style: consume the canonical cool design tokens from openadapt-web

.github/workflows/ci.yml

@@ -51,3 +51,28 @@ jobs:
 
       - name: Build canonical documentation in strict mode
         run: uv run mkdocs build --strict
+
+  design-tokens:
+    name: Vendored design tokens match openadapt-web
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
+
+      - uses: actions/setup-python@5fda3b95a4ea91299a34e894583c3862153e4b97 # v7.0.0
+        with:
+          python-version: "3.12"
+
+      # The online half of the palette guard. The offline half runs in
+      # `pytest tests/` above and needs no network: it pins the digest of each
+      # vendored copy and fails on a retired palette value in any shipped
+      # stylesheet, template, script, SVG, or recorded image measurement.
+      #
+      # This job is what catches the other direction: a token that changes in
+      # OpenAdaptAI/openadapt-web while this site keeps serving the old value.
+      # Failing here is the intended outcome of that drift, not a flake. The
+      # fix is `python scripts/vendor_design_tokens.py --write`, reviewed and
+      # committed. Nothing is installed, so it cannot fail on a lock refresh.
+      - name: Check the vendored tokens against openadapt-web@main
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: python scripts/vendor_design_tokens.py --check

docs/assets/screenshots/PROVENANCE.txt

@@ -2,9 +2,11 @@ Screenshot provenance
 =====================
 
 All screenshots below are rendered from OpenAdapt's own tooling in local /
-mock mode. None are mocked-up or hand-drawn. The Cloud captures carry the
-visible "LOCAL MOCK MODE" banner because they were taken against the app
-running with OPENADAPT_CLOUD_MODE=mock (seeded demo data, no real tenant).
+mock mode. None are mocked-up or hand-drawn. Each capture that shows product
+chrome is a byte-identical copy of a file in OpenAdaptAI/openadapt-web, which
+owns the published visual record for every OpenAdapt surface;
+docs/assets/visual-palette.json records the measured dominant colours of every
+image on this site.
 
 program-graph-cloud.png
   The compiled-program visualizer in the OpenAdapt Cloud workspace
@@ -34,8 +36,11 @@ settings-governance.png
 
 dashboard-overview.png
   The /dashboard overview of the OpenAdapt Cloud workspace at exact commit
-  f2300ee69b555d1676e69348999f910c1cc8cd8c. Cypress Electron captured the
-  focused 2560x1440 viewport in OPENADAPT_CLOUD_MODE=mock. It shows grouped
-  navigation and the visible LOCAL MOCK MODE banner. The fixture is synthetic
-  OpenEMR patient registration. SHA-256:
-  ea3b8065cc139f94202ce2422443566ad4fb74f159d0af255e3802f0b7e74a01.
+  021a9caf99a6832bff9f32e14b1b34d42ec2e290, captured at 2560x1440 with
+  Playwright Chromium through openadapt-cloud scripts/capture-public-previews.mjs
+  against the local Cloud mock UI. The runtime-mode banner and the account chip
+  are hidden during capture because both report the environment the capture ran
+  in, not the product. The fixture is synthetic OpenEMR patient registration.
+  Byte-identical source: OpenAdaptAI/openadapt-web,
+  public/product-preview/dashboard-workflows.png. SHA-256:
+  f9f894a60e44b266f16b8d138d8747e1a1fb639bf92370db6adbf4d4aab5bde4.

docs/assets/ui/PROVENANCE.txt

@@ -1,46 +1,69 @@
 Current UI visual provenance
 =============================
 
+Every file below is a byte-identical copy of a file in OpenAdaptAI/openadapt-web,
+which owns the published visual record for every OpenAdapt surface. That
+repository's provenance manifests name the upstream Cloud or Desktop commit
+each capture came from. docs/assets/visual-palette.json records the measured
+dominant colours of every image on this site, and tests/test_design_tokens.py
+fails if a copy stops matching its record.
+
 desktop-qualification-cockpit.png
-  SHA-256: 14cf86ae66ce91c0255eba74ac2c4fa3b8db188a01eb9c5874e90afe37300f9d
-  1600 x 1100. Unmodified source: OpenAdaptAI/openadapt-web, Web PR #332
-  final head 0e57c9ce7c7bd92e37af7e4ca4bf54833d949fe7,
-  public/desktop-preview/cockpit/10_dashboard_workflows.png. The documentation
-  copy is byte-identical to that source file. It contains synthetic data.
+  1600 x 1100. SHA-256:
+  83b854f7375bea220902bb6fca6d1ca40e31123691c0da4a7f4a435de37da65a.
+  Source: openadapt-web public/desktop-preview/cockpit/10_dashboard_workflows.png.
+  Headless Chromium rendered the Desktop React cockpit through
+  scripts/capture-cockpit-previews.mjs in openadapt-desktop at PR #139 head
+  d8c08567e42ae45e9687211d757ae3536c97e460, answering the engine IPC surface
+  from that repository's committed synthetic fixture table. No Tauri window,
+  engine sidecar, account, Cloud session, recording, physical input, customer
+  data, or live execution was used. It shows a synthetic local workflow library
+  with one halted workflow and one verified workflow.
 
 mobile-decision-request.jpg
-  SHA-256: 6249a4288a807a6f33ec05c3df6f1f52ef169a3ee25edf99feebbd89de37464e
-  314 x 784. Phone-shell crop from an OpenAdaptAI/openadapt-cloud production
-  build at component commit f250ef34962e9237e49a3d04b37c6fead3b565f9.
-  The supported production-browser capture used device scale factor 1. The
-  files are not post-upscaled.
-  The retained capture set is in commit
-  7e462d489b4b9fc249cefb5b70c9a6009c737a0c. The route is
-  /demo/attention in Cloud PR #276. The data is synthetic OpenEMR application
-  data.
+  314 x 784. SHA-256:
+  c50b46b5febaafafcef2a58e6ec2d9a6c5221a4386bdeacaa3a2f3836886b9b6.
+  Source: openadapt-web public/attended-decision/identity-request.jpg. Phone
+  shell at route /demo/attention, captured with Playwright Chromium through
+  openadapt-cloud scripts/capture-public-previews.mjs against the local Cloud
+  mock UI at commit 021a9caf99a6832bff9f32e14b1b34d42ec2e290. The runtime-mode
+  banner is hidden during capture because it reports the environment the
+  capture ran in, not the product. It shows an open identity request before an
+  operator answer. The data is synthetic OpenEMR application data.
 
 mobile-decision-pending.jpg
-  SHA-256: fa616b67e072ed82291b7990e8623bc5d263f64547c5bbd269daf121d77791a5
-  314 x 784. From the same source and capture sequence as the request. It shows
-  the decision_pending state. The signed answer was accepted once. No runner
-  receipt exists, and the workflow remains paused.
+  314 x 784. SHA-256:
+  a4d4af28c3627de9356ef1ee09d78137e3a2d530edc3cfa68c0d242299288b46.
+  Source: openadapt-web public/attended-decision/decision-pending.jpg. From the
+  same capture sequence as the request. It shows the decision_pending state:
+  the signed answer was accepted once, no runner receipt exists, and the
+  workflow remains paused.
 
 mobile-decision-result.jpg
-  SHA-256: 3096f8b5064a6271894c894e98c15919668a14ef0bd3aa0c59319710d75d69a6
-  314 x 784. From the same source and capture sequence as the request. It is the
-  completed runner result with the verified_and_resumed reason code. The data
-  is synthetic OpenEMR application data.
+  314 x 784. SHA-256:
+  e64f5de7ba84457baa5280cfac0cdcfea12f8c38009bd8dfb3443b794ed01042.
+  Source: openadapt-web public/attended-decision/identity-verified.jpg. From the
+  same capture sequence as the request. It is the completed runner result with
+  the verified_and_resumed reason code. The data is synthetic OpenEMR
+  application data.
 
 business-decision-qualification-capture.png
-  SHA-256: 1876efdba92d3995cffcdca126d32d9e2493f7e4ffaf46b131f8580c09994862
-  860 x 3542. Byte-identical source: OpenAdaptAI/openadapt-cloud,
-  Cloud PR #281 merge commit 176595db43c3bfbf0a0bd4a1100e035e922396c0,
-  public/demo/qualification-judgment/judgment-capture-request.png. It is a
-  synthetic qualification capture. It is not a live execution request.
+  860 x 3240. SHA-256:
+  013d3dedd76fb02c01a05b06ebfb64da4d51d6ae442f9cd5e2eac0b66126d7e9.
+  Source: openadapt-web public/qualification-judgment/judgment-capture-human-node.png.
+  Route /demo/qualification-judgment in openadapt-cloud, captured with
+  Playwright Chromium through that repository's
+  scripts/capture-public-previews.mjs at commit
+  021a9caf99a6832bff9f32e14b1b34d42ec2e290. It is a synthetic qualification
+  capture. It is not a live execution request.
 
 business-decision-runtime-request.png
-  SHA-256: cf9ab3f36f6869ec80758ae70ff6aa2466d73615997de5d5eed351d2cdca1a65
-  520 x 880. Byte-identical source: OpenAdaptAI/openadapt-cloud,
-  Cloud PR #278 merge commit 167b2f0843229c5938f3d9e97de46f68b66e3476,
-  public/demo/judgment/business-decision-request.png. It is a synthetic
-  mobile live-run decision request. It does not report an execution result.
+  430 x 859. SHA-256:
+  829347392941e65d8ec2e895cd503a9b600260ed638ddb662e15fcc423b797dd.
+  Source: openadapt-web public/business-decision/business-decision-request.png.
+  Route /demo/judgment?view=request in openadapt-cloud at commit
+  baf08a94e38618c9de4b47cfa85c95750ed51b2a, captured with Playwright Chromium
+  through that repository's scripts/capture-public-previews.mjs. The shot is
+  clipped to the phone bezel's border box and carries an alpha channel. It is a
+  synthetic mobile live-run decision request. It does not report an execution
+  result.

docs/assets/visual-palette.json

@@ -0,0 +1,221 @@
+{
+  "$note": "Measured palette of every image docs.openadapt.ai ships. The docs site kept a retired warm palette after the rest of the product moved to the canonical cool one, and nobody could see it: a screenshot's ground reads as 'off-white' either way, and a search for token names finds nothing inside a PNG. So this file records a measurement instead of an opinion. scripts/measure_asset_palette.py writes it; tests/test_design_tokens.py enforces it offline.",
+  "$cool_rule": "The DOMINANT colour of every image must have blue >= red. The dominant colour is the page ground, and the ground is what makes a capture read warm or cool: the canonical grounds are cool by construction (#FFFFFF, #F5F7FA, #EDF1F6, #0B1220 all satisfy it) and every retired warm ground fails it (#f2f1ec, #fbfaf6, #fdfcf9, #fdfcf8, #f4f3ef, #f4f3ed all have red > blue). The rule needs no allowlist and no warm ground can satisfy it. It deliberately stops at the dominant colour: a warm tint below it is usually --accent-halt doing its job, which is a canonical token, not a retired palette.",
+  "$provenance": "Each entry names the file in OpenAdaptAI/openadapt-web this copy is byte-identical to. openadapt-web owns the published visual record; that repository's own provenance manifests name the upstream Cloud or Desktop commit each capture came from. The prose descriptions stay in the PROVENANCE.txt beside the images.",
+  "assets": {
+    "favicon.ico": {
+      "source_repository": null,
+      "source_path": null,
+      "note": "Neutral greyscale mark. It carries no brand hue in either palette.",
+      "sha256": "56f16c3d7104a06436643c659cfe360b869633ebe4e1f1d49ac1a90dc0a0f03b",
+      "width": 48,
+      "height": 48,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 68.01
+        },
+        {
+          "hex": "#595959",
+          "share_percent": 4.86
+        },
+        {
+          "hex": "#e9e9e9",
+          "share_percent": 3.73
+        }
+      ]
+    },
+    "screenshots/dashboard-overview.png": {
+      "source_repository": "OpenAdaptAI/openadapt-web",
+      "source_path": "public/product-preview/dashboard-workflows.png",
+      "sha256": "f9f894a60e44b266f16b8d138d8747e1a1fb639bf92370db6adbf4d4aab5bde4",
+      "width": 2560,
+      "height": 1440,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 46.48
+        },
+        {
+          "hex": "#f5f7fa",
+          "share_percent": 30.14
+        },
+        {
+          "hex": "#fafafa",
+          "share_percent": 2.51
+        }
+      ]
+    },
+    "screenshots/for-each-program-graph.png": {
+      "source_repository": null,
+      "source_path": null,
+      "note": "Output of `openadapt flow visualize`, not product chrome. It was already neutral and is unchanged by the palette move.",
+      "sha256": "0a6e524c4066a07c14db8732bd416d1e940e5b525d46c749ee548f1a6b94b244",
+      "width": 1280,
+      "height": 455,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 78.27
+        },
+        {
+          "hex": "#f6f8fa",
+          "share_percent": 10.69
+        },
+        {
+          "hex": "#f7f9fb",
+          "share_percent": 0.76
+        }
+      ]
+    },
+    "showcase/demo.gif": {
+      "source_repository": null,
+      "source_path": null,
+      "note": "Terminal recording of a replay run on a dark ground, not product chrome.",
+      "sha256": "cec6464e0e7acabc5c551e047bc906488bcc0be5e8c62a2cc082a21a36e16f7f",
+      "width": 1272,
+      "height": 535,
+      "dominant_colors": [
+        {
+          "hex": "#0d0b1e",
+          "share_percent": 94.68
+        },
+        {
+          "hex": "#0c0a1d",
+          "share_percent": 0.56
+        },
+        {
+          "hex": "#0b091c",
+          "share_percent": 0.38
+        }
+      ]
+    },
+    "ui/business-decision-qualification-capture.png": {
+      "source_repository": "OpenAdaptAI/openadapt-web",
+      "source_path": "public/qualification-judgment/judgment-capture-human-node.png",
+      "sha256": "013d3dedd76fb02c01a05b06ebfb64da4d51d6ae442f9cd5e2eac0b66126d7e9",
+      "width": 860,
+      "height": 3240,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 21.37
+        },
+        {
+          "hex": "#f5f7fa",
+          "share_percent": 9.64
+        },
+        {
+          "hex": "#0b1220",
+          "share_percent": 2.33
+        }
+      ]
+    },
+    "ui/business-decision-runtime-request.png": {
+      "source_repository": "OpenAdaptAI/openadapt-web",
+      "source_path": "public/business-decision/business-decision-request.png",
+      "sha256": "829347392941e65d8ec2e895cd503a9b600260ed638ddb662e15fcc423b797dd",
+      "width": 430,
+      "height": 859,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 33.93
+        },
+        {
+          "hex": "#edf1f6",
+          "share_percent": 18.81
+        },
+        {
+          "hex": "#fdfcfe",
+          "share_percent": 4.53
+        }
+      ]
+    },
+    "ui/desktop-qualification-cockpit.png": {
+      "source_repository": "OpenAdaptAI/openadapt-web",
+      "source_path": "public/desktop-preview/cockpit/10_dashboard_workflows.png",
+      "sha256": "83b854f7375bea220902bb6fca6d1ca40e31123691c0da4a7f4a435de37da65a",
+      "width": 1600,
+      "height": 1100,
+      "dominant_colors": [
+        {
+          "hex": "#f5f7fa",
+          "share_percent": 75.8
+        },
+        {
+          "hex": "#ffffff",
+          "share_percent": 16.49
+        },
+        {
+          "hex": "#f8f9fb",
+          "share_percent": 0.93
+        }
+      ]
+    },
+    "ui/mobile-decision-pending.jpg": {
+      "source_repository": "OpenAdaptAI/openadapt-web",
+      "source_path": "public/attended-decision/decision-pending.jpg",
+      "note": "The second colour, #fdfcf8, is the halt-state card wash derived from the canonical --accent-halt #9A5B0A. It is warm on purpose. The ground above it is #ffffff.",
+      "sha256": "a4d4af28c3627de9356ef1ee09d78137e3a2d530edc3cfa68c0d242299288b46",
+      "width": 314,
+      "height": 784,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 26.33
+        },
+        {
+          "hex": "#fdfcf8",
+          "share_percent": 16.52
+        },
+        {
+          "hex": "#edf0f5",
+          "share_percent": 14.03
+        }
+      ]
+    },
+    "ui/mobile-decision-request.jpg": {
+      "source_repository": "OpenAdaptAI/openadapt-web",
+      "source_path": "public/attended-decision/identity-request.jpg",
+      "sha256": "c50b46b5febaafafcef2a58e6ec2d9a6c5221a4386bdeacaa3a2f3836886b9b6",
+      "width": 314,
+      "height": 784,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 24.65
+        },
+        {
+          "hex": "#edf0f5",
+          "share_percent": 7.81
+        },
+        {
+          "hex": "#fbfdfc",
+          "share_percent": 2.91
+        }
+      ]
+    },
+    "ui/mobile-decision-result.jpg": {
+      "source_repository": "OpenAdaptAI/openadapt-web",
+      "source_path": "public/attended-decision/identity-verified.jpg",
+      "sha256": "e64f5de7ba84457baa5280cfac0cdcfea12f8c38009bd8dfb3443b794ed01042",
+      "width": 314,
+      "height": 784,
+      "dominant_colors": [
+        {
+          "hex": "#ffffff",
+          "share_percent": 21.85
+        },
+        {
+          "hex": "#fbfdfc",
+          "share_percent": 12.27
+        },
+        {
+          "hex": "#edf0f5",
+          "share_percent": 11.5
+        }
+      ]
+    }
+  }
+}