Repository navigation
Conversation
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: |
UX Review (Opus 5.5) — ✅ PASSUX-level review of UX-Verdict: PASS Nothing blocks. The blind reader read the new paused chip correctly: "the pinned chats now also go through the other filters". It rated that reading only "a guess" and called the wording "hard". Notes
[UX-REVIEWED] 3901207 Description read[DESCRIPTION-READ] 4f4e01c58cb74b1447aabcb28eaff55117ec7473a0b1f8d03a803f3bbaf01373 That sha256 names everything this verdict read: the PR description, plus the 7 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 |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: the paused Pinned chip's special meaning is scoped to the status dimension, labelled on the chip itself, recorded in the spec, and backed by four screenshots. Notes
[DESIGN-REVIEWED] 3901207 Description read[DESCRIPTION-READ] 6ea5e408b987c399157fa5260763ca1544f35c28f8948f629dea25b99bbac2d4 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 |
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: |
First Principles Review (Opus 5.5) — ✅ PASSPremise-level review of First-Principles-Verdict: PASS Nothing blocks. The fix is at mechanism level ( Notes
What this change shipsInventory (7 items) — 4 justifiedIntent: let someone who pins many sessions see only the pinned ones the other status chips match, by pausing Pinned. FIX.
[FIRST-PRINCIPLES-REVIEWED] 3901207 Description read[DESCRIPTION-READ] b7d7b1bac45f32824fe8c55859810fbc0d72c6e0fe2392d4ecbcd9f54bd6be70 That sha256 names the PR description this verdict read. Recompute it with |
a4b5b1d to
134ab68
Compare
134ab68 to
cfcd11c
Compare
cfcd11c to
9af21aa
Compare
Code change approver recommendation: A person should decideWhy: the decider estimates a 12% chance that this change introduces a defect, within the 12% limit for approving on its own, but it is held at tier 2: design-doc-gate: D. Pausing the Pinned chip while another status chip is on now stops pinned rows from passing the status filter automatically. The tag filter and folder pins still let them through, as the PR's non-goals require. The filter cache is keyed on the active and paused chip sets, so it recomputes when they change. The caption and the chip and menu wording all use the same check as the filtering. Reveal now clears the status chips only when they really hide the row. Lifecycle cases (pause all, resume all, turning a chip on while the others are paused) fall back to the old behaviour. New tests cover each change and would fail without it, and CI is green. I found no regression or contradiction with the spec, only notes on documentation and wording. Tier: 2, held for a person: design-doc-gate: D. An automated recommendation, not a review: a maintainer still decides. It approves on its own only when the estimated chance of a defect is at most 12% and the change has at most 400 changed lines. Decider v3c (217fe607) · commit |
|
This PR edits docs/feature-map/README.md, so a maintainer needs to approve it. This reviewer does not approve changes to that file. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/). Review result on this head: no blocking findings.
Done well: one flag in the one place that declares the filter dimensions, with a mutation-sensitive test for each over-broad version of the fix. Optional:
|
9af21aa to
c6668e8
Compare
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Thanks for the focused fix. One evidence gap blocks this head.
website/src/pages/chat-sidebar/SearchDock.tsx:467: when Pinned is paused, the filter-menu row now ends in· filtered tooinstead of· paused. None of the three attached screenshots shows the menu open, so the UX review could not evaluate that row ("cannot evaluate: missing the filter-menu row for a paused Pinned chip"). A changed control with no rendered evidence counts as a block, even though the lane line reads PASS. Fix: runwebsite/scripts/capture-status-chip-pause.mjson this head and attach the frame with the menu open, In progress on and Pinned paused (shot-05). Give it a header naming the head commit, then re-run the UX lane.
The rest of the review found no blocking defect. One flag in the status dimension drives filtersRow, hides and exemptsPinned together, and the new rows.test.ts block goes red when any one production hunk is reverted. This PR also edits docs/feature-map/README.md, so a maintainer gives the final approval once the evidence is in.
Optional:
- With a tag filter on, In progress on and Pinned paused, an idle pin is hidden, but
pinnedExemptNarrowing(rows.ts:972) stays true because the tag dimension still exempts pins. The "Pinned · always shown" caption then still shows. The tooltip's "follow the other filters" also overclaims, because pins still skip the tag filter. "the other status chips" would be accurate. - The paused Pinned count (
rows.ts:799) counts pins that the status chips keep, but a model, creator or search filter can still drop them. So "the pinned sessions still listed" in the body and the feature map is not exact. That rule also conflicts with the "counts read the predicates" comment atrows.ts:632-635. - The status reveal
hidesnow reads the chips' own predicate for every row, not only pinned rows. This is a sound change, but "What changed" andhistory.mddo not mention it. Users who already have Pinned paused will also see fewer rows after upgrading. Stating that under Backwards compatibility would help readers.
ec78f2c to
c1e27a9
Compare
|
This PR edits docs/feature-map/README.md, so a maintainer needs to approve it. This reviewer does not approve changes to that file. A maintainer will post the decision here. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/). Earlier request on ec78f2c: fixed. The body now has the shot with the filter menu open (In progress on, Pinned paused). It shows Review result on this head: no blocking findings.
Done well: the reveal loop test (rows.test.ts:552-571) turns the status Readiness still shows the UX and First Principles BLOCKs with no disposition. Each one needs a fix or a disposition before a maintainer can merge. Optional:
|
Withdrawn: every item in this review is fixed or answered on the current head (see the follow-up comment). Other holds on this PR are listed there.
c1e27a9 to
fc5bf4b
Compare
fc5bf4b to
1916d73
Compare
1916d73 to
3901207
Compare
|
This changes behaviour users already rely on, so it needs a design sign-off from a maintainer before it can merge. A maintainer will post the decision here. This PR also edits docs/feature-map/README.md. This reviewer does not approve changes to that file. To propose a design, follow the request-for-change process in CONTRIBUTING.md (docs/request-for-change/). Earlier request on ec78f2c: fixed. The body has the frame with the filter menu open (In progress on, Pinned paused), and the frame has a header that names the build. Review result on this head: no blocking findings.
The design question for the maintainer: with Pinned off, pins stay exempt. With Pinned paused, pins now follow the other status chips. For every other chip, paused narrows nothing, like off. Users whose Pinned chip is already paused will see fewer rows after they upgrade. Done well: one shared Optional:
|
Problem / Motivation
Goal: Pausing the Pinned status chip makes the other status chips narrow pinned sessions too, so the pause has a visible effect.
Click the Pinned chip to pause it while In progress (or Unread) is on. The chip reads
Pinned (17) · paused, but every pinned session stays listed and the "Pinned · always shown" caption stays. The pause looks broken.Why it matters
Anyone who pins many sessions and filters by Unread or In progress cannot see just the pinned sessions that match. The Pinned chip offers a pause that does nothing, which reads as a bug.
Not a goal
What changed (motivation → approach → change)
Pinned rows skip the status chips, whatever the Pinned chip says. So once another status chip was on, the Pinned chip never decided anything, active or paused.
Now a paused Pinned chip turns that skip off for the status chips. A pinned session stays only if an unpaused chip matches it, such as In progress. The "Pinned · always shown" caption is not shown while that happens, even when a tag filter also narrows.
The paused Pinned chip says what it does on its face: it reads
Pinned (N) · paused, filtered too, and so does its row in the filter menu. It keeps the word "paused" that every other paused chip shows, and adds what the pause changes: the pinned sessions are filtered too. This wording and the tooltip appear only while another status chip is on. With no other chip narrowing, pins still pass every filter, so a paused Pinned chip reads plain "paused" and "Resume Pinned filter", like any other chip. Its countNis the pinned sessions still listed. Its tooltip reads "Pinned sessions follow the other status chips. Click to resume." The two new strings,pages.chatSidebar.pinned_paused_suffixandpages.chatSidebar.paused_pinned_chip_titleinwebsite/src/i18n/locales/en.manual.json, are translated in every locale file. Tag filters and folder pins keep their exemption.Revealing a session from the chat header clears only the filters that actually hide it. The status chips still ask list membership first, as before. They now also ask their own match, so a reveal of a pinned row that only the search dropped keeps the chips. Without that, pausing Pinned made such a reveal clear both chips.
The spec, the feature map and
website/scripts/capture-status-chip-pause.mjsmatch this.Backwards compatibility
Compatible: no key, field or stored value changes, and the stored chip values
'0','1','2'read as before.Behaviour change: someone whose Pinned chip was already paused (stored
'2') beside another status chip sees fewer pinned rows after upgrading, because the pause now narrows pins. Resuming Pinned shows them again.Removes nothing: no reader of a key, field or stored filter value is removed.
Tests
rows.test.tsgainsa paused Pinned chip lets the other status chips narrow pinned rows:The narrowing and count cases fail on main's
rows.ts; the others guard what must not change.ChatSidebar.filterSuspend.test.tsxchecks that only the paused Pinned chip reads "paused, filtered too" and carries the new tooltip, that a paused Unread chip still reads "paused" with "Resume Unread filter", and that resuming Pinned restores "Pause Pinned filter". It fails without the chip change.Proof: not provable -- frontend test (prove.py runs Python only)
Manual verification
Captured the sidebar with stubbed sessions (two pinned, one running each way) on main
11e9453ae3and on branch head1916d737b0, with In progress on. Each shot carries a header naming its build and state.Pinned (2) · pausedPinned (1) · paused, filtered tooOn the branch the paused chip's
titlereads "Pinned sessions follow the other status chips. Click to resume." (read from the DOM; a headless browser does not paint native tooltips).Screenshots / video
Main, Pinned paused: the idle pin is still listed and the chip says only "paused".
Branch, Pinned on (unchanged):
Branch, Pinned paused: only the running pin is left, the caption is gone, and the chip reads
Pinned (1) · paused, filtered too.Branch, Pinned paused with the filter menu open: the Pinned row reads
Pinned (1) · paused, filtered toobeside the pause icon.Related Issues
Closes #38883
Sibling sweep
Swept: git grep -nE "pinnedBypass|matches.pinned(" -- website/src
Sibling: website/src/pages/chat-sidebar/rows.ts:635 -- not affected -- the tag filter's pin exemption (its filter and reveal sides) is a separate answer and stays on
Pattern harvest
Not generalizable: the defect is one exemption that ignored the state of the chip whose predicate it shares; no other chip carries an exemption.
Checklist
feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...)