Repository navigation
test(website): check the long description, FAQ and models in workflow server HTML - #21550
MaanilVerma wants to merge 7 commits into
Conversation
… server HTML The catalog detail audit now reports a workflow page whose server HTML does not contain its long description, each FAQ question and answer, or its model list. The SSR test runs the audit on two CCC workflow records, one with a Run form and one without.
📦 Bundle Size
🧩 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
Findings and check detailsThe matrix runs on same-repository version-bump PRs with relevant changes, main-branch pushes, and manual dispatches. This candidate does not qualify, 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. Findings and check detailsNo new findings in the changed files. Reproduce locally# the base the gate compared against
npx fallow audit --root . --quiet --format json \
--changed-since cfcff3a504815861c835bdb52cee50ca1d0f9d6a
Full docs: ADR-DEVEX-LINT-0015 🏛️ Architecture: ✅ PassedFindings and check detailsWarnings (132)
Reproduce locallypnpm architecture:checkFull docs: Frontend domain catalog 🌐 Website E2ETip All tests passed. Test results
🔗 Website PreviewWebsite Preview: https://comfy-website-preview-pr-21550.vercel.app Models is built and gated by PostHog (workshop-enabled), using staging Cloud. Deployment detailsThis commit: https://website-frontend-6284ts948-comfyui.vercel.app Last updated: 2026-10-21T16:53:19Z for |
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configuration
You can disable this status message by setting the Use the checkbox below for a quick retry:
Comment |
This comment has been minimized.
This comment has been minimized.
benjcooley
left a comment
There was a problem hiding this comment.
Review: no blockers. A few audit-accuracy gaps.
The approach is sound: the paragraph split, the FAQ flattening and the model-list selection all match WorkflowPage.vue, FAQSplit01.vue and WorkflowFacts.vue. Blank FAQs are dropped upstream in faqDetails(). Vitest config auto-restores the env and fetch stubs, so nothing leaks.
The main issue: the audit treats authored record text as HTML. Text that looks like a tag is dropped from the expected value, so a long description that contains ComfyUI prompt syntax falsely fails (inline). The audit also matches substrings across the whole page, so some checks are weaker than they look.
Note: auditCatalogDetailBody has no caller outside the vitest SSR tests. check-models-html.ts does not use it, so this guards fixtures and not the built dist. That is fine for this PR, but don't read it as a production audit.
| Severity | Count |
|---|---|
| BLOCKER | 0 |
| SHOULD FIX | 3 |
| TRIVIAL | 2 |
Tested: scripts/models-html-audit.test.ts passes 28/28 on a copy of the branch. Mutation testing was done on that copy, not on the branch. The SSR test was not run locally because of a missing dependency.
benjcooley
left a comment
There was a problem hiding this comment.
Re-review: no fixes yet
The only new commit is ac5ec67f23, a merge from work/ccc-mar-integration-20261014. The three PR files are byte-identical to cb22522837, so none of the five round-1 threads are addressed. All five are still open:
| Thread | Severity | Status |
|---|---|---|
has() passes authored text through textOf, which strips <…> (false "missing" on <lora:…> prompt syntax) |
SHOULD FIX | open |
whole-page substring matching (SD/Flux pass from the h1 or summary) |
SHOULD FIX | open |
| no test drops the FAQ question (the mutation survives) | SHOULD FIX | open |
faqAnswerText duplicates faqAnswerPlainText |
TRIVIAL | open |
MODEL records with catalogDetails.models false-fail (latent) |
TRIVIAL | open |
The merge brought in a FAQSplit01.vue change (optional faq.link), which I checked. The audit's FAQ input comes from faqDetails() and never sets link, so the audit is unaffected.
CI note: website-unit failed on cb22522837 only because of src/styles/catalog-styles.test.ts, which is unrelated. This PR's audit and SSR tests passed at the default timeout, which answers your open QA item about the 5 s timeout.
| Severity | Open |
|---|---|
| BLOCKER | 0 |
| SHOULD FIX | 3 |
| TRIVIAL | 2 |
The first SHOULD FIX has a one-line committable suggestion on its thread. Please push fixes or reply on each thread before I approve.
Codecov Report✅ All modified and coverable lines are covered by tests. @@ Coverage Diff @@
## work/ccc-mar-integration-20261014 #21550 +/- ##
=====================================================================
- Coverage 86.45% 86.34% -0.11%
=====================================================================
Files 4030 3924 -106
Lines 184642 181388 -3254
Branches 50825 49716 -1109
=====================================================================
- Hits 159627 156622 -3005
+ Misses 22316 22216 -100
+ Partials 2699 2550 -149
Flags with carried forward coverage won't be shown. Click here to find out more.
... and 122 files with indirect coverage changes 🚀 New features to boost your workflow:
|
benjcooley
left a comment
There was a problem hiding this comment.
Re-review (round 3): approve
I checked the five round-1 threads against 9e048b4371, df8ff9cda7 and 7d0fd565d9. All five are fixed, and I resolved them.
- Authored text is matched as-is. The
<lora:…>,<b>,&and</>cases pass, and the summary check got the same fix. - Section scoping. All five markers exist in the real components (
WorkflowPage.vue:183,210,WorkflowFacts.vue:97,LocalModelPage.vue:55,75) and are the first occurrence of their string. A renamed marker gives an empty section, so the check fails loudly. - The
workflowIdgate is correct.workflowIdis required on workflow records (models-catalogue-data.ts:51). CCC records set it (catalog-presentation.ts:488), andCatalogDetailPage.vue:85uses the same field to pickWorkflowPage. So the model check still runs in the real SSR tests, and all 30 seed workflows havetemplate.models. - The FAQ-question test exists. The
faqAnswerPlainTextreuse is in.
Verification: the unit audit tests pass 34/34 locally. Mutation probes on the new code were all caught: depth tracking, whole-page scope, the workflowId gate, the FAQ ||, the summary <> handling and the model scope. CI is green on 7d0fd565d9, with 55 checks passing.
Two inline TRIVIAL notes, neither blocking. Two optional test gaps:
- No case puts About or FAQ text after its section's close tag. The facts case covers the shared
elementWithclose logic. - No SSR case renders a local MODEL record with
catalogDetails, somodel-aboutandmodel-faqare checked only against fixtures.
| Severity | Count |
|---|---|
| BLOCKER | 0 |
| SHOULD FIX | 0 |
| TRIVIAL | 2 inline + 2 optional |
benjcooley
left a comment
There was a problem hiding this comment.
Re-review (round 4): approve
7310c18316 fixes both TRIVIAL notes from round 3, and I resolved both threads.
- The model check now reads
data-testid="workflow-runs-on". That marker exists once, inWorkflowModelLinks.vue:16, whichWorkflowFactsrenders once. The new author rowFlux Kontext Fanmakes the runs-on case a real test: with the oldworkflow-factsmarker, 2 cases fail. elementWithreturns''when the section is never closed. I traced the new unclosed-FAQ case: the depth goes 1 → 2 → 1 through the runs-on<section>and never reaches 0, so it hits the new branch. Withreturn restrestored, it fails.
Verification: the unit audit tests pass 36/36 locally. Every mutation of the two changed lines is caught. website-unit and typecheck were still pending when I submitted.
Optional nit: the case the model in its facts block now really means outside its runs-on list.
| Severity | Count |
|---|---|
| BLOCKER | 0 |
| SHOULD FIX | 0 |
| TRIVIAL | 1 (optional) |
|
Moved to Comfy-Org/comfy-website#25 (the website now lives there). This PR stays open for its review history. |
|
Closing: the website moved; this change lives in Comfy-Org/comfy-website#25. |
Summary
The catalog detail audit now proves that the server HTML of a CCC workflow page contains its long description, its FAQ and its model list, each in its own section. Before this PR, the SSR test used only seed records, and seed records have no long description and no FAQ. Thus nothing proved that this content reaches search engines and AI crawlers.
The page body was already in the server HTML, because Astro server-renders the
client:loadisland ofCatalogDetailPage. On a runnable workflow, the model list sits in a hidden tab panel.withoutHiddenMarkupcounts hidden-tab content as present. Visitors see no change. The only runtime change is adata-testid="model-about"attribute on the About section ofLocalModelPage.vue.Changes
auditCatalogDetailBodyinscripts/models-html-audit.tsnow reports a page whose server HTML does not contain each paragraph ofcatalogDetails.longDescription, each FAQ question and answer, or each model name.aria-labelledby="workflow-about-heading"ordata-testid="model-about"), the FAQ in#workflow-faqor#model-faq, and the models indata-testid="workflow-runs-on".elementWithfinds the element that holds a section marker by counting nested tags of the same name. It returns no text when the close tag is missing, so a section that does not close gives a report.workflowId. A workflow withoutcatalogDetailstakes its model list fromworkflow.template.models, so the seed records get the model check too.faqAnswerPlainTextfrom@/utils/faqAnswer, so an inline link inside an answer does not cause a false report.<lora:detail:0.8>or1 < 2 & 3 > 2matches its escaped form in the HTML. The summary check now uses the same rule.bodyErrorscalls one function for each section:descriptionErrors,faqErrorsandmodelErrors.LocalModelPage.vueaddsdata-testid="model-about"to its About section, so the audit can find it.catalog-detail-ssr.test.tsnow server-renders three CCC workflow records and runs the audit on them: one with a Run form, one without, and one with only a long description.QA Focus
pnpm exec vitest run scripts/models-html-audit.test.tsinapps/website. Expected: 36 tests pass. Author: tested at7310c18316, 36 pass. Before the first audit change, 5 of the first 28 tests did not pass (red).pnpm exec vitest run scripts/models-html-audit.test.ts src/routes/models/catalog-detail-ssr.test.tsinapps/website. Expected: all tests pass. Author: tested atdf8ff9cda7, 227 pass (with 34 audit tests at that commit). Not run together again after7310c18316added two audit tests.faqErrors, remove!inFaq(question) ||and run the audit tests. Expected: the caserejects a workflow body without an FAQ questiondoes not pass. Author: tested,1 failed | 33 passed.textWithinback to whole-page text and run the audit tests. Expected: the out-of-section cases do not pass. Author: tested,4 failed | 30 passed.workflow-factsand run the audit tests. Expected: the casethe model in its runs-on listdoes not pass. Author: tested,2 failed | 34 passed.elementWithreturnrestwhen the close tag is missing and run the audit tests. Expected: the casea closed FAQ sectiondoes not pass. Author: tested,1 failed | 35 passed.workflowIdgate inworkflowModelNamesand run the audit tests. Expected: the casechecks the body of a model record without a model listdoes not pass. Author: tested,1 failed | 33 passed.textOfcall back on the record side and run the audit tests. Expected: the casekeeps authored angle brackets and ampersands as textdoes not pass. Author: tested,1 failed | 33 passed. The red commit is9e048b4371.WorkflowPage.vuenot render (v-if="false") and run the SSR test again. Expected: both CCC workflow cases that have an FAQ report the missing FAQ. Author: tested in the first review round, both cases did not pass with "server HTML is missing the FAQ". I restored the file after the run.WorkflowPage.vuenot render (v-if="false"on the about section) and run the CCC workflow cases again. Expected: all three cases report the missing long description. Author: tested in the first review round, 3 of 3 did not pass with "server HTML is missing the long description". I restored the file after the run.curlon a CMS workflow page of the preview with JavaScript off. Expected: the h1, the summary, the long description, the FAQ and the model names are in the HTML. Author: not tested, because the preview reads seed data, and seed records have no long description and no FAQ.catalog-detail-ssr.test.tsat the default 5 s timeout. Expected: all tests pass in CI. Author: on this busy machine, the first test timed out on the base branch and on this branch, so the CI run decides it.Risky areas
scripts/models-html-audit.tsis also used byscripts/check-models-html.tsandscripts/models-gallery-audit.ts. A model record withcatalogDetailsnow gets the long description and FAQ checks indata-testid="model-about"and#model-faq. A model record gets no model list check, because it has noworkflowId.aria-labelledby="workflow-about-heading",data-testid="model-about",#workflow-faq,#model-faqanddata-testid="workflow-runs-on". If a later change renames one of them, the audit reports missing content for every CCC page.display:none. The audit accepts this, because the text is in the HTML. Google can give lower weight to text in a hidden tab. A visible model list is a UI change and is not in this PR.PR dependency map
main(base)main)Merge #21204 before this PR. #21551 merges after it.