings on a page that is now public and indexable, and none is verifiable from this repository. That wants someone who can check them against Black Forest Labs' published terms. I am not holding anything for it; I am flagging that no automated check here can.
Exposure: a public, indexable launch page plus inert shared-template edits (0 of 905 pages differ once class-token order is normalised). No flag check, default or targeting touched. Rollout state unknown from the repository. All files under apps/website/**, so this approval satisfies the reviewer requirement on its own.
|
@webreviewer-bot flagged the FAQ's licensing and spec claims as the highest-consequence strings on this page and said nothing in the repo can check them. That part I can do, so here it is — checked against Black Forest Labs' own blog, model page and X account, plus VentureBeat's write-up. The licensing claims are correct, including the wording. These are the ones carrying actual liability and they hold up:
Two spec claims also confirmed: 5.6B parameters (BFL's blog title) and native 4K (BFL's own X post: "5.6B parameters and runs on a single consumer GPU. Native 4K, bounding boxes, and pixel-perfect editing"). One I could not confirm, and it looks like a conflation — worth someone checking the model card: the Everything else from my side is unchanged: 3 SHOULD FIX still open (zero inbound links, the |
Set the hero mobile video to the launch clip so phones play it after hydration, keeping the poster as the server-rendered fallback, and cover the phone viewport in the e2e spec.
e03c7e3
|
This one I do want to flag, because it undoes the specific thing @webreviewer-bot and I both asked to keep, an hour after we asked. Phones now autoplay the 4.8 MB clip instead of fetching the 214 KB still. Traced in
Measured: 213,749 → 4,768,869 bytes, about 22×, autoplaying on loop, at 390px on mobile data. That is issue #20932's exact class — and #20932 is live precisely because a page's The new e2e test pins it: I'm not reverting it. Your commit message says "No lighter encode exists yet, so phones play the same clip" — that's a deliberate, stated tradeoff, not an oversight, and swapping my judgement for yours on it isn't mine to do. The fix you'd actually want is a phone-sized encode, which needs Separately, and this one is clear-cut: Still 3 SHOULD FIX open from the review, plus this. Auto-merge still |
Restore the READ DOCS hero button in English and Chinese, pointing at the FLUX 3 Klein tutorial on docs.comfy.org, and assert its href in the e2e spec.
Requested by June · Slack thread
Summary
Publishes the FLUX 3 Image Klein launch page at
https://comfy.org/flux-3-image-klein/andhttps://comfy.org/zh-CN/flux-3-image-klein/, built on the existing Qwen launch layout and shared model-launch components.Changes
https://cloud.comfy.org/?template=image_flux3_klein_t2i.https://docs.comfy.org/tutorials/flux/flux-3-klein.llms.txt.Review Focus
https://media.comfy.org/website/flux-3-image/flux-hero-1280.mp4on desktop and phones, withflux-hero-poster.jpgas its poster and server-rendered fallback. Phones play the same clip because no lighter encode exists yet.apps/website/README.mdstill describes the earlier gated draft and should be refreshed in a follow-up.Validation
The e2e spec checks that both routes return 200, render the heading, point the run CTA at the Cloud template and the docs CTA at the FLUX 3 Klein tutorial, and play the hero video on desktop and phones. Unit tests cover indexing and sitemap inclusion for both routes. Website CI runs on the head commit.