Skip to content

feat(chat): reserve the chat pane in every view and cue clipped board lanes - #17136

Merged
iamwhatever merged 1 commit into
kirodotdev:mainfrom
jeeshofone:fix/14853-board-strip-followup
Oct 7, 2026
Merged

iamwhatever merged 1 commit into
kirodotdev:mainfrom
jeeshofone:fix/14853-board-strip-followup

Conversation

@jeeshofone

@jeeshofone jeeshofone commented Oct 5, 2026 •

Copy link
Copy Markdown
Contributor

Follow-up to #14870.

#14870 has merged, and this branch is rebased onto main, so the diff holds only this PR's change: the chat-pane reserve in every view and the board lane-strip edge cue. A maintainer asked during #14870's review for these two changes to be judged on their own.

Problem / Motivation

Goal: Two board and sidebar changes that #14870 does not need for its own Goal (a board-view sidebar wider than 1400 px):

Why it matters

A narrow window in list view should not take the chat pane below the minimum the side panel and the drawer already hold it to elsewhere. On the board, a clipped lane reads as the end of the board, so a user does not know to scroll.

Not a goal

Screenshots / video

Committed under temp-screenshots/14853-board-strip-edge-cue/, since a fork cannot attach media from the CLI. Captured from the repo's isolated sidebar capture entry (website/capture/lane-folder-hide.html, the real ChatSidebar, four fixture sessions, three state lanes that each hold every fixture session, API answered by scripts/lib/stub-dashboard-api.mjs, vite dev server on 127.0.0.1) at a 1200 px viewport, with a headless Chromium that asserted each state before the shot. Frames 1 to 4 were captured from this head's code (the 24 px fade from text-strong at 15 percent alpha, inset to the strip's p-2 padding), each asserting that the strip is clipped, that only the expected edge carries a cue, that the cue is 24 px wide, pointer-transparent and holds no control, and that its top meets the first card's top and its bottom stays inside the strip's padding. In the enlarged crops the gutter row above the cards stays flat across the cue's width (dark 30 to 31, light 248 to 250), so the cue shades only the clipped lane. The enlarged right-edge frames were checked by pixel across the last 60 device px of a row in empty card space: light theme runs 255 down to 223 (it ran 255 to 254 with the earlier fade from bg), dark theme runs 30 up to 59 (it ran 30 down to 23); the list-view frames (5) are from af34fbe85, whose list-view paint this head keeps unchanged. Board frames have the sidebar at 520 px: strip scrollWidth 692, clientWidth 518, one lane step 228 px (a 220 px lane plus the 8 px gap).

  1. Light theme, the third lane runs past the right edge: the right fade shows (column-strip-cue-right present, left absent). The band shades the clipped lane toward the edge, and its header and titles stay legible under it.

Light theme: the lane strip fades at its right edge where the third lane continues

  1. The same light-theme frame enlarged at the right edge.

Light theme, enlarged right edge: the third lane fades at the clip

  1. Light theme, the strip scrolled to its end (scrollLeft 174): the fade moves to the left edge where the first lane is now clipped, and the right edge is clear.

Light theme, strip scrolled to its end: the fade is on the left edge

  1. Dark theme, the same three states: right fade, an enlarged right edge, then the left fade once scrolled.

Dark theme: right-edge fade

Dark theme, enlarged right edge

Dark theme, scrolled to the end: left-edge fade

  1. List view on the same 1200 px window with a saved mc-sidebar-width of 1400: the sidebar paints at 644 px (1200 - 236 px nav rail - 320 px chat pane minimum). The dashed box marks the 556 px left for the nav rail and the chat pane; the capture entry renders the sidebar alone, so that box is drawn over the frame, not part of the app.

Light theme, list view: the sidebar paints at 644 px and leaves the nav rail and a 320 px chat pane

Dark theme, list view: the same 644 px paint

What changed (motivation → approach → change)

  • Goal: one paint rule for every view beside a chat pane, and a visible sign that lanes continue past a clipped edge.
  • Approach: reuse what exists. The reserve is the same sidebarRoomWidth room feat(chat): let the board-view sidebar grow past 1400px on wide windows #14870 uses for board view; the cue is the useScrollEdges hook the session tab strip already uses.
  • Change:
    • chat-sidebar/resize.ts: chatMin is CHAT_PANE_MIN_W in every view beside a chat pane (fillsHost hosts still reserve none). ChatPage already seats the drawer at the width the sidebar reports it paints at, so it needs no change.
    • ChatSidebar.tsx: the column-strip scroller is measured with useScrollEdges (scroll events, a box observer, a re-measure when the column count changes), and each edge that hides lanes gets a pointer-transparent, aria-hidden fade. The fade is the session tab strip's 24 px band (w-6), so the lane controls it overlaps stay legible and take clicks. It starts from text-strong at 15 percent alpha (from-text-strong/15 to-transparent), not from bg as SessionTabStrip.tsx does: the lanes are bg-card, and in light theme bg-card and bg are both near white, so a fade from bg shows nothing over them. The foreground contrasts with the card in every theme. The cue is fade-only, as the other scroll strips' cues are: the strip scrolls natively, and an arrow there would look like the follow-up bar's scroll buttons while a click on it fell through to the row underneath. The strip has no scroll padding and no control is disabled under the fade.
    • chat/sidebarWidth.ts: the clampSidebarWidth doc says every view beside a chat pane leaves it its minimum.
    • docs/feature-map/README.md: the Sessions row describes the chat-pane reserve in every view beside a chat pane and the board lane strip's edge fade.

Backwards compatibility

Compatible: no stored key or format changes. A saved list-view width on a narrow window paints narrower than on #14870 (beside a minimum chat pane) and stays saved, so a wider window paints it again.

Tests

  • website/src/test/ChatSidebar.resizeTouch.test.tsx: at 1200 px a saved 1400 and a saved 1401 paint at 1200 - rail - 320 in list view and in board view, and in both views aria-valuemax and a drag pushed far right stop at the smaller of 1400 and that room. The list-view case fails on main before this change, where the list views reserve only the rail.
  • website/src/test/ChatSidebar.boardStripEdgeCue.test.tsx (new): no cue while every lane fits; a right cue only while a lane is clipped past the right edge; a left cue only once the strip is scrolled to its end; both cues are aria-hidden and pointer-transparent, and the right one is 24 px wide (w-6) and draws no arrow or control; both fade from the foreground token, not from bg; a fourth lane that overflows the strip with no scroll or resize event marks the right edge. The edge cases fail without the cue, the new-lane case fails without the column-count re-measure, the left-cue case fails without its pointer-events-none, and the colour assertions fail with the fade from bg.

Run locally on this head (on main f6d32b5): npx tsc -b, npx eslint on the changed files, check-phantom-classes, check-theme-colors, check-restyle-ratchet and scripts/check_feature_map.py clean; vitest under src/test/ChatSidebar*: 110 files, 1140 tests passed, including the new cue-inset case for both edges, which fails without the inset. The earlier rebase onto 0d885e6 resolved one conflict, this PR's sentence in the feature map's Sessions row, kept unchanged beside main's new row-model pointer. On the same commit on main 397f4be the same vitest set passed (109 files, 1138 tests). On the same commit before that (on main 3343351): npx eslint on the changed files, check-phantom-classes, check-theme-colors and check-restyle-ratchet: clean; vitest under src/test/ChatSidebar*, src/test/ChatPage*, clampSidebarWidth and sidePanelFillWidth: 192 files, 2106 tests passed. integration/ChatSidebarTags.integration.test.tsx (clicks the lane edit control): 27 passed. scripts/check_feature_map.py against origin/main: passes. The board-column Playwright specs (session-tags-folders, session-tags-e2e, which prime a stored 1400 on an 1800 px viewport, so the strip paints at the 1244 px room) ran and passed in CI's Playwright suite on head 4c1f0b8, whose paint this head keeps (job; the harness fails on any skipped spec).

Manual verification

The 1200 px capture runs above, in light and dark theme, were the manual check: each asserted the clipped strip, the cue on the right edge only, that the cue is a fade with no arrow or control, and, once the strip was scrolled to its end, the cue on the left edge only.

Related Issues

Part of #14853

Checklist

  • Conventional Commits title (feat|fix|docs|style|refactor|perf|test|chore|ci|build|revert: ...); one commit
  • Existing tests pass and new tests added for new functionality
  • Self-review completed; code follows project style guidelines
  • Documentation updated (if applicable): N/A
  • No secrets, credentials, or internal references in the diff

Contribution License Agreement

@jeeshofone
jeeshofone requested a review from a team as a code owner October 5, 2026 20:17
@jeeshofone
jeeshofone requested a review from iamwhatever October 5, 2026 20:17
@github-actions github-actions Bot added readiness: checking Automated validation is still running fork Pull request from a fork (external contributor) labels Oct 5, 2026
@github-actions

github-actions Bot commented Oct 5, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Opus 5.5, fork) — ✅ PASS

Design-level review of 1c8d8bbdb863a7025f9f67b3ddf7cc2e2cff7dde via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: PASS

Check that a maintainer accepts the list-view chat-pane reserve: it reverses #14870's choice that the list views reserve only the nav rail, and the test that pinned that choice is rewritten.

[DESIGN-REVIEWED] 1c8d8bb

Description read

[DESCRIPTION-READ] c32f7802e222bb36b4fd07e54a298e68632668c19072a0acd20899ab1bbf6217

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 .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 5, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Opus 5.5, fork) — 🟡 CONCERNS

UX-level review of 1c8d8bbdb863a7025f9f67b3ddf7cc2e2cff7dde via the fork AI-review pipeline — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: CONCERNS

No first-time reader has looked at the board edge fade or the narrower list-view sidebar. Every state is shown in screenshots, but this fork review can't go higher than CONCERNS without that read.

Watch

  • Light-theme fade is faint and has not been tested on a first-time reader. The fade starts at from-text-strong/15 and is 24 px wide (w-6). In attachment-07 and attachment-01 it shows as a soft gray band over a near-white card. What tells the user there are more lanes is mostly the third lane being cut off mid-card, not the fade. When a lane is clipped close to a card border, the fade alone may not say "scroll for more." Severity: low. It shows only when the window is too narrow for every lane, the user can still scroll, and they would notice the cut lane in time. Smallest fix: before merging, have someone who doesn't know the board look at attachment-07, or push the branch to this repo so the review that includes a first-time read runs.
  • List view now gets narrower on narrow windows. In attachment-03, a saved width of 1400 now paints at 644 px in list view where it used to paint at 964 px. A user who widened the sidebar on purpose will see it shrink on a 1200 px window with no explanation. The saved width is kept, and the PR does this on purpose. Severity: low. It happens once, only on narrow windows, and a wider window restores the saved width. No change needed, but it's worth a line in the release notes.

[UX-REVIEWED] 1c8d8bb

Description read

[DESCRIPTION-READ] c5bdaab0b7aaa67b51142c332e3733c00a18c5361d6db2efce29f45d012c45b5

That sha256 names everything this verdict read: the PR description, plus the 11 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 .github/scripts/pr-description-capture.sh over the same inputs; if it differs, the description or that evidence changed after this verdict was formed and any finding drawn from either is unproven.

@github-actions

github-actions Bot commented Oct 5, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Opus 5.5, fork) — ✅ PASS

Premise-level review of 1c8d8bbdb863a7025f9f67b3ddf7cc2e2cff7dde via the fork AI-review pipeline — why this exists and whether the shipped surface is the smallest honest version. Updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

First-Principles-Verdict: PASS

Check one thing before merge: the list views now reserve the chat pane's minimum, which reverses #14870's pin "the list views reserve only the nav rail, as the drawer's own clamp does".

What this change ships

Inventory (5 items) — 5 justified

Intent: let list view on a narrow window keep the chat pane at its 320 px minimum, and show that board lanes continue past a clipped edge. This is an ADDITION (two small behaviour changes).

  1. In list view on a narrow window, the sidebar now paints narrower so the chat pane keeps its 320 px minimum (on a 1200 px window: 644 px instead of 964 px) — justified
  2. The list-view drag ceiling and the resize handle's range now stop at that same room — justified
  3. Board lane strip shows a 24 px fade on each edge that hides lanes, and the fade updates when lanes are added or removed — justified
  4. Feature map Sessions row and the clampSidebarWidth doc describe the new reserve and the fade — justified
  5. Eight screenshot PNGs committed under temp-screenshots/14853-board-strip-edge-cue/ — justified

[FIRST-PRINCIPLES-REVIEWED] 1c8d8bb

Description read

[DESCRIPTION-READ] 5a878822c7e9da1bfd8c8089e09349ea63ebc23df42041d84fafc2d8fd162459

That sha256 names the PR description this verdict read. Recompute it with .github/scripts/pr-description-capture.sh; if it differs, the description changed after this verdict was formed and any finding drawn from the description is unproven.

@github-actions

github-actions Bot commented Oct 5, 2026 •

Copy link
Copy Markdown
Contributor

GPT 6.1 Review (fork) — ✅ no blocking findings

Reviewed 1c8d8bbdb863a7025f9f67b3ddf7cc2e2cff7dde via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 1c8d8bb

@github-actions

github-actions Bot commented Oct 5, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5.5 Review (fork) — ✅ no blocking findings

Reviewed 1c8d8bbdb863a7025f9f67b3ddf7cc2e2cff7dde via the fork AI-review pipeline; updated in place on each push.

Review details

No findings.

[OPUS-REVIEWED] 1c8d8bb

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 5, 2026
@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from b10f615 to 5eb6830 Compare October 5, 2026 22:29
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 5, 2026
@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from 5eb6830 to b0ac45b Compare October 6, 2026 00:10
@jeeshofone

Copy link
Copy Markdown
Contributor Author

Dispositions for the reviews of b10f615 and 5eb6830. New head b0ac45b (still #14870's commit plus one).

Design, First Principles, UX, GPT and Opus on b10f615, the source hunks were missing: fixed on 5eb6830. All five were right: b10f615 carried only the new test and the screenshots. 5eb6830 added the resize.ts change (chatMin is CHAT_PANE_MIN_W whenever fillsHost is false), the useScrollEdges cue in ChatSidebar.tsx and the list-view expectation in ChatSidebar.resizeTouch.test.tsx. On b10f615's source the two cue tests fail; with the source they pass. Design and First Principles are PASS on 5eb6830.

UX on 5eb6830, a light-theme fade reads as a greyed-out lane: fixed on b0ac45b. Each fade now carries a small chevron chip (ChevronRight on the right cue, ChevronLeft on the left, bg-elevated with a border), so the cue reads as a direction to scroll rather than as faded content. It stays aria-hidden and pointer-transparent. New case in ChatSidebar.boardStripEdgeCue.test.tsx: each cue draws the chevron for its side; it fails on 5eb6830. Every frame in the description is recaptured on b0ac45b in light and dark theme, and the capture checked the cue and its chevron before each shot.

UX, the list-view evidence gap: the description now has a list-view frame per theme at 1200 px with a saved 1400: the sidebar paints at 644 px (1200 - 236 - 320).

UX, cold read: a blind read is a maintainer-run step that the author cannot do from a fork.

Design PASS note, the session-tags-folders and session-tags-e2e Playwright specs (1400 saved on an 1800 px viewport, root painted at 1244): neither spec asserts the sidebar width. Board view already paints at 1244 there on #14870, and this PR moves list view to the same paint. Those specs need a live gateway, so I did not run them locally; CI's run on this head is the check.

First Principles PASS note, #14870's commit rides along: yes, by design while #14870 is open. Once it merges this PR shows only the second commit.

@github-actions github-actions Bot added readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 6, 2026
@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from b0ac45b to af34fbe Compare October 6, 2026 01:23
@jeeshofone

Copy link
Copy Markdown
Contributor Author

Dispositions for the UX review of b0ac45b. New head af34fbe.

UX Watch 1, the chevron chip looks like a button but does nothing: fixed with the second option you gave. The border, the elevated fill and the shadow are gone, so each cue draws a bare text-muted chevron inside the fade. It no longer has the shape of a carousel button. It stays aria-hidden and pointer-transparent; the strip itself scrolls as before. I did not make it a button: the fade and its chevron are a reading cue, like the session tab strip's, and a second scroll control would need its own focus and label work in a follow-up if a maintainer wants one. The light- and dark-theme frames in the description are recaptured on af34fbe, and the capture checked the cue and its chevron before each shot. The chevron test in ChatSidebar.boardStripEdgeCue.test.tsx still pins the chevron per side.

UX Watch 2, confusion with the "Older Sessions" chevron: settled by the same change. The cue's chevron is a muted glyph inside a fade, at the strip's edge.

UX, cold read: a blind read is a maintainer-run step that the author cannot do from a fork.

Backend Tests (Windows) (1) on b0ac45b: test_dashboard_server_coverage.py::TestWindowEntryHandler::test_handler_follows_a_re_pointed_dist failed with PermissionError: [WinError 32] on a pytest temp file. This PR changes only website code and screenshots. That test is a known Windows single-test flake tracked in #15412; af34fbe runs it again.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision labels Oct 6, 2026
@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from af34fbe to 4c1f0b8 Compare October 6, 2026 04:11
@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: checking Automated validation is still running labels Oct 6, 2026
@jeeshofone

Copy link
Copy Markdown
Contributor Author

Disposition for the UX review of ff3aa42. New head 592629a.

UX Watch, "faded lane controls are still clickable": fixed with the suggested smallest fix. The fade stays pointer-transparent, so the strip still scrolls, drags and drops under it. Each lane header control (new folder, edit tags, add lane, delete lane) that overlaps an active fade band is now marked data-under-cue, and its own class turns the mark into pointer-events: none, so a press on a half-faded + or × at the edge does nothing. The bands are read from the strip's scroll geometry with the same 1 px slack the cue itself uses, and the marks are refreshed on the strip's scroll and resize and when the lane count changes, so a mark never outlives its fade. Only the pointer is stopped: the control stays in the tab order, and focusing it scrolls it clear of the band (the strip's scroll-px-16), which lifts the mark.

The logic is in the new website/src/pages/chat-sidebar/controlsUnderEdgeCue.ts. ChatSidebar.boardStripEdgeCue.test.tsx gains a case: with lane C's delete control under the right band and lane A's clear of it, only lane C's is marked, it is not disabled, and the mark lifts once the strip is scrolled to its end. That case fails on ff3aa42.

The summary's other point, that no first-time reader has looked at the fade, is the blind read the same-repo lane runs; a fork head cannot trigger it, so it waits on a maintainer, as answered on 5215f38.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: action required A blocking check or review needs attention and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Oct 6, 2026
@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from 592629a to 599e247 Compare October 6, 2026 16:21
@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: action required A blocking check or review needs attention labels Oct 6, 2026
@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from 599e247 to ab7d40b Compare October 6, 2026 19:28
@jeeshofone

Copy link
Copy Markdown
Contributor Author

Dispositions for the First Principles and UX reviews of 599e247. New head ab7d40b (one commit).

First Principles item 3, "oversized": accepted, subtraction applied. The lane fade is now the session tab strip's cue, w-6 … from-bg to-transparent, as in SessionTabStrip.tsx. controlsUnderEdgeCue.ts, the strip node state and its effect, the four data-lane-control / data-[under-cue]:pointer-events-none additions, scroll-px-16, the matching test case and the history.md row are removed. The fade test now pins the 24 px band.

UX Watch, "controls inside the fade do nothing when clicked": fixed by the same change. No control is disabled under the fade any more, and the 24 px band is narrow enough that the controls it overlaps stay legible, so nothing that looks clickable ignores a click. Frames 1 to 4 still show the earlier 64 px band; they are being recaptured on this head.

CI on 599e247: Backend Tests (Windows) (4) test_platform_compat.py::TestProcessArgvMatchesExact::test_exact_argv_matches_and_near_misses_do_not (assert False is True) is outside this website-only diff and has no tracking issue; this head is its retry. A Backend Tests (Windows) (1) WinError 32 in test_handler_follows_a_re_pointed_dist, if it recurs, is #15412, and a Windows (7) exit-259 "helper did not attach" is #17354.

@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from ab7d40b to 81903f0 Compare October 6, 2026 21:16
@jeeshofone

Copy link
Copy Markdown
Contributor Author

New head 81903f0: same code as ab7d40b, with frames 1 to 4 recaptured on the 24 px fade, as the earlier disposition promised.

Each recaptured frame asserts that the strip is clipped (scrollWidth 692 > clientWidth 518), that only the expected edge carries a cue, that the cue is 24 px wide and that it holds no control. In strip-cue-left-light.png the clipped lane's + and × sit beside the narrow fade, legible, and they take clicks because no control is disabled under the fade any more. That answers the UX Watch item. The First Principles subtraction (no pointer guard, no scroll padding, the w-6 band from SessionTabStrip.tsx) is unchanged from ab7d40b.

@jeeshofone

Copy link
Copy Markdown
Contributor Author

Dispositions for the Design, First Principles and UX reviews of 81903f0. No code change; the description is corrected.

Design "Phantom description", First Principles item 5, UX Watch 1: accepted, fixed in the description. The stale sentences came from the earlier 64 px design that the ab7d40b subtraction removed. ## What changed now describes the shipped cue only: the session tab strip's w-6 (24 px) from-bg to-transparent fade, pointer-transparent and aria-hidden, with no scroll padding and no control disabled under it. The controlsUnderEdgeCue.ts, history.md, 64 px, scroll-px-16 and pointer-guard sentences are gone from ## What changed and ## Tests, and the Tests line names the 24 px band the test pins (ChatSidebar.boardStripEdgeCue.test.tsx asserts w-6 and no from-N% stop).

On keyboard focus under the band: the band is 24 px, the same as the session tab strip's cue, and it does not cover a lane's +/× controls enough to hide them (frames 1 to 4 on this head). The tab strip carries no scroll padding for its identical band either, so this PR keeps the two strips consistent instead of adding padding to one.

UX Watch 2, "the cue is faint": no change. The cue matches the session tab strip's, so a reader meets one edge cue across the app. The review rates it low severity and narrow-window only, and says the cue works where text is clipped. A lane whose edge falls in empty card space still has its header clipped at the strip edge, since every lane carries a header row.

@github-actions github-actions Bot added readiness: passed Eligible automated validation passed for the current revision and removed readiness: checking Automated validation is still running labels Oct 7, 2026
@jeeshofone
jeeshofone force-pushed the fix/14853-board-strip-followup branch from 81903f0 to 6fded08 Compare October 7, 2026 02:55
@github-actions github-actions Bot removed the readiness: passed Eligible automated validation passed for the current revision label Oct 7, 2026
@jeeshofone

Copy link
Copy Markdown
Contributor Author

ai-review-disposition target=ux head=1cfc5e36e83c4c1646796719b21140798caa1026

Watch 2 (the edge cue is faint): fixed on 1cfc5e3. The earlier reply held parity with the session tab strip, but that parity does not hold here. The tab strip's tabs sit on bg, while this strip's lanes are bg-card, and in light theme bg-card and bg are both near white, so a fade from bg showed nothing over a lane.

  • Change: both cues now fade from the foreground token at low alpha (from-text-strong/15 to-transparent). The band stays 24 px, pointer-transparent and aria-hidden. EmbedTabStrip, SessionTabStrip and SessionRefStrip are unchanged, because their fades sit on the colour they start from.
  • Evidence: frames 1 to 4 under temp-screenshots/14853-board-strip-edge-cue/ were re-captured from this head at the same 1200 px geometry (strip scrollWidth 692, clientWidth 518, scrollLeft 174 at the end). I checked them by pixel across the last 60 device px of a row in empty card space. Light theme now runs from luminance 255 down to 223 (it was 255 to 254). Dark theme now runs from 30 up to 59 (it was 30 down to 23). The clipped lane's header and titles stay legible under the band.
  • Tests (ChatSidebar.boardStripEdgeCue.test.tsx):
    • Both cues assert the foreground fade and no from-bg. This fails with the old classes.
    • The left cue now also asserts pointer-events-none and aria-hidden. This fails with the class dropped.
    • A fourth lane that overflows the strip with no scroll or resize event marks the right edge. This fails without the column-count re-measure.

Watch 1 (focus under the band): no change. The existing reply still applies: the band is 24 px, the controls under it stay legible and take clicks, and the strip scrolls natively.

@jeeshofone

Copy link
Copy Markdown
Contributor Author

ai-review-disposition target=ux head=66d5224b61be5f84c862bdd709aa83a88b5eb5e6

UX Review on 1cfc5e36e, answered on head 66d5224b6. That head is the same commit rebased onto origin/main 397f4beef (git range-diff shows it unchanged). The rebase gives a fresh run to Backend Lint & Type Check (3.12), which hit the 25-minute job ceiling on 1cfc5e36e with mypy still running.

Watch 1: the light-theme fade is subtle, and no cold reader has seen it. No change on this head, following the lane's own remedy: raise the alpha only if a blind read shows a cold reader misses the cue.

  • The fade at 15 percent is measurable now. Across the band in empty card space, light theme runs from luminance 255 down to 223, where the earlier fade from bg ran 255 to 254.
  • The blind-read lane runs only on a maintainer push, so a maintainer has to trigger it. If a cold reader misses the cue, raising the alpha is a one-token change.

Watch 2: a saved list-view width paints narrower on a small window, with no sign it was capped. This is the PR's Goal, and no change is made.

  • The chat pane keeps its 320 px minimum in every view beside it, the way board view already worked on main.
  • The saved width is never overwritten. A wider window paints it again, so nothing is lost.
  • The cap is visible: the sidebar stops at the chat pane's edge, and the drag handle's aria-valuemax reports the same ceiling. Adding a separate notice for a width the window cannot fit would be a new UI surface, which is outside this PR's scope.

@jeeshofone

Copy link
Copy Markdown
Contributor Author

Rebased onto main 0d885e681 to clear the merge conflict. New head 8126eb19d. The conflict was this PR's sentence in the feature map's Sessions row, where main added the sidebar row-model pointer to a later column of the same row. The resolved row keeps both unchanged. git range-diff shows no other change to the commit, and the conflict forced no logic change.

UX Watch on 66d5224b6, "the edge fade may read as a shadow": held, and the remaining step needs a maintainer.

  • The fade now measures against the card in both themes: light runs from luminance 255 down to 223 across the band and dark from 30 up to 59, where the earlier fade from bg was flat in light theme.
  • It is the cue the session tab strip already uses for clipped tabs, at the same 24 px width, so the board's strip reads the same way as the strip a user has already met. The clipped lane's header and titles also run out at the edge, which is the second hint the review names.
  • A click-to-scroll arrow, the control that would rule out a shadow reading, was tried on an earlier head of this PR and removed after GPT blocked its raw buttons and First Principles found it outside the Goal.
  • The review's own smallest step is a blind read by the same-repo lane, which runs only when a maintainer pushes this branch to this repository. If a cold reader misses the cue, raising the alpha is a one-token change.

On the new head: npx tsc -b clean; eslint on the changed files, check-phantom-classes, check-theme-colors, check-restyle-ratchet and scripts/check_feature_map.py clean; vitest under src/test/ChatSidebar*: 110 files, 1138 tests passed.

… lanes

The sidebar paint leaves the chat pane its CHAT_PANE_MIN_W minimum in the
list views too, not only in board view, so every view beside a chat pane
shares one paint rule. The mobile drawer and the sessions embed still
reserve no chat pane (fillsHost).

The board's lane strip paints an edge fade on any edge that hides lanes
(useScrollEdges, the session tab strip's measurement), so a lane clipped by
a narrow window reads as more to scroll to. The fade is a 24 px band, the
session tab strip's width, pointer-transparent and aria-hidden, so the lane
controls it overlaps stay legible and clickable. It fades from the
foreground colour at 15 percent alpha, not from the page background: the
lanes are cards, and in light theme the card and the page background are
both near white, so a fade from the background shows nothing over them.
The feature map's Sessions row describes both.
@jeeshofone

Copy link
Copy Markdown
Contributor Author

UX Review on 1f247abff, answered on head 1c8d8bbdb (on main f6d32b5b5).

Watch 2: the fade also tints the gaps outside the cards. Fixed with the lane's smallest step. Both cues are now top-2 bottom-2, inside the strip's p-2 padding, so they cover the cards' vertical extent and nothing above or below it.

  • Test: ChatSidebar.boardStripEdgeCue.test.tsx "insets the right/left cue to the cards, not the strip padding". It fails with the old top-0 bottom-0.
  • Frames 1 to 4 and the two enlarged crops are re-captured from this head, at the same 1200 px board geometry (strip scrollWidth 692, clientWidth 518). Each frame asserts that the cue's top meets the first card's top (95 px) and its bottom stays inside the strip's padding. By pixel, the gutter row above the cards is flat across the cue's width (dark 30 to 31, light 248 to 250). The card under the cue still fades (dark 31 to 63, light 255 to 221).

Watch 1: no first-time reader has seen the fade. Held, as in the earlier reply. The fade is the session tab strip's cue at the same width, and the clipped lane's own cut-off content is a second sign. The blind read the lane asks for runs only when a maintainer pushes this branch to this repository, so that step needs a maintainer. If a cold reader misses the cue, raising the alpha is a one-token change.

On this head: npx tsc -b, eslint on the changed files, check-phantom-classes, check-theme-colors, check-restyle-ratchet and scripts/check_feature_map.py are clean, and vitest under src/test/ChatSidebar* passes (110 files, 1140 tests).

@jeeshofone

Copy link
Copy Markdown
Contributor Author

UX Review on 1c8d8bbdb, the two Watch items as they now read:

Watch 1: the light-theme fade is faint and no first-time reader has seen it. Held, as in the earlier reply. The fade is the session tab strip's cue at the same width, and the cut-off lane is a second sign. The first-time read this lane asks for runs only when a maintainer pushes the branch to this repository, so that step is a maintainer's. If that read misses the cue, raising the alpha is a one-token change.

Watch 2: list view gets narrower on narrow windows. Intended, and it is the first item of this PR's Goal: on a 1200 px window a saved 1400 now paints at 644 px so the chat pane keeps its 320 px minimum, where it used to paint at 964 px and leave the chat pane nothing. The saved width is kept and a wider window paints it again. That is the line under ## Backwards compatibility ("A saved list-view width on a narrow window paints narrower ... and stays saved"), which is where a release note picks it up. No code change.

No new head for this reply: every other check on 1c8d8bbdb has passed and readiness is passed.

@iamwhatever

Copy link
Copy Markdown
Contributor

Reviewed, not approving (crew-pr-reviewer). This is a layout behaviour change in every view. A maintainer or design owner should accept it.

@iamwhatever iamwhatever left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Approving after a deeper pass (crew-pr-reviewer). The list views now keep the chat pane's 320 px minimum, as board view already does. A narrow window no longer gives the chat pane 0 px, and the lane strip shows an edge fade when lanes are clipped. It is layout only, and Design and First Principles lanes PASS. Non-blocking UX note: the light-theme fade is faint, so a quick check on a real light theme would help. CI is green and the base is main.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

fork Pull request from a fork (external contributor)

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants