Skip to content

fix(ui): glass dock follow-ups: neutral focus, chip hairline, hover fallback - #14799

Closed
CrysisDeu wants to merge 1 commit into
mainfrom
fix/glass-followups
Closed

CrysisDeu wants to merge 1 commit into
mainfrom
fix/glass-followups

Conversation

@CrysisDeu

@CrysisDeu CrysisDeu commented Sep 28, 2026 •

Copy link
Copy Markdown
Contributor

Problem / Motivation

Goal: The chat composer reads as one piece of Liquid Glass: focus is the glass getting clearer, nothing turns the theme colour, and every glass pane shows the same edges.

Follow-ups on the glass dock that shipped in #14568, and one re-tune of the material itself against the reference:

  • The composer dock still wore composer-halo: an --accent-glow shadow on focus plus a 1px accent ring in dark themes, and the input wrapper turned its border accent/50. Every other glass pane (Settings capsule, cards, chips) already used the neutral glass-shadow.
  • The follow-up chips lost their top and bottom hairline. The composer and the Settings capsule show it; the chips did not.
  • In the three solidifying fallbacks (@supports not backdrop-filter, prefers-reduced-transparency, prefers-contrast: more) a hovered chip did not change at all, and the jump-to-bottom pill sat 8px above the dock, glass on glass.
  • Measured against the reference material in dark mode, ours was a flat grey slab: the dark tint was a 40% BLACK fill (the glass could only ever land below the page), the side line read +20 over the page where the reference reads about +100, and a 6px blur smeared a white bar's edge over 14px where the reference smears it over about 5. Text typed into the composer used the theme's --text, not the white the reference types in.
  • The context shelf (agent, project, branch, model chips) sits below the glass on the bare transcript, and the dock floats over content, so a chip label was reading against whatever scrolled under it.

Why it matters

The composer is the most-focused element in the product. A theme-coloured glow there, and none anywhere else, made the glass look like two materials. A missing hairline on chips is the same tell in miniature. A hover that does nothing in the reduced-transparency modes is a lost affordance for exactly the users who asked for a plainer surface.

Not a goal

  • The light-mode tint, the bands and the hairline alpha are unchanged from feat(ui): one glass dock for the composer band #14568. What moves: the dark tint, the dark side line, the frost, and the primitive's inner bevel glow (universal, but measured invisible on a light pane; see the table).
  • No new knob on Glass / LiquidGlass: the retune is token and recipe values.
  • The Settings capsule's own focus cue is unchanged; it already used glass-shadow.
  • No change to the approval glow's colour or pulse (approval-glow): a pending decision still lights the dock warm. Under prefers-reduced-motion it now holds a steady ring instead of no ring (see below).

What changed (motivation → approach → change)

The composer's focus is the glass, not a theme color

Symptom: focusing the composer painted an accent halo and an accent border. Cause: the dock carried composer-halo and the wrapper focus-within:border-accent/50, both predating the shared material. Change: the dock wears glass-shadow like every other pane, and the wrapper's border stays transparent in the default memory mode (incognito / temporary keep their warn / aim border, which is a mode signal, not a focus cue). Focus is the material getting a step clearer: --glass-tint steps to --glass-tint-focus and the side lines to --glass-edge-focus, the same cue as the Settings capsule and the cards. composer-halo and its -warn / -aim forms are deleted from index.css.

While a tool approval is pending the dock is glass-shadow approval-glow: the glow takes the box-shadow slot (--approval-shadow, animated as before) and the tint + edge focus step still applies, so the textarea keeps a visible focus cue in that state too, in every theme and in the three fallbacks (their .glass-shadow:focus-within outline now matches).

Measured on this head's build (harness asserts it on every run):

rest shadow focus shadow tint wrapper border
dark rgba(0,0,0,.30) 0 0 18px rgba(0,0,0,.55) 0 0 18px rgba(127,134,147,.30) -> rgba(127,134,147,.42) transparent, unchanged
light rgba(0,0,0,.06) 0 0 18px rgba(0,0,0,.14) 0 0 18px #eeeef373 -> #ffffffeb transparent, unchanged

Composited interior through the input: dark #383742 -> #444451 (luminance 56 -> 68), light 247 -> 255; dark side line 143 -> 213. With an approval pending: shadow stays the warm glow, tint and edge step the same way.

The dark material is the reference's colour, not a darkening

Symptom: in dark mode the dock read as a flat grey slab beside the reference, and the blur hid what scrolled under it. Cause: --glass-tint was rgba(30,30,34,.40), a black fill, so the glass interior could only land below the page; the side line was 14% white (+20 over the page, the reference's rim is about +100); frost was 6 / 4. Change (tokens and the two recipe numbers only):

before after
dark --glass-tint rgba(30,30,34,.40) (black) rgba(127,134,147,.30): the reference's composited colour #37393D solved back into a translucent fill over the app's own dark page (#181818): 24 + 0.30 x (127 - 24) = 55. Measured on the build: #383742 at rest. (Solving it over pure black instead -- a 24% tint of rgb(222,234,255) -- matched the reference on black but composited to #6A6B79 on the real page while focused, far too bright.)
dark --glass-tint-focus rgba(84,84,92,.72) rgba(127,134,147,.42): #444451 measured
dark --glass-edge / -focus 14% / 55% white 35% / 70% white
frost (Glass recipe) panel 6, chip 4 panel 3, chip 2: a white bar's edge under the pane now smears over ~5px, was 14
inner bevel glow (liquid-glass.tsx) 0.2 x light alpha, 1x reach BEVEL_GLOW = 0.48, BEVEL_REACH = 2.5: the reference's rim carries a white glow fading inward over ~5pt (+25-35 over the page); ours had almost none. Dark top edge inward, 2x pixels: 66 64 60 56 52 49 44 -> 113 112 109 107 104 101 98. This layer paints in light mode too, where white on a near-white pane is invisible: every light edge profile is within 1/255 of the previous head.
composer ink theme --text / --muted new --glass-text / --glass-placeholder: #ffffff / #9c9c9c on smoked glass, the theme's --text / --muted on a light pane (white would vanish)

Light-mode tint, bands and hairline are untouched. The tokens stay polarity-fixed (website/docs/theming-contract.md). The retune applies to every Glass pane, the mobile Settings capsule included (frames below); the capsule has no shelf under it, so the fade is the composer's alone.

The context shelf stands on a short fade to page colour

Symptom: the chips under the pane (agent, project, branch, model) read against transcript text scrolling under the floating dock. Change: the shelf is glass-shelf, a ::before that fades from nothing at the pane's bottom edge to solid --bg by 60% of the shelf's height. It is positioned against the dock's relative z-10 dock-inert wrapper (the shelf itself is deliberately not positioned), so it spans exactly the dock root -- which already stops short of the scroller's reserved scrollbar column, so the thumb stays uncovered -- and its z-index: -1 resolves in that wrapper's stacking context, BEHIND the pane and its .glass-shadow: the shadow falls onto the solid strip the way it fell onto the page. (A first cut painted the fade over the shadow, which whitened the shadow out part-way down and left a bright band under the pane that read as a reflection.) ChatInput hands it the shelf's border-box height as --glass-shelf-h so it starts at the pane's edge. Nothing paints inside the pane: the glass above the shelf is untouched (the harness asserts the pane's own ::before carries no gradient). Measured with a message body scrolled under the whole dock (2x pixels): text under the glass peaks at 152 over the 108 interior; under the shelf it is 37 at the fade's top and gone (page 24) by its lower half; light 255 throughout.

Secondary ink and the strip edge cues follow the material

Cause (Opus, round 3): the retuned dark tint composites to about rgb(67,71,81) over the default page, where the theme's --muted #7f7f88 keeps barely 2:1 -- half of what it had on the near-black fill. Change: .liquid-glass { --muted: var(--glass-muted) }, a polarity-fixed #b4b4bc on smoked glass and the theme's own --muted on a light pane, so every text-muted inside a pane (chip labels, the memory chip, card hints, the control row) brightens with the glass. Measured on the follow-up chips, dark: muted ink 129 -> 154 (max 181); light frames byte-identical.

Cause (Opus, round 3): the file-preview strip's and the control row's overflow cues faded from-bg-elevated, an opaque colour the glass no longer has, so at narrow widths they would have painted a dark wedge on the pane. Change: the cues are backdrop-blur wedges (backdrop-blur-[3px], masked to nothing toward the strip), which soften the clipped chip instead of covering it; same elements, same useScrollEdges gate, same tests. Where the pane solidifies (the three fallback blocks, including @supports not backdrop-filter, where a blur wedge paints nothing) glass-strip-cue-l/-r swap the wedge for a fade from --bg-elevated, the pane's own colour there (Design, round 3).

A pending approval keeps its glow under reduced motion

Cause (Opus, round 2): the global reduced-motion rule runs every animation once for 0.01ms, which parks approval-glow at its 0% keyframe, where --approval-shadow is fully transparent; the dock that hands its shadow slot to that variable then showed no shadow and no glow. Change: @media (prefers-reduced-motion: reduce) { .approval-glow { animation: none; --glow-strength: .6 } }, a steady mid-pulse ring.

Chips get their top and bottom hairline back

Symptom: no hairline on chips. Cause: the chip row is overflow-x-auto; CSS makes the other axis auto too, and the 0.5px --glass-hairline is drawn just outside the box, so the row clipped it top and bottom. Change: the row gets py-px -my-px; the line paints, the row height does not change.

Chip edge profile, outer -> inner (2x pixels): dark top 24 -> 12 -> 57 47 38 (was 24 -> 67 60 56, no line); light top 255 -> 204 -> 250 249 247.

Hover survives the solidifying fallbacks; jump pill floats clear

Cause: .glass-hover steps --glass-tint, which only the frost layer reads, and all three fallback blocks hide that layer. Change: each block moves the hover step onto the solid fill (.liquid-glass.glass-hover:hover), beside the glass-accent / glass-warn lines already there; the prefers-contrast: more block uses twice the mix (16%), so the users who asked for more contrast get the clearest hover, not the faintest. The jump-to-bottom pill floats 24px clear of the dock (-top-14); the layout comments in ChatScrollChrome.tsx and ChatPage.tsx name the new figure.

Backwards compatibility

Compatible: CSS classes, tokens and two recipe numbers; composer-halo* had no consumer outside ChatInput.tsx. --glass-text, --glass-placeholder and --glass-muted are new polarity-fixed tokens, not theme roles; .liquid-glass scopes --muted to the pane. No API, config or stored data changes.

Tests

  • ChatInput.liquidGlass.test.tsx: the dock wears glass-shadow at rest and glass-shadow approval-glow while a decision is pending; the wrapper never carries focus-within:border-accent; index.css has no composer-halo rule, the neutral .glass-shadow rules carry no theme colour, and the three fallback blocks each carry .liquid-glass.glass-hover:hover and the .glass-shadow:focus-within outline.
  • ChatInput.liquidGlass.test.tsx also pins the new token values for both polarities, the glass-shelf fade rule (containing block = the dock wrapper, no viewport-wide insets, no stacking context of its own), its class and --glass-shelf-h on the shelf (and that no in-pane scrim exists), the fallback rules that turn the strip cues back into --bg-elevated fades, the scoped --glass-muted, the blur cues (no from-bg-elevated left in the file), and the reduced-motion approval-glow rule; LiquidGlass.test.tsx pins the frost 3 recipe.
  • The chip hairline and the pill clearance are geometry, pinned by the capture harness below rather than a unit test.

Manual verification

  • tsc, eslint (touched files), i18n-check, check_focus_cue.py (diff-scoped, base origin/main): green.
  • capture-liquid-glass.mjs on the built SPA: 80 frames (this head, dock-r22). The long scene now also scrolls a message body under the whole dock and asserts the shelf fade is present, starts at the pane's edge and does not touch the pane. It asserts the focus deltas above (rest vs focused, with and without a pending approval), the chip hairline and the pill clearance, and fails on any unexpected dialog. Its new contrast scene emulates prefers-contrast: more and asserts the fallback itself: the frost layers are display: none, a hovered chip changes its solid fill (dark rgb(33,29,37) → rgb(48,44,52), light 255 → 240), and the focused dock wears the 2px solid accent outline. No local test run; CI runs the suites.

Screenshots / video

Rest / focus, dark and light (the retuned dark material: reference colour, brighter side line, frost 3):

composer dark rest
composer dark focused
composer light rest
composer light focused

Approval pending, focused: the glow stays, the glass still steps clearer.

composer dark approval focused
composer light approval focused

Chips with their hairline:

chips dark
chips light
chips dark picked

Solidifying fallback (prefers-contrast: more, emulated): every pane is a flat card, the hovered chip (first chip, "用 A") takes the 16% text mix, and the focused composer gets the standard accent outline. The other two fallback blocks paint the same frame with the 8% hover mix.

contrast fallback dark, chip hovered
contrast fallback dark, composer focused
contrast fallback light, chip hovered
contrast fallback light, composer focused

Long transcript, pill floating clear of the dock:

long transcript dark
long transcript light

Mobile Settings search capsule on the retuned material, REST (left) and FOCUSED (right) in one frame per polarity. The focus cue on glass is deliberately the material itself, never a theme colour or a ring (maintainer decision): the tint steps clearer and the side lines step darker/brighter, the shadow deepens. Measured: dark interior 57 -> 69 with the side line 14% -> 55% white; light interior 247 -> 255 with the side line 24% -> 60% black and the shadow .06 -> .14 (the edge under the capsule 249 -> 239). The caret is the third cue.

settings capsule dark, rest vs focused
settings capsule light, rest vs focused

Message body scrolled under the whole dock: text shows through the glass, the shelf under it sits on the fade to page colour:

long transcript under the shelf, dark
long transcript under the shelf, light

Related Issues

Follow-up to #14568.

Pattern harvest

Rule candidate: review-prompt
Pattern: a conditional className that swaps one shadow class for another drops every other cue the first class carried (here the focus step); prefer stacking (base state) over swapping.

Checklist

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

@CrysisDeu
CrysisDeu requested a review from a team September 28, 2026 18:42
@CrysisDeu
CrysisDeu requested a review from a team as a code owner September 28, 2026 18:42
@github-actions github-actions Bot added the readiness: checking Automated validation is still running label Sep 28, 2026
@github-actions

github-actions Bot commented Sep 28, 2026 •

Copy link
Copy Markdown
Contributor

Design Review (Fable 5) — 🟡 CONCERNS

Design-level review of 7fdc2b9d03ea279bde3310f179140603ea356b0d — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

Design-Verdict: CONCERNS

The shelf scrim's geometry hangs on an implicit containing-block chain across ChatPage/ChatInput, asserted only by a script CI never runs.

Watch

.glass-shelf::before positions against ChatPage's relative z-10 dock-inert wrapper only because the shelf "is deliberately not positioned" — and because nothing between them (Composer → ChatInput's containers) ever gains position, transform, or filter. In a codebase that animates these very containers with framer-motion (which writes inline transforms), a later unrelated change silently re-scopes the fade to the wrong box, the chips lose their scrim, and nothing goes red: the geometric assertion lives in capture-liquid-glass.mjs, whose own header says "Nothing in CI runs this file"; the vitest pins are source-text regexes that can't see an ancestor's transform.
Clears when: a CI-run test asserts the shelf's computed containing block is the dock wrapper (or the scrim moves onto the wrapper itself, making the pseudo-element's containing block explicit).

[DESIGN-REVIEWED] 7fdc2b9

Description read

[DESCRIPTION-READ] c6f303698bd4d8ad6b3e4aa662da4c382b6b1cba3ad195e16928508a9812b856

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 Sep 28, 2026 •

Copy link
Copy Markdown
Contributor

GPT 5.6 Review — ✅ no blocking findings

GPT 5.6 completed its review of 7fdc2b9d03ea279bde3310f179140603ea356b0d and found no blocking issues.

This comment is updated in place on each push.

Review details

No findings.
[GPT-REVIEWED] 7fdc2b9

False positive or not applicable? A repository writer can comment:
/ai-review override gpt 7fdc2b9d03ea279bde3310f179140603ea356b0d: <one-sentence reason>

@github-actions

github-actions Bot commented Sep 28, 2026 •

Copy link
Copy Markdown
Contributor

First Principles Review (Fable 5) — ✅ PASS

Premise-level review of 7fdc2b9d03ea279bde3310f179140603ea356b0d — 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

Every measured claim (tint composite, blur smear, bevel profile) is asserted only by capture-liquid-glass.mjs, which nothing in CI runs — run it once on this head.

What this change ships

Inventory (10 items) — 10 justified

Intent: make the composer read as one piece of the shared glass material — neutral focus, matching edges, working fallbacks — and retune the dark glass against the reference. This is a FIX.

  1. Composer focus is now the glass clearing (tint + edge step), not an accent glow/border; composer-halo deleted — justified
  2. Dark glass retuned: lighter blue-grey tint, stronger side lines, half the blur, lit rim — justified
  3. Composer types in white with a grey placeholder on smoked glass — justified
  4. Secondary muted ink brightened inside dark panes — justified
  5. Follow-up chips regained their top/bottom hairline — justified
  6. Hovering a glass button now visibly steps in the three solidified fallbacks — justified
  7. Jump-to-bottom pill moved 24px clear of the dock (was 8px) — justified
  8. Context-shelf chips stand on a short fade to page colour — justified
  9. Strip edge cues are blur wedges on glass; colour fades only where the pane solidifies — justified
  10. Reduced motion: a pending approval holds a steady glow instead of none — justified

List capped at 10: the 11th item is the new capture-harness scenes (contrast, shelf/focus/approval assertions) — dev tooling, rides along, harm-free.

Evidence notes: the deleted dark-ring pin ("the glow alone did not separate a focused capsule from a resting one") is reversed with declared, measured replacement evidence (edge step; side line 143→213, interior 56→68), so lens 8 is satisfied. Grepped <Glass across website/src (12 call sites): the FollowUpBar scroll row is the only overflow container holding Glass chips, so the hairline fix has 0 unfixed siblings. composer-halo has 0 remaining consumers after deletion. The --glass-text/--glass-placeholder/--glass-muted tokens follow the theming contract's documented polarity-fixed mechanism, updated in the same commit.

[FIRST-PRINCIPLES-REVIEWED] 7fdc2b9

Description read

[DESCRIPTION-READ] bce0e4d243142054fe1b5f6dfced7980e689286cca9451af2bba872c70cf9454

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 Sep 28, 2026 •

Copy link
Copy Markdown
Contributor

Opus 5 Review — ✅ no blocking findings

Reviewed 7fdc2b9d03ea279bde3310f179140603ea356b0d — this comment is updated in place on each push.

Review details

Dark-polarity composer focus indicator drops below 3:1 with nothing else in its place.

FINDING — website/src/components/ChatInput.tsx:4274 — the wrapper's focus cue is now unconditional border-transparent and composer-halo (whose dark rule painted 0 0 0 1px color-mix(in srgb, var(--accent) 55%, transparent), a full-perimeter ring measuring 3.62:1 against --bg:#12141a) is deleted, so with the textarea still outline-hidden the only remaining cue on the default dark theme is .glass-shadow:focus-within's whole-pane tint step (rgba(127,134,147,0.30)→0.42 ≈ 1.24:1), two 1px side lines (--glass-edge 0.35→0.70 white ≈ 2.46:1) and an outer shadow the file itself records as "no delta" — a keyboard user tabbing into the composer gets no indicator meeting WCAG 2.4.11's 3:1 → Fix: keep a real indicator on the focused control in dark polarity — restore focus-within:border-accent/50 on the wrapper, or widen the focused edge to a ≥2px perimeter whose focused/unfocused pixels differ by ≥3:1.

[OPUS-REVIEWED] 7fdc2b9

Verdict parsed from the review's SHA-scoped output markers for commit 7fdc2b9d03ea279bde3310f179140603ea356b0d.

False positive or not applicable? A repository writer can comment:
/ai-review override fable 7fdc2b9d03ea279bde3310f179140603ea356b0d: <one-sentence reason>

@github-actions

github-actions Bot commented Sep 28, 2026 •

Copy link
Copy Markdown
Contributor

UX Review (Fable 5) — 🔴 BLOCK

UX-level review of 7fdc2b9d03ea279bde3310f179140603ea356b0d — updated in place on each push. A BLOCK verdict blocks PR readiness; PASS/CONCERNS are advisory.

UX-Verdict: BLOCK

cannot evaluate: missing shots of the repainted overflow edge cues (file strip / control row) and the reduced-motion steady approval ring.

Watch

  • The frost 3 retune lets transcript text through hard enough to read as a draft: the blind reader, on the PR's own shot-14/shot-18, "for a second I thought the blurred lines were text I had typed" and "I couldn't tell whether the box was empty or had something in it." Every scrolled dark-mode conversation, glance-level misread, every time — a human should weigh a step more tint/frost behind the input line against the reference match.

Evidence gaps

  • preview-strip-cue-* / control-row-cue-* changed from an opaque from-bg-elevated fade to a 3px blur wedge, but no screenshot shows any strip overflowing — whether the wedge still reads as "content continues" is unverifiable; close with a dark-mode shot of an overflowing file strip and control row (plus one solidified-fallback frame).
  • @media (prefers-reduced-motion:reduce){.approval-glow{--glow-strength:.6}} introduces a newly visible steady ring; no screenshot shows an approval pending under reduced motion — close with one emulated frame.

[UX-REVIEWED] 7fdc2b9

Description read

[DESCRIPTION-READ] e7ef8846308b4dc23b21f49b22449fabc30af82f80220769e7ebcf2b90f8815f

That sha256 names everything this verdict read: the PR description, plus the 22 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.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=1b3fc4cbdee4 textarea has no focus indicator while a tool approval is pending: fixed

Fixed in 54e5538: the dock is glass-shadow approval-glow while a decision is pending. .approval-glow keeps its animated warn shadow in a --approval-shadow custom property and a combined .glass-shadow.approval-glow (rest and :focus-within) rule hands that value to the shadow slot, so the glow still wins the shadow while the .glass-shadow:focus-within tint + edge step (and the three fallback blocks' accent outline) now apply on that branch too. ChatInput.liquidGlass.test.tsx pins the stacked classes; the capture harness's approval scene focuses the textarea with the bar up and asserts tint #1e1e2266 -> #54545cb8 / edge #ffffff24 -> #ffffff8c (dark) and #eeeef373 -> #ffffffeb / #0000003d -> #0009 (light) while the shadow stays the glow. Frames in the PR body.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=d9ed244f0565 ChatScrollChrome layout contract still says 40px: fixed

Fixed in 54e5538: the contract now reads 56px above it (\-top-14`), and the ChatPage comment names -top-14`.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=747fb44e07b6 theming-contract.md and the harness header still cite the composer halo: fixed

Fixed in 54e5538: both sentences now name the neutral glass-shadow rest shadow.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=8247d5c4aacc floor assertion on one non-default theme pack for the focus step: accepted

Noted, not in this round: the harness already fails when tint or edge does not change on focus in the two default polarities; a pack loop needs the settings stub to install a pack, which is a harness feature of its own. Tracked as the next harness step; the tokens are documented as focus-bearing in theming-contract.md so a pack author sees the constraint.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=78298e2bf67a Item 1 — neutral focus reverses the 'accent glow is the composer's alone' pin: declined

The reversal is the maintainer's explicit product decision, made after seeing the shipped #14568 build: the composer's focus is the glass getting clearer, with no theme colour, glow or ring on any glass pane (owner direction, 2026-09-28, recorded in the PR body's Goal). The pins this diff deletes recorded the earlier state, not a user report; the harm removed is the composer being the one pane that lit up in the theme colour while every other glass surface did not. The maintainer confirms the reversal by authoring this PR.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=24b70625e169 light-theme focus cue barely reads after the accent border / glow removal: accepted

Measured, kept: on 54e5538 the light focus step is tint #eeeef373 -> #ffffffeb (interior 247 -> 255), side edge #0000003d -> #0009 (24% -> 60% black) and shadow .06 -> .14; dark is 27 -> 68 plus edge 14% -> 55% white. A theme-coloured border or glow is ruled out by the maintainer's decision (the glass is the cue). The caret and the edge step are the light cue; if a live check with a human still reads as zero the next knob is --glass-edge-focus on light, a one-token change.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=99c9d75317c6 prefers-contrast: more hover fill is the faintest 8% mix: fixed

Fixed in 54e5538: the prefers-contrast: more block uses a 16% --text mix for .liquid-glass.glass-hover:hover; the other two fallback blocks keep 8%. ChatInput.liquidGlass.test.tsx pins the per-block mix.

@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 Sep 28, 2026
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=26b6c244ee73 harness guard message names the deleted rule, not the re-added one: fixed

Fixed in 9661330: the guard now reads "the accent focus border is back -- the composer's focus cue must be the glass, not a themed border", so a reintroduced focus-within:border-accent/50 points the reader at the rule that came back.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=8a6b60d55811 Item 1 — neutral focus reverses the 'accent glow is the composer's alone' pin: declined

Same premise as the previous round (span 78298e2bf67a): the reversal is the maintainer's explicit product decision after seeing the shipped #14568 build -- focus on any glass pane is the material stepping clearer, never a theme colour, glow or ring. The deleted pins recorded the earlier state, not a user report, and the maintainer confirms the reversal by authoring this PR; no linked report is needed for an owner-directed design change.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=e8f3ec7f6b80 light-theme focus cue is near-threshold: accepted

Measured and kept, as in the previous round (span 24b70625e169): light focus is interior 247 -> 255, side edge 24% -> 60% black, shadow .06 -> .14; the caret is the third cue. A theme-coloured border or glow is ruled out by the maintainer's decision. The next knob, if a human read still calls it a guess, is the light --glass-edge-focus token alone. The two evidence gaps that carried the BLOCK are closed on 9661330: the harness's new contrast scene emulates prefers-contrast: more and the PR body now shows the hovered chip and the focused composer under that fallback in both polarities.

@github-actions github-actions Bot added readiness: checking Automated validation is still running readiness: passed Eligible automated validation passed for the current revision and removed readiness: action required A blocking check or review needs attention readiness: checking Automated validation is still running labels Sep 28, 2026
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=26ba1dddb132 approval-glow shadow slot paints nothing under prefers-reduced-motion: fixed

Fixed in c35e7ce: @media (prefers-reduced-motion: reduce) { .approval-glow { animation: none; --glow-strength: .6 } } holds a steady mid-pulse ring, so --approval-shadow is never the transparent 0% keyframe value and the dock keeps both its glow and a shadow. Pinned in ChatInput.liquidGlass.test.tsx.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=d5541e3ef7af retained comment describes the deleted composer-halo focus behaviour: fixed

Fixed in c35e7ce: the motion.div comment now says the element paints no shadow or focus cue of its own, that both belong to the Glass dock pane wrapping it (.glass-shadow, tint + edge step, no theme colour) and that a pending approval swaps the shadow slot for approval-glow.

@github-actions github-actions Bot added readiness: checking Automated validation is still running and removed readiness: passed Eligible automated validation passed for the current revision labels Sep 28, 2026
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=7617c9284cd8 bevel retune (BEVEL_REACH / BEVEL_GLOW) not named by the description: fixed

The PR description now names the bevel retune in Goal / Not-a-goal and the dark-material table: BEVEL_GLOW = 0.48, BEVEL_REACH = 2.5 is the reference's inner rim glow (+25-35 over ~5pt), part of the maintainer's chosen material. It paints in light mode too, where white on a near-white pane is invisible: every light edge profile measured within 1/255 of the previous head (top edge inward 251 250 250 250 249 249 249 248 198 251 before and after). Description amended before bbbde82 was pushed.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=bd606188eaae retune applies to the Settings capsule with no evidence: fixed

Four Settings-capsule frames (rest + focused, dark + light) from this build are in the description; the capsule has no shelf under it, so the shelf fade is the composer's alone. The in-pane control-row scrim this finding names is gone in bbbde82: the fade now sits under the context shelf, outside the pane.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=2831d833de9c Item 4 -- bevel retune undeclared and oversized: declined

Not reverted: the inner glow is part of the material the maintainer picked against the reference (dark rim measured 66 -> 113 inward, the reference's +25-35). The PR description now names the bevel retune in Goal / Not-a-goal and the dark-material table: BEVEL_GLOW = 0.48, BEVEL_REACH = 2.5 is the reference's inner rim glow (+25-35 over ~5pt), part of the maintainer's chosen material. It paints in light mode too, where white on a near-white pane is invisible: every light edge profile measured within 1/255 of the previous head (top edge inward 251 250 250 250 249 249 249 248 198 251 before and after). Goal / Not-a-goal amended by the maintainer's account.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=0f2b6f49930e placeholder legibility on the clearer glass: declined

Frost 3 and the #9c9c9c placeholder are the maintainer's material decisions after measuring against the reference (a 6px blur smeared a white bar's edge over 14px where the reference smears ~5). The row-confined scrim this finding refers to is removed in bbbde82; what stands on a fade now is the context shelf below the pane, where chips read against scrolling transcript. Inside the pane the typed ink is pure white; the placeholder is the one deliberately quiet ink.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=7deb4748419e neutral focus cue may not register as a state to a cold reader: accepted

Measured, not changed: dark interior 74 -> 108, dark side line 143 -> 213 on focus; light 247 -> 255 with the side line 24% -> 60%. The user causes the transition, as the finding notes. A theme-coloured ring on any glass focus is excluded by the maintainer's product decision.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=f1dca20ac6c3 light-mode panes inherit the universal bevel retune: declined

The glow layer is universal by design (one recipe), and on a light pane it is invisible: every light edge profile is within 1/255 of the previous head, so no light-mode pane widened or brightened. Restoring 0.2 * light would remove the reference's rim glow from the dark pane, which is the material the maintainer chose. Declared in the description.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=26ba1dddb132 muted ink loses half its contrast on the retuned dark tint; scrim not neutralised in fallbacks: fixed

Fixed in bbbde82: .liquid-glass { --muted: var(--glass-muted) } with --glass-muted: #b4b4bc on smoked glass (theme --muted on a light pane), so every text-muted inside a pane brightens with the glass -- follow-up chips measured 129 -> 154. The .glass-toolbar scrim is removed entirely; the shelf fade that replaces it uses var(--bg), the page colour, which needs no fallback override.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=d5541e3ef7af control-row edge cues fade from bg-elevated, a colour the surface no longer has: fixed

Fixed in bbbde82: both the control row's and the file-preview strip's cues are backdrop-blur wedges masked toward the strip (backdrop-blur-[3px]), which soften the clipped chip instead of painting an opaque colour on the glass. Same elements, same useScrollEdges gate, same tests; the comment that called from-bg-elevated a match is rewritten.

@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 Sep 29, 2026
…allback

Three follow-ups on the glass dock that shipped in #14568.

The composer's focus cue is the glass itself, no theme color. The dock
wore `composer-halo` (an accent glow, plus a 1px accent ring in dark
themes) and the input wrapper turned its border `accent/50` on focus;
every other glass pane wore the neutral `glass-shadow`. Now the dock
wears `glass-shadow` too and the wrapper's border stays transparent:
focus is the material getting a step clearer (`--glass-tint-focus` and
`--glass-edge-focus`), the same cue as the Settings capsule and the
cards. `composer-halo` and its memory-mode forms are gone.
While a tool approval is pending the dock stacks `approval-glow` on
`glass-shadow`: the glow takes the shadow slot, the focus step stays.

The follow-up chips lost their top and bottom hairline: the chip row is
`overflow-x-auto`, which also clips vertically, and the 0.5px line is
drawn just outside the box. The row gets `py-px -my-px` so the line
paints at the same row height.

Hover feedback on glass buttons survives the three solidifying
fallbacks: `.glass-hover` steps `--glass-tint`, which only the hidden
frost layer reads, so each fallback block moves the step onto the solid
fill (the contrast block at twice the mix). The jump-to-bottom pill floats 24px clear of the dock instead of
8px, where glass on glass read as one pane bleeding into the other.
@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=4fd463b10146 blur-wedge cues paint nothing where the pane solidifies: fixed

Fixed in 7fdc2b9: the cue elements carry glass-strip-cue-l / -r, and all three solidifying blocks (@supports not backdrop-filter, prefers-reduced-transparency, prefers-contrast: more) reset the blur and mask and paint a fade from --bg-elevated, the pane's own colour there. Pinned per block in ChatInput.liquidGlass.test.tsx.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=3ca7f4932175 Settings-capsule rest vs focused frames do not land at cold read: fixed

Evidence reworked in 7fdc2b9: each polarity is now ONE frame with REST and FOCUSED side by side and labelled, with the measured deltas in the caption (dark interior 57 -> 69, side line 14% -> 55%; light 247 -> 255, side line 24% -> 60%, shadow .06 -> .14, edge under the capsule 249 -> 239). The cue itself is unchanged by design: on glass, focus is the material stepping clearer plus darker/brighter side lines and a deeper shadow, never a theme colour or a ring -- the maintainer's decision, stated in the description; the caret is the third cue for a keyboard user.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=26ba1dddb132 shelf fade escapes the dock gutter / placeholder contrast: fixed

Fixed in 7fdc2b9 (the -100vw insets): the shelf is no longer positioned, so .glass-shelf::before takes the dock's relative z-10 dock-inert wrapper as its containing block and spans exactly the dock root, which already stops short of the scrollbar column; ChatInput hands it the shelf height as --glass-shelf-h. The placeholder point is now moot in the direction it argued: the dark tint was re-solved over the app's page (rgba(127,134,147,.30), composited #383742 at rest, the reference's #37393D) after the maintainer measured the previous focused composite at #6A6B79; #9c9c9c on #383742 is about 4.1:1, up from 3.39:1. #9c9c9c itself is the maintainer's spec for the placeholder and stays.

@CrysisDeu

Copy link
Copy Markdown
Contributor Author
  • span=d5541e3ef7af overflow cues render no pixels without backdrop-filter: fixed

Fixed in 7fdc2b9: same fix as the Design item -- glass-strip-cue-l/-r get a --bg-elevated fade (blur and mask reset) in all three fallback blocks, so the cue is visible exactly where the pane is opaque and the wedge would paint nothing.

@CrysisDeu CrysisDeu closed this Sep 29, 2026
@github-actions github-actions Bot removed the readiness: action required A blocking check or review needs attention label Sep 29, 2026
@bolichen97
bolichen97 deleted the fix/glass-followups branch November 6, 2026 07:17
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant