Mobile UX polish, finish v2 SEO/OG, smooth Material Source curve

Mobile sliders
- Multi-row vertical layout (label | slider | info-row) that mirrors desktop;
  no horizontal-overlap concerns.
- Slider track centered, fixed-but-responsive width (min(60vw, 220px)) so
  every row has the same length and slider edges align across ingredients.
- Names left-aligned (consistent with the 'min' bound just below).
- Smaller 16px thumb on mobile; --thumb-half CSS var keeps the JS
  slider-preview-marker math in sync with the rendered thumb size.
- Vertical alignment of name and value-input via 'align-items: center'
  on the label flex container.
- Mobile font hierarchy bumped for legibility on touch devices and older
  eyes: label 0.85rem, panel-body insight text 0.95rem, h2 0.95rem.
  Hide W/B readout on mobile to keep GWP+Cost on one line.
- Replaced ingredient-name 'border-bottom: 1px dashed' with
  'text-decoration: underline dashed' to save a few pixels of vertical
  height per row (no descender expansion).

Click-to-edit composition values
- Type a number + Enter, blur, or unit-toggle commit; Escape reverts;
  out-of-range clamps; non-numeric reverts. Material Source value is a
  <span> (non-editable). Pre-toggle unit context preserved on commit.
- 'Temp' renamed to 'Temperature' in displayed labels (column name
  unchanged so the model layer is untouched).

Material Source toggle
- Curve-level 350ms ease-in-out blend between pre/post-toggle posteriors
  in drawStrengthCurve (avoids feeding fractional categorical input to
  the GP). State exposed via window.__test for deterministic testing.
- scheduleInsightUpdate() on toggle so the Mix Insight no longer goes
  stale when the new composition is outside the training set.

Finish v2: SEO + social
- robots.txt and sitemap.xml.
- README link to the deployed explorer.
- (Open Graph / Twitter Card / JSON-LD already in head.)

Bug fixes
- Filter min/max placeholders no longer clipped by Chromium spinner UI
  ('appearance: textfield' + zeroed -webkit spin pseudo-elements).
- Slider preview marker now offsets by slider.offsetLeft so it lands on
  the visible track on mobile (and stays a no-op on desktop).
- displayPreviewComp now stays in sync with currentComposition on
  Material Source toggle (no more dashed-preview ghosting).

Refactor / cleanup
- logSpacedTimes() and computeStds() helpers consolidate three duplicated
  copies of the time-grid generator and the GP-noise std computation.
- Removed pre-existing dead declarations in updateMixInsight and
  getParetoTag.
- Renamed unused @parameterized.expand 'name' params to '_name' in
  test/test_utils.py so the linter no longer flags them.

Tests (10 new specs + extended sliders)
- preview-curve: displayPreviewComp sync, mix-insight refresh on toggle,
  curve-transition state hook (deterministic, no screenshot timing race).
- font-uniformity: insight bodies share computed font-size.
- og-meta + seo: required tags present, og-image fetch in budget,
  robots.txt and sitemap.xml reachable.
- mobile-slider-layout: multi-row no overlap, names left-aligned,
  uniform slider widths, centered+narrower slider, vertical centerline,
  text-decoration affordance, font hierarchy, value glyph aligns with
  max bound, tap target via offsetHeight, preview marker on track.
- mobile-value-fit: no panel overflow at default, fractional, imperial,
  imperial-max, and 320px viewports.
- scatter-filter: placeholders fit fully (no spinner clip).
- readouts-strip: desktop has all three, mobile hides W/B and stays on
  one line.
- sliders.spec: extended with 7 click-to-edit cases.
- test/e2e/README.md invariants table updated.

Author: SebastianAment

README.md

@@ -1,6 +1,8 @@
 # BOxCrete: Bayesian Optimization for Sustainable Concrete Mix Design
 
-Concrete, the second most widely used material in the world, accounts for **6–8% of global anthropogenic CO₂ emissions**, largely due to Portland cement production (~0.8 tons CO₂ per ton of cement). Partial replacement with Supplementary Cementitious Materials (SCMs) such as fly ash, slag, and natural pozzolan reduces embodied carbon and often improves durability, but high SCM usage makes compressive strength a highly nonlinear function of multiple interacting mix parameters, rendering traditional design empirical and trial-and-error driven. To systematically navigate this complex composition space, data-driven frameworks are needed.
+> 🌐 **[Try the interactive explorer →](https://facebookresearch.github.io/SustainableConcrete/)** — predict concrete strength from mix composition in your browser, no installation needed.
+
+Concrete, the second most widely used material in the world, accounts for **6–8% of global anthropogenic CO₂ emissions**, largely due to Portland cement production (~0.8 tons CO₂ per ton of cement).
 Here, we introduce BOxCrete, an open-source Bayesian optimization framework for probabilistic strength curve prediction and sustainable mix design.
 We invite researchers and practitioners from all disciplines including AI, machine learning, computer science, materials science, and civil engineering
 to collaborate on discovering more sustainable concrete formulations that are applicable

docs/index.html

@@ -5,6 +5,71 @@
   <meta name="viewport" content="width=device-width, initial-scale=1.0, viewport-fit=cover">
   <title>BOxCrete – Interactive Concrete Strength Explorer</title>
   <meta name="theme-color" content="#e8edf3">
+
+  <!-- SEO -->
+  <meta name="description" content="Interactive Bayesian Optimization explorer for sustainable concrete: predict 28-day compressive strength, GWP, and cost from mix composition in your browser.">
+  <meta name="author" content="Sebastian Ament; BOxCrete authors">
+  <link rel="canonical" href="https://facebookresearch.github.io/SustainableConcrete/">
+
+  <!-- Open Graph (Facebook, LinkedIn, Slack, iMessage, Discord) -->
+  <meta property="og:type" content="website">
+  <meta property="og:site_name" content="BOxCrete">
+  <meta property="og:title" content="BOxCrete – Interactive Concrete Strength Explorer">
+  <meta property="og:description" content="Predict concrete compressive strength, GWP, and cost from mix composition. A Bayesian Optimization model running entirely in your browser.">
+  <meta property="og:url" content="https://facebookresearch.github.io/SustainableConcrete/">
+  <meta property="og:image" content="https://facebookresearch.github.io/SustainableConcrete/og-image.jpg">
+  <meta property="og:image:width" content="1200">
+  <meta property="og:image:height" content="630">
+  <meta property="og:image:alt" content="BOxCrete: a close-up photograph of fresh concrete, used as the social-share thumbnail for the interactive explorer.">
+
+  <!-- Twitter / X Card -->
+  <meta name="twitter:card" content="summary_large_image">
+  <meta name="twitter:title" content="BOxCrete – Interactive Concrete Strength Explorer">
+  <meta name="twitter:description" content="Predict concrete compressive strength, GWP, and cost from mix composition. A Bayesian Optimization model running entirely in your browser.">
+  <meta name="twitter:image" content="https://facebookresearch.github.io/SustainableConcrete/og-image.jpg">
+
+  <!-- iOS share sheet uses apple-touch-icon as a fallback thumbnail -->
+  <link rel="apple-touch-icon" href="og-image.jpg">
+
+  <!-- Schema.org structured data: WebApplication + scholarly citations -->
+  <script type="application/ld+json">
+  {
+    "@context": "https://schema.org",
+    "@type": "WebApplication",
+    "name": "BOxCrete",
+    "alternateName": "BOxCrete Interactive Concrete Strength Explorer",
+    "url": "https://facebookresearch.github.io/SustainableConcrete/",
+    "description": "Bayesian Optimization model for predicting concrete compressive strength, global warming potential, and cost from mix composition. Runs entirely in the browser.",
+    "applicationCategory": "EngineeringApplication",
+    "operatingSystem": "Any",
+    "browserRequirements": "Requires JavaScript and a modern browser",
+    "offers": {
+      "@type": "Offer",
+      "price": "0",
+      "priceCurrency": "USD"
+    },
+    "author": [
+      {"@type": "Person", "name": "Bayezid Baten"},
+      {"@type": "Person", "name": "M. Ayyan Iqbal"},
+      {"@type": "Person", "name": "Sebastian Ament"},
+      {"@type": "Person", "name": "Julius Kusuma"},
+      {"@type": "Person", "name": "Nishant Garg"}
+    ],
+    "citation": [
+      {
+        "@type": "ScholarlyArticle",
+        "name": "BOxCrete: A Bayesian Optimization Open-Source AI Model for Concrete Strength Forecasting and Mix Optimization",
+        "url": "https://arxiv.org/abs/2603.21525"
+      },
+      {
+        "@type": "ScholarlyArticle",
+        "name": "Sustainable Concrete via Bayesian Optimization",
+        "url": "https://arxiv.org/abs/2310.18288"
+      }
+    ]
+  }
+  </script>
+
   <link rel="icon" type="image/svg+xml" href="favicon.svg">
   <link rel="stylesheet" href="style.css">
   <link rel="preconnect" href="https://fonts.googleapis.com">
@@ -154,7 +219,7 @@ <h2>Predicted Strength Curve</h2>
         <div class="readouts" id="readouts">
           <div>GWP: <span class="value" id="gwp-value">–</span> <span class="unit" id="gwp-unit">kg CO₂/m³</span></div>
           <div>Cost: <span class="value" id="cost-value">–</span> <span id="cost-uncertainty"></span> <span class="unit" id="cost-unit">$/m³</span></div>
-          <div>W/B: <span class="value" id="wb-value">–</span></div>
+          <div class="wb-readout">W/B: <span class="value" id="wb-value">–</span></div>
           <!-- TODO: Add Slump readout once model/slump.json is available. -->
         </div>
       </div>

docs/og-image.jpg

@@ -0,0 +1,7 @@
+# BOxCrete interactive explorer — crawler hints.
+# All assets under docs/ are public; no auth or rate-limited surfaces here.
+# Sitemap path is informational: GitHub Pages doesn't auto-discover robots.txt
+# at project subpaths, so this file is mainly for Lighthouse and direct fetches.
+User-agent: *
+Allow: /
+Sitemap: https://facebookresearch.github.io/SustainableConcrete/sitemap.xml

docs/robots.txt

@@ -0,0 +1,9 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9">
+  <url>
+    <loc>https://facebookresearch.github.io/SustainableConcrete/</loc>
+    <lastmod>2026-05-07</lastmod>
+    <changefreq>weekly</changefreq>
+    <priority>1.0</priority>
+  </url>
+</urlset>

docs/sitemap.xml

@@ -427,23 +427,70 @@ h2 {
   box-shadow: 0 2px 8px rgba(0, 0, 0, 0.08), inset 0 0 0 1px var(--border);
 }
 
-/* Sliders panel */
-.sliders { grid-column: 1; overflow-y: auto; max-height: 80vh; position: relative; }
+/* Sliders panel
+   Note: no `max-height` / `overflow-y` here — the panel grows to fit all
+   ingredient rows naturally on desktop. The references panel in column 3 is
+   typically taller, so the composition panel still ends within the page flow. */
+.sliders { grid-column: 1; position: relative; }
 .sliders-header {
   display: flex;
   align-items: center;
   justify-content: space-between;
   margin-bottom: 0.25rem;
 }
 .sliders-header h2 { margin: 0; }
-.slider-group { margin-bottom: 0.75rem; }
+.slider-group { margin-bottom: 0.55rem; }
 .slider-group label {
   display: flex;
+  /* Vertically align name text with the value-input's text. Without
+     `align-items: center` the default `stretch` makes the name span and
+     the value-input fill the row's full height, but the name text sits
+     at the top of the line box while the value-input's text sits in its
+     own vertical middle — visually offset. Centering the children's
+     content axes pins them to the same baseline. */
+  align-items: center;
   justify-content: space-between;
   font-size: 0.8rem;
   margin-bottom: 2px;
 }
 .slider-group label span { font-weight: 600; }
+
+/* Click-to-edit composition value (regular sliders only — not Material Source).
+   Visual treatment mirrors `.toggle-btn`: rectilinear with 4px radius and
+   accent-colored focus border, matching the existing glass-morph language.
+   Width is 6.5ch to accommodate the worst-case formatted value across both
+   unit systems — e.g. water 700 kg/m³ in lb/yd³ = `1180.6` (six chars). */
+.slider-value {
+  font: inherit;
+  font-weight: 600;
+  text-align: end;
+  width: 6.5ch;
+  min-width: 6.5ch;
+  padding: 2px 4px;
+  border: 1px solid transparent;
+  border-radius: 4px;
+  background: transparent;
+  color: inherit;
+  cursor: text;
+  transition: background 0.15s, border-color 0.15s;
+  /* Hide native number-input spinners (defensive — input type is `text` here
+     but `inputmode=decimal` may map to spinner controls in some engines). */
+  appearance: textfield;
+  -moz-appearance: textfield;
+}
+.slider-value::-webkit-inner-spin-button,
+.slider-value::-webkit-outer-spin-button {
+  -webkit-appearance: none;
+  margin: 0;
+}
+.slider-value:hover {
+  background: rgba(var(--card-bg-rgb), 0.4);
+}
+.slider-value:focus {
+  outline: none;
+  border-color: var(--accent);
+  background: rgba(var(--card-bg-rgb), 0.6);
+}
 .slider-group input[type=range] {
   width: 100%;
   cursor: pointer;
@@ -481,21 +528,35 @@ h2 {
 /* Slider preview marker (shown on scatter hover) — hollow circle like a ghost thumb */
 .slider-group {
   position: relative;
-}
-/* Clickable ingredient names */
+  /* Half the rendered thumb width, exposed as a CSS variable so the JS
+     marker-positioning math (`showSliderPreview` in ui.mjs) can derive
+     the same value via `getComputedStyle(...).getPropertyValue('--thumb-half')`.
+     Mobile overrides this to 8 px because the thumb is shrunk to 16 px
+     in the @media block below; without that override, the preview marker
+     sat 1 px off-track on mobile. */
+  --thumb-half: 9px;
+}
+/* Clickable ingredient names.
+   Use `text-decoration` instead of `border-bottom` for the dashed
+   underline so the line sits at the font's natural baseline-offset
+   (just under the glyphs) rather than below the descender — saves a
+   few pixels of vertical space per row, especially noticeable on
+   mobile where every ingredient row is stacked. */
 .ingredient-name {
   cursor: pointer;
-  border-bottom: 1px dashed var(--text-secondary);
-  transition: color 0.15s, border-color 0.15s;
+  text-decoration: underline dashed var(--text-secondary);
+  text-decoration-thickness: 1px;
+  text-underline-offset: 4px;
+  transition: color 0.15s, text-decoration-color 0.15s;
 }
 .ingredient-name:hover {
   color: var(--accent);
-  border-color: var(--accent);
+  text-decoration-color: var(--accent);
 }
 .ingredient-name.active {
   color: var(--accent);
-  border-bottom-style: solid;
-  border-color: var(--accent);
+  text-decoration-style: solid;
+  text-decoration-color: var(--accent);
 }
 
 /* Ingredient insight panel */
@@ -735,6 +796,17 @@ h2 {
   background: rgba(var(--card-bg-rgb), 0.5);
   color: var(--text);
   transition: border-color 0.15s;
+  /* Hide native number spinner buttons. Chromium shows them by default,
+     which steals ~14 px on the right and clips the "min"/"max"
+     placeholders to "mi"/"ma". Mobile browsers don't render spinners,
+     which is why the bug only manifested on desktop. */
+  appearance: textfield;
+  -moz-appearance: textfield;
+}
+.filter-row input[type=number]::-webkit-inner-spin-button,
+.filter-row input[type=number]::-webkit-outer-spin-button {
+  -webkit-appearance: none;
+  margin: 0;
 }
 .filter-row input[type=number]:focus {
   border-color: var(--accent);
@@ -865,7 +937,7 @@ h2 {
   margin: 0.1rem 0;
 }
 .ref-desc {
-  font-size: 0.75rem;
+  font-size: 0.82rem;
   color: var(--text);
   line-height: 1.4;
   margin: 0.2rem 0 0;
@@ -1224,11 +1296,129 @@ a.ref-link {
     font-size: 0.78rem;
   }
 
-  /* Sliders content inside unified panel */
-  .mobile-sliders-view .slider-group { margin-bottom: 0.4rem; }
-  .mobile-sliders-view .slider-group label { font-size: 0.72rem; }
+  /* Sliders content inside unified panel.
+     Mobile uses a vertical multi-row layout:
+       row 1: <label> with name (LEFT) and value-input (RIGHT) — full panel
+              content width via the base flex-between rule.
+       row 2: <input type=range> CENTERED, narrower than the panel so the
+              row 1 / row 3 edge labels (name↔min on the left, value↔max on
+              the right) stack vertically without the slider getting in
+              their way.
+       row 3: <div class="info-row"> with min (LEFT) and max (RIGHT) —
+              also full panel content width.
+     Vertical spacing is tightened so the three rows feel like a single
+     compact block (small label margin-bottom, small slider-group
+     margin-bottom). */
+  .mobile-sliders-view .slider-group {
+    margin-bottom: 0.3rem;
+    /* `.slider-group` is `position: relative` for the slider preview marker
+       (set globally above). */
+  }
+  .mobile-sliders-view .slider-group label {
+    /* Inherits `display: flex; justify-content: space-between` from the base
+       rule — name on the left, value-input on the right. 0.85rem matches
+       the mobile h2 size and bumps legibility on touch devices; both the
+       name span and the value-input inherit this size (the input uses
+       `font: inherit`). The mobile-value-fit specs assert the resulting
+       6.5ch input still fits within the panel at 320 px. */
+    font-size: 0.85rem;
+    margin-bottom: 0;
+  }
+  .mobile-sliders-view .slider-group .slider-value {
+    /* Tap-target tuning: ensure ≥ 32px height for thumbs while keeping
+       the rectilinear 4px radius. */
+    min-height: 32px;
+    padding: 4px 6px;
+  }
+  .mobile-sliders-view .slider-group input[type=range] {
+    /* Constrain the slider track to a centered, fixed-but-responsive width
+       so it sits in the middle of the panel rather than spanning the full
+       width. `min(60vw, 220px)` is comfortable on phones from 320 px up:
+         320 px viewport → 60vw = 192 px (capped to 192 px)
+         412 px viewport → 60vw = 247 px (capped to 220 px)
+       `display: block; margin: 4px auto;` centers horizontally and adds a
+       small breathing margin from the label above and info-row below. */
+    width: min(60vw, 220px);
+    display: block;
+    margin: 4px auto;
+  }
+  /* Smaller thumb on mobile reduces the intrinsic thumb-half-width inset
+     (was 9 px for the 18 px desktop thumb). 16 px is still a comfortable
+     touch target combined with the row's vertical generosity. The CSS
+     variable below keeps the JS slider-preview-marker math in sync —
+     `showSliderPreview` reads `--thumb-half` from the slider-group. */
+  .mobile-sliders-view .slider-group {
+    --thumb-half: 8px;
+  }
+  .mobile-sliders-view .slider-group input[type=range]::-webkit-slider-thumb {
+    width: 16px;
+    height: 16px;
+  }
+  .mobile-sliders-view .slider-group input[type=range]::-moz-range-thumb {
+    width: 16px;
+    height: 16px;
+  }
+  .mobile-sliders-view .slider-group .info-row {
+    /* Right-align the max-bound text to the same pixel column as the
+       `.slider-value` input's rendered glyph end. The value input has
+       1 px border + 6 px right padding = 7 px inset; matching that here
+       lines up the displayed value with the printed maximum exactly.
+       Left padding is zeroed so the minimum bound aligns with the
+       left-aligned ingredient name above. */
+    padding: 0 7px 0 0;
+  }
+  /* Material Source row on mobile: hide the redundant value-span (the active
+     toggle button is self-labeling). The toggle row sits below the label as
+     a normal block-flow child, just like an `input[type=range]` would. */
+  .mobile-sliders-view .material-source-group label > span:last-child {
+    display: none;
+  }
 
-  h2 { font-size: 0.85rem; }
+  /* --- Mobile legibility tuning ---
+     Touch-device read distances and older-eye comfort favour ~16 px (= 1rem)
+     for body copy. Bump the panel-body text and section headers uniformly
+     so the hierarchy stays consistent. We deliberately leave layout-tight
+     components (`.controls`, `.filter-row`, `.site-header h1`) at their
+     existing sizes to avoid wrapping or clipping in chrome that's already
+     space-constrained on small screens. */
+  h2 { font-size: 0.95rem; }
+  .mix-insight-text,
+  .ingredient-insight-text,
+  .ref-desc {
+    /* Bumped from 0.82–0.85rem. Keeps the three panels typographically
+       unified — the `font-uniformity.spec.ts` invariant still holds. */
+    font-size: 0.95rem;
+    line-height: 1.5;
+  }
+  .mix-insight-title,
+  .ingredient-insight-title {
+    /* Uppercase pill labels: bump enough to be comfortably readable but
+       still smaller than the body so the hierarchy reads "header → body". */
+    font-size: 0.85rem;
+  }
+  .references-panel h3 {
+    font-size: 0.95rem;
+  }
+  .ref-authors {
+    /* Author byline directly under the ref title. Was 0.75rem; bump
+       slightly to keep up with the 0.95rem ref-desc above. */
+    font-size: 0.85rem;
+  }
+  .readouts {
+    /* Override the earlier `font-size: 0.78rem` block above with a more
+       legible 0.88rem for the GWP / Cost / W/B readout strip. */
+    font-size: 0.88rem;
+  }
+  /* Hide the W/B (water-to-binder ratio) readout on mobile. With the bumped
+     0.88rem font, GWP + Cost already fill the strip; adding W/B forces a
+     wrap that steals vertical space from the strength curve canvas above.
+     Power users on mobile still have the slider readouts to infer water/
+     binder; W/B remains visible on desktop where there's room. */
+  .wb-readout { display: none; }
+  .mobile-hint {
+    /* Italic instructional hint inside the unified mobile panel header. */
+    font-size: 0.8rem;
+  }
 }
 
 /* Landscape mobile — panels side by side */