Repository navigation
Conversation
|
👋 Hi! This PR is currently in draft status. Workflow runs won't be auto-approved until it's marked as ready for review. When you're ready, click "Ready for review" and the workflows will be approved on the next cycle automatically. |
GPT 5.6 Review (fork) — ✅ no blocking findingsReviewed Review detailsFINDING -- website/src/components/library/LibraryTable.tsx:181 -- |
Design Review (Fable 5, fork) — 🔴 BLOCK (blocking)Design-level review of Design-Verdict: BLOCK Cannot evaluate: missing screenshot or recording of the offline artifacts library — dimmed controls, menu reason row, sticky table footer, held-name note. BlockersMissing rendered evidence for a reshaped visible surface. WatchThe description describes a different revision: it names [DESIGN-REVIEWED] b445445 |
UX Review (Fable 5, fork) — 🔴 BLOCKUX-level review of UX-Verdict: BLOCK cannot evaluate: missing screenshots of every offline state this PR adds — both attachment lists are empty ("Screenshots / video: None attached"). Evidence gaps
The temp-screenshots policy question the description cites doesn't apply here: attaching the images to the PR description (github user-attachments) is the path this gate reads, and commits nothing. [UX-REVIEWED] b445445 |
First Principles Review (Fable 5, fork) — 🟡 CONCERNSPremise-level review of First-Principles-Verdict: CONCERNS A declared two-control fix ships five undeclared surfaces — move-submenu gate, swatch gate, doc star, a draft-holding editor, sticky table rows — and misdescribes what shipped. Not justified as shipped
What this change shipsInventory (10 items) — 4 justifiedIntent: stop offline artifact-library writes from silently dropping by dimming and refusing them, per the existing
[FIRST-PRINCIPLES-REVIEWED] b445445 |
Opus 5 Review (fork) — ✅ no blocking findingsReviewed Review detailsNo blocking defect — two advisory gaps in the new offline gating. FINDING — website/src/pages/ArtifactsPage.tsx:1299 — drag-to-file is the table/tree's only artifact-move path and stays ungated while the menu's Move is now refused ( FINDING — website/src/components/library/LibraryTable.tsx:440 — the offline branch adds [OPUS-REVIEWED] b445445 |
7af35f5 to
2ebdf95
Compare
2ebdf95 to
56f0db6
Compare
56f0db6 to
3997323
Compare
3997323 to
53b569b
Compare
5b4b6ca to
6703736
Compare
6703736 to
6f960f6
Compare
6f960f6 to
b41e7fc
Compare
b41e7fc to
f2a6122
Compare
Star, Delete and the folder menu wrote through the gateway with no connected check, so an offline click looked like it worked. Split from a larger sweep.
|
🤖 Kiro Crew [operator: NicholasRBowers#a942f9ca]: This PR has been idle since 2026-09-24 with two blocking lanes and a conflicting base. I've assessed the blockers and they appear resolvable — I'll push fixes directly to this branch as a co-author. Assessment: Nothing blocking here is a code defect. Design and UX both block for the same reason — no rendered evidence for a reshaped visible surface — so the work is capturing the five listed offline states (row star and Delete, the folder menu with its reason row, the sticky row mid-scroll, Worth flagging: no AI lane has judged the current head at all — GPT, Opus, Internal Content Scan, Security Scope, Design, First Principles and UX all failed closed in the 2026-09-24 rate-limit storm, so the verdicts above are from If you'd prefer I don't touch this PR, add the |
|
PR triage (automated, against Not on main yet. The overlap is with the sibling split PRs, open PRs #9485 and #10911 (both checked as PRs, not issues). The three PRs build incompatible versions of the same shared primitives:
Two consequences:
Suggested path: pick one owner for the shared No breaking change and no P0 found. The Opus advisory about the ungated drag-to-file path in |
|
Thank you @rnoack1 for this fix. This branch was about 3,700 commits behind main, and #20749 (your #9485) has since landed the shared offline primitive, so I carried the change forward in a same-repo PR: #20989. Your commit keeps your authorship. The port uses the API main now has ( Closing this one in favour of #20989; the branch is left in place. |
Problem / Motivation
In the artifacts library, the row's star toggle, the row's Delete, and the folder menu's
rename / delete items all wrote through the gateway with no
connectedcheck. With thegateway down the click was accepted, the optimistic state flipped, and the write was
dropped — so an offline delete looked like it worked.
Every other write surface in the sidebar already refuses these clicks visibly through
offlineProps. This table did not, and the two row controls were the last ungated pairin a file whose folder menu had already been gated.
Why it matters
A silent drop is worse than a refusal. A user who sees a row disappear believes the
artifact is gone; on reconnect it returns, and the only explanation available to them is
that the app is unreliable. The failure mode is invisible at exactly the moment the user
has least information — no error, no notice, no retry affordance.
What changed (motivation → approach → change)
ArtifactRowand the folder menu now readuseConnected()and gate at the sink, which isthe rule the rest of the codebase already follows:
LibraryTable.tsx— the star toggle and Delete carry theopacity-40dim, theofflinePropstooltip naming the verb, and a guard at the handler so a click thatreaches it anyway cannot write.
OfflineMenuReasonrenders last inside the folder menu, because a nativelydisabledrow carries
pointer-events-noneand so opens no tooltip — an explanation gated onactivation cannot reach it.
utils.offlineverbs (star_artifacts,delete_artifacts,delete_folders,rename_folders) plus the menu'soffline_reason, across all 12 catalogues.OfflineMenuReason.tsxis a new 17-line component. It is added identically by the siblingPRs listed below, so whichever lands first carries it and the rest merge cleanly.
This PR is one unit of a split. A single larger change gated ~18 controls at once and
came to 4224 added lines, which is not reviewable. It is being cut into independent units;
this is the artifacts-table one, and it stands alone — it depends on no sibling.
Tests
LibraryTable.folderMenuOffline.test.tsx,ArtifactsPageCoverage.test.tsx,ArtifactsPage.test.tsx— 118 passed.asserts the write still reaches the server when the gateway is up, so the tests
discriminate rather than merely passing.
tsc -p tsconfig.app.jsonclean. i18n suites 683 passed;check-i18n-keys.mjsandcheck-i18n-strings.mjsboth OK.Manual verification
Toggled the gateway off with the artifacts library open: the star and Delete controls dim,
their tooltips name the verb, the folder menu shows the reason row, and no request is
issued. On reconnect the controls return to full weight and write normally.
Screenshots / video
None attached. Screenshot evidence for this change is under an unresolved repo-policy
question — the convention this repo's evidence gate rewards is committing frames under
temp-screenshots/, which.gitignoredesignates as never-committed. Rather than pick aside inside a split PR, no frames are committed here and none are attached. Happy to
supply them in whichever form a maintainer prefers.
Related Issues
None.
Pattern harvest
useConnected()at the component that owns the write, rather than threading aconnectedprop down: this file already did that at the folder level, so the row followed it instead
of introducing a second convention.
Checklist
Contribution License Agreement
I confirm this contribution is made under the terms of the project's contribution licence.