Repository navigation
fix(design-critique): say the composer below is the retry on a failed ask turn - #21891
Conversation
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Check the title on a failed turn that is not the last one in a thread. "Ask again below" then sits above later answers, not directly over the composer. [DESIGN-REVIEWED] b8dd13a Description read[DESCRIPTION-READ] cab108fea464afc23dc378e11e05e7d92bed5b2def0845c2afe3c72b367b8d80 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing to check. What this change shipsInventory (3 items) — 3 justifiedIntent: when an ask turn fails, the user should know they can retry by typing in the Ask composer just below it. This is a FIX.
[FIRST-PRINCIPLES-REVIEWED] b8dd13a Description read[DESCRIPTION-READ] b0dc58548d0d5c8ca26f6bcba2175f582d3f39af297289fb2fa9a7bd86f2573e That sha256 names the PR description this verdict read. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
UX Review (Opus 5.5) — 🟡 CONCERNSUX-level review of UX-Verdict: CONCERNS The new title says "ask again below", but the box below says "Ask a follow-up…", so users can't tell they must retype the failed question. Watch
Suggestions
[UX-REVIEWED] b8dd13a Description read[DESCRIPTION-READ] 01f70a123e9b72b713d7963ccd52f405d622c4d47ed40571ee10b00a9b9943a2 That sha256 names everything this verdict read: the PR description, plus the 5 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
|
|
|
|
da60a3e to
b8dd13a
Compare
|
|
|
|
|
|
Review-ready at b8dd13a: PR Readiness passed (103 checks, 0 failing), green is fresh against main, one commit, and every review item on this head has a disposition. One wording question is left for the maintainer in the needs-a-decision reply (keep "Answer failed — ask again below" or switch to "Answer failed — type your question again below"). The placeholder and refill follow-ups are tracked in #21894. Not merged, approved or set to auto-merge; that is the maintainer's call. |
mbriones98
left a comment
There was a problem hiding this comment.
Checks green and readiness passed at b8dd13a; no behavior change.
Problem / Motivation
Goal: When a Design Critique ask turn fails, the error tells the user to ask again in the composer just below it.
The failed turn showed only the raw failure text, such as "Agent session ended before it replied". Nothing said the Ask box under it is where you retry. A blind reader in the UX Review lane on #13461 said: "if the thing that answers is gone, I don't know if typing here would work."
Why it matters
A user whose question failed does not know whether typing again will work, so they may give up on the annotation or reload the page and lose the thread.
Not a goal
messagestays the raw string, because it is the error journal's lookup key.What changed (motivation → approach → change)
The failed turn's
ErrorNoticeinwebsite/src/apps/design-critique/AskLayer.tsxnow has a bold title: "Answer failed — ask again below". The raw failure text sits on its own line under it (messagePlacement="below"), so the guidance leads and the server string reads as the detail.The title is a new
apps.designCritique.askLayer.answer_failed_ask_again_belowkey, written by hand inen.manual.jsonand translated in every locale catalogue underwebsite/src/i18n/locales/. Theen-XApseudolocale was regenerated withgen-pseudolocale.mjs.Backwards compatibility
Compatible: the notice gains a title and a layout prop; the
messageand its journal key are unchanged.Removes nothing: the failure text still renders in full, only under the new title.
Tests
website/src/test/DesignCritiqueAskLayerCov80.test.tsx: a new case renders a failed turn and asserts the alert carries both the guidance title and the unchanged failure text. It fails on the parent commit (title absent) and passes here.--sequence.shufflerun.src/i18nsuite (50 files, 1022 tests) withI18N_BASE_REFset, covering catalogue parity, pseudolocale and per-locale style checks;i18n:check,i18n:render,gen-pseudolocale.mjs --check,tsc -p tsconfig.app.jsonand eslint on the touched files all exit 0.Manual verification
Rendered
AskLayerwith a failed turn in Chromium (Vite dev server on loopback, English, dark theme) before and after the change; screenshots below.Screenshots / video
Before: only the failure text.
After: the guidance title, with the failure text under it.
Related Issues
Closes #13487
Pattern harvest
Not generalizable: a missing piece of guidance copy on one notice, not a code defect class.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)