Knob now built across all 5 layers, validating the exercise. #7 confirmed
first-hand: the Knob build hit phantom theme tokens + uncontracted provider
CSS vars (silent drift), exactly as predicted.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The relocated Knob didn't render (only the number showed). Root causes, all
found by inspecting the loaded stylesheet (not guessing):
- A stray `*/` inside the header comment (`--scale-*/--primitive-*`) closed the
comment early and corrupted the FIRST rule, so the base `[data-knob]` rule
(which sets `--_knob-diameter`) never loaded → the control collapsed to height 0.
- Phantom tokens inherited from the Fable draft that don't exist in the theme:
`--color-neutral-content` (→ `--color-neutral-text`), `--state-hover`
(→ a `filter: brightness` hover); the pointer was transparent because its
fallback token didn't exist.
Design (per the user's reference image + directives):
- Value readout centred in the dial's hole (grid place-items).
- Donut arc: replaced an improvised hard-edged radial mask (aliased inner edge)
with an opaque face plate over the conic pie — a border-radius circle is
GPU-antialiased, so the inner ring is clean. Colours use theme-adaptive
tokens (surface-raised face, neutral-track ring, neutral-text value,
neutral-text-strong pointer).
- Value + label fonts scale with the knob via the same size coordinate as the
diameter (`--size-{k}-*` bundle), not an invented "one step below" rule.
Demo: value composed inside the Control (centre), plus a `needle` control that
conditionally composes `<Knob.Indicator>` (the compositional way to show/hide it).
Verified structurally (sizes, tokens, toggle) via DOM inspection; the screenshot
tool timed out all session, so the final pixel check is the user's. Recipe
contract green.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Two corrections after verifying the relocated Knob against the recipe contract:
1. The Knob was NEVER a contract violation — every `--knob-*` alias in the css
is consumed WITH a fallback (`var(--knob-x, default)`), which the
recipe-css-contract treats as runtime-optional, so no recipe-token
declaration is required. Removing `knob` from `WIP_TRACKS` keeps the full
eidos suite green (289) — it's now first-class, not an excluded WIP track.
2. But `size` was a visual no-op: the css hardcoded `--size-md-*` everywhere,
so `data-size` did nothing. Introduce private `--_knob-*` size vars selected
by `[data-size]` off the global `--size-{k}-*` bundle (public `--knob-*`
still overrides; the size-driven private is only the default; value readout
sits one step below). Browser-verified: control diameter now sm 60 / md 72 /
lg 88 px (was fixed 72 at any size).
npm run check unchanged at 59 (baseline); no console errors.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Canonical-depth demo at web/routes/uix/components/knob mirroring the Slider
demo (Live / API / morfo / sema / recipe / a11y tabs; every public prop is a
live control; sema event trace + play buttons; soma/eidos code snippets).
Serves as the visual/behavioral verification the relocation lacked. Confirmed
live at /uix/components/knob:
- renders role=slider, aria-valuemin/max/now, aria-roledescription resolved
from the langs catalog ("dial giratorio" under the es locale);
- provider-published `--knob-angle` / `--knob-progress` drive the conic-gradient
arc + rotated indicator (angle = startAngle + progress*sweep, verified exact);
- keyboard loop works end-to-end (ArrowUp/PageUp → value 40→52, aria-valuenow +
CSS vars + bound ValueText all update); no console errors.
npm run check: 0 errors in the demo (total unchanged at the 59 baseline).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Ships the net that batches A1-A3 / B3 land against, mirroring
scripts/docs-check.ts (textual invariants, no Svelte-graph imports):
- A-readme (error): every src/arts/{name}/ has a README.md. ethereal is
exempt via a self-cleaning README_PENDING set until batch A3 authors it.
- A-naming (error): the Engine*/Active* split holds — no Engine* root is
declared in a .svelte.ts (engines are stateless), and every reactive
`class Active*` root lives in a .svelte.ts (Active...Error exempt).
- A-index (warn): the src/arts/README.md Map lists every art. WARN, not
error — the four current gaps (bus, ethereal, orca, prefs) are batch B3's
worklist and drive to zero there.
Green now: 0 errors, 4 warnings (the B3 worklist), 22 arts.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
User decided (mirroring F7's corpus-shape call): arts stay
E5-in-place-reconciled with only active-app promoted to an architecture
chapter; orca stays E5 for now; the DESIGN docs move to docs/decisions/
translated. B1/B2 locked; A1/A2 (drift + language) unaffected.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The docs reconciliation (F6 drift-fix + F7 corpus-libro) covered
active-uix (morfo/soma/sema/eidos) but left the active-app side — the 22
arts/ runtime artifacts — as E5-in-place, un-reconciled. This plan does
for arts what F6+F7 did for the UIX layers, grounded in a 22-agent
per-art audit (every drift verified against the code): ~40 doc-vs-code
mismatches across 15 arts, 12 READMEs needing ES->EN, ethereal missing a
README, 3 historical DESIGN docs, dead route/demo cross-refs. Batches
A0 (guard) / A1 (drift sweep) / A2 (ES->EN) / A3 (ethereal README) / B1
(active-app -> architecture chapter, the missing analogue of
active-uix) / B2 (DESIGN docs -> decisions/) / B3 (index+TOC). Three
structural open questions for the user (arts E5 vs book; orca chapter;
DESIGN docs move) mirror F7's corpus-shape decision.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The Fable-built Knob lived entirely under `src/uix/active-uix/knob/` — the
wrong home for a multi-layer component — with guessed APIs (103 type errors),
missing sub-part wrappers, junk `mnt/user-data/outputs/` paths, and its angle
math inlined against a fabricated `Gesture.base` surface. Dismember it into the
canonical per-layer directories, against the real APIs, consuming the real
`Gesture.rotate` (STUMBLES S4):
- **morfo** → `morfo/components/knob.ts`. Applies the S3 `part-absent` condition
to `aria-label` (emit only when no Label part) — the file's own NOTE asked
for it once the condition existed.
- **langs** → `langs/components/knob.ts` (+ registered in the langs index).
- **sema** → `sema/components/knob.ts` (+ registered in the sema index); valid
nominal earcons + haptic kinds, `channels` declared, risk-commit composes
(`op: 'add'`) over the intent profile per doctrine.
- **soma** → `soma/components/knob/`: provider (rewired to `Gesture.rotate` in
the constructor — mapping `angleAbsolute` onto the [startAngle,endAngle]
sweep, folding [0,360) into a symmetric range), `state<T>()` for the label id,
the missing `KnobIndicatorProvider`, root + five sub-part wrappers, exports,
index, and a provider test (ARIA slider contract + gesture wiring + keyboard
stepping/commit + disabled guard + dblclick reset).
- **eidos** → `eidos/components/knob/`: recipe css, size-aware root wrapper
(`ActiveEidos.resolve`), five pass-through wrappers, types, namespace index.
The old `active-uix/knob/` (untracked, added mid-session) is removed. Knob is
added to `recipe-css-contract`'s WIP_TRACKS until its recipe tokens land (css
still hardcodes md sizes, no variants). `npm run check`: 0 errors in every knob
file (total back to the 59 pre-existing baseline, from 162). Tests: 4 knob
provider + no regressions across eidos/sema/langs (174).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Ran the full 65-test port suite on WebKit + Firefox (UIX_CROSS_BROWSER=1),
not just chromium. That caught one real robustness gap: Firefox stores
scrollTop as a subpixel value (~59.93 for a 60px scroll), so ScrollState's
exact-equality assertion is relaxed to toBeCloseTo — the port faithfully
passes the browser's own scrollTop through.
New coverage:
- tabbable.ts WRAPPER (getTabbableCandidates / getTabbableEdges /
getTabbableFrom / getTabbableFromFocusable) — the surface soma's
focus-scope actually consumes, previously untested.
- resource: AbortController cancellation + onCleanup on rapid source change
(aborted request must not clobber state with an AbortError).
- FiniteStateMachine: from/to/event/args lifecycle metadata, send-arg
forwarding, and no-action-for-event (no transition).
All green on chromium + webkit + firefox. (Repo check baseline is ~214 due
to the unrelated in-progress Knob component; 0 errors in these files.)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Fourth gesture specialization alongside base/drag/resize, for radial controls
(Knob, and future AngleSlider / colour wheel / rotary time picker). The Knob
built by Fable (active-uix/knob) currently does its angle math inline against a
guessed `Gesture.base` API and flags the gap — this is the primitive it needs.
`GestureRotate extends GestureBase`, mirroring the existing peers (protected
onMove/onRelease/onCancelled overrides + its own opts callbacks):
- `angleAbsolute` — pointer angle around a centre, 0° = up (12 o'clock),
clockwise positive, [0, 360). Matches CSS `rotate()`, so it pipes straight
into a `--rotation` var (the convention the real Knob consumer uses).
- `angleDelta` — signed rotation since start, unwrapped across the 0/360 seam
(355°→5° is +10°, not −350°); multi-turn friendly.
- `angularVelocityDeg` — deg/ms via a dedicated VelocityTracker fed the
accumulated angle, so it reads d(angle)/dt (radius-independent).
- Detents (`AngularSnap`: fraction / degrees / '45deg' / '50%'), wrap-aware
distance, dismiss by accumulated degrees, `center: 'element' | {x,y}`
(captured once at pointerdown — no per-move layout read).
Deviations from the scoped plan, all deliberate: angle convention taken from
the real Knob consumer (`atan2(dx,-dy)`, not the plan's `atan2(dy,dx)`);
`referenceSize` omitted (no meaning for a radius-independent angle);
CW/CCW direction locking deferred to v1 as the plan itself noted.
Two release-time snap bugs (found by adversarial review) fixed before landing:
- angular velocity projection now scales for the deg/ms domain and caps below a
half-turn, instead of reusing the px-scaled `VelocityTracker.project` (a
1 deg/ms flick would otherwise project 400° and cross a full turn);
- the sequential-detent guard measures neighbour adjacency modulo the detent
count under `wrap`, so the first/last detents are treated as physically
adjacent (the linear `|i-current|>1` from drag broke across the seam).
Tests: 9 new cases (angle convention on all cardinals, seam unwrap, onRotate,
detent resolution, dismiss by degrees + no-dismiss under threshold,
single-pointer, cancel). `npm run check` = 0 errors in the gesture layer.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The emit contract explains discrete events (one gesture → one emit → one
hold) but says nothing about direct-manipulation surfaces (Slider, Drawer,
future knob/swipe) where the pointer moves at sample rate while the value
updates continuously. Add a `## Continuous components (drag, swipe, hold)`
section after the emit contract, grounded entirely in the real Slider/Drawer
code (adversarially fact-checked, zero discrepancies):
- One door: `runtime.trigger()` is the sole path (there is no `emitEvent`).
Its Promise represents the whole occurrence incl. the visual hold — so
`void trigger()` is fire-and-forget (correct for high-frequency emits) and
`await trigger()` blocks until the hold ends (correct when a structural
change must observe the resolved signal, e.g. `close`).
- `coincident` vs `post` for a moving value: the move (handle-drag,
drag-progress) is `coincident` (signal + mutation indivisible); the commit
(commit-set) is `post` (handler settles, then the signal celebrates).
Slider is the worked example.
- Throttle continuous POINTER emits, not keyboard: pointer emits are coalesced
to animation frames + floored to a component-tuned interval via ActiveDom's
requestFrame (Slider DRAG_SIGNAL_MS, Drawer DRAG_PROGRESS_SIGNAL_MS);
keyboard commit-set fires once per keydown (already human-paced). Tuning ms
point to the code, never hard-coded here.
- The per-emit `overrides` payload owns pitch/gain/contour; a continuous
event's cascade rule sets only `channels` and MUST NOT override those
primitives, or it would clobber the live payload every frame.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
A line-by-line audit of the ports against runed's original source found two
faithfulness regressions in unused F3 utilities:
- IsIdle seeded `lastActive` to 0 (epoch) instead of the construction
timestamp, so `Date.now() - lastActive` read ~56 years until first activity.
Now `$state(Date.now())`, matching runed.
- IsInViewport passed `threshold: undefined`, which the IntersectionObserver
spec coerces to 0 (fires at a single pixel). runed defaults it to 0.1;
restored (and rootMargin -> '0px').
Guarded IsIdle's baseline in its test. check=59.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
65 tests in real chromium covering every ported utility, so the removed
runed + tabbable deps can be relied on with confidence:
- tabbable-core: order, tabindex, disabled, real display/visibility, inert,
disabled fieldset, radio groups, contenteditable, details, shadow DOM.
- $libs/reactive runes: watch / watchOnce / watch.pre, Context, Previous,
Debounced/debounce, Throttled/throttle, interval, StateHistory,
FiniteStateMachine, resource, IsMounted, onCleanup.
- $adom DOM runes: ElementSize, ElementRect, ActiveElement, IsFocusWithin,
IsIdle, IsDocumentVisible, IsInViewport, onClickOutside, PressedKeys,
ScrollState, TextareaAutosize, AnimationFrames.
Patterns: $effect.root + flushSync for runes, vi.useFakeTimers for
debounce/throttle/interval/idle, vi.waitFor for real observers/rAF, and
rendered fixtures (context-probe, scroll-state-probe) for the two utilities
that need a component context. `npm run check` = 59 (baseline).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Completes the runed + tabbable removal (F1+F2 in d0f41fb2). Ports runed's 11
DOM-reactive utilities into $adom, window-correct by resolving the window from
the node (getWindow) or an optional window/document — not runed's defaultWindow:
AnimationFrames, ActiveElement (+ activeElement singleton), IsDocumentVisible,
IsFocusWithin, IsInViewport, onClickOutside, PressedKeys, ScrollState,
TextareaAutosize, ElementRect, IsIdle.
Follows the committed ElementSize precedent (raw observers/listeners live in the
arts/adom implementation layer) rather than threading a `dom` param.
PersistedState is intentionally NOT ported: the $storage art already supersedes
it (its README names runed/PersistedState as the thing it replaces) — same call
as boolAttr being covered by boolToEmptyStrOrUndef.
`npm run check` = 59 (baseline, 0 in ported files); server tests unchanged
(pre-existing failures only). The runed + tabbable port is now complete.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The common ARIA pattern "emit aria-label only when there is no Label part"
was not declarable. Add `{ when: 'part-absent', part }` as the logical
inverse of `part-present`, plumbed through the four condition sites:
- types.ts: MorfoCondition union member + JSDoc.
- schema.ts: discriminated-union member (structural accept) + extend the
checkCondition unknown-part guard to cover part-absent (throws on unknown
part, same as part-present — the prior guard only discriminated part-present).
- compile.ts: collectConditionDeps adds the referenced part to partRefs, so
the runtime re-evaluates when the part's presence toggles.
- resolver.ts: shouldEmitMorfoEntry returns !bindings.parts?.[part] — emit
when the part is absent/falsy, suppress when present.
Runtime + keyboard paths delegate to these helpers (evalAttrPlan /
shouldEmitMorfoEntry), so no runtime change is needed. Correct by symmetry
with part-present, which reads the same presence-accurate bindings.parts.
Tests: schema accepts a part-absent condition on a known part and throws on
an unknown part; the compiler collects the part dep; the resolver emits the
value when the part is absent and undefined when present.
Docs: morfo.md condition list gains the part-absent form.
Non-breaking: no existing morfo uses it.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Removes the two runtime dependencies that broke the "framework depends on
nobody" doctrine, porting them into our own code (same move as
floating-ui -> $ethereal).
F1 (dropped the deps):
- tabbable -> src/libs/dom/tabbable-core.ts (getComputedStyle/CSS.escape
resolved from the node's own window, iframe-safe); consumed by the
existing tabbable.ts wrapper and surfaced via $adom.
- runed Context + watch/watch.pre/watchOnce -> $libs/reactive.
- runed ElementSize -> $adom/element-size.svelte.ts, rebuilt window-correct
on the node's own window (not defaultWindow).
- Rewired ~33 watch + 3 Context imports; floating (ElementSize) and
focus-scope (tabbable) now consume $adom.
F2 (rest of the pure reactive runes -> $libs/reactive):
- Previous, IsMounted, onCleanup, interval, debounce/Debounced,
throttle/Throttled, StateHistory, FiniteStateMachine, resource.
- Dropped runed's React-style use* prefix; reused framework toValue
(=extract) + MaybeActiveOrGetter; added Setter<T>.
runed + tabbable removed from package.json. `npm run check` = 59 (baseline,
0 in ported files); tests unchanged (pre-existing failures only).
F3 (remaining DOM-reactive runes -> $adom on ActiveDom, + PersistedState
-> $storage) is pending: see docs/process/continue-runed-tabbable-port.md.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Continuation doc for tomorrow: full session state (C7 + eidos triage +
onion-menu/surface-subtle + functional flags + C10 reframe + S1
vocabularies), plus the three high-confidence implementation plans the
scope-stumbles-s2-s6 workflow verified against the real code (S3
part-absent condition, S2 continuous-components doctrine, S4
Gesture.rotate) with exact file:line edit maps. S5/S6 flagged for
re-scope. Foreign runed/tabbable WIP in the tree left untouched.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
An agent building a component from the docs-book alone could not assign
archetypes / holds / haptic kinds: the closed sets live only in code consts
and the corpus (correctly) forbids copying them into prose, so they were
invisible from the docs. STUMBLES #1, the top stumble.
Fix: docs/canon/vocabularies.md is GENERATED from the consts by
'npm run docs:vocabularies' — the ONE sanctioned place the lists are spelled
out (generated = no drift objection), which every other doc links to. It
covers part archetypes (with a one-line role each), sema families with their
default hold/persistence + per-intent overrides, verbs by family, intents,
the perceptual-duration scale, haptic kinds, the 33 palette scales, sizes,
variant archetypes, and the shared common.* strings.
- scripts/docs-vocabularies.ts imports the consts (robust — no fragile
JSDoc/regex parsing) and formats them; the generation is an exported
function so docs-check can compare.
- Added ARCHETYPE_DESCRIPTIONS to morfo/types.ts (co-located with
ARCHETYPE_VOCABULARY, satisfies Record<MorfoArchetype,string>) as the
authoritative one-liners the appendix reads — descriptions become
first-class data instead of inline union JSDoc a tool would have to parse.
- docs-check gains I7: it regenerates in-memory and fails if the committed
file drifts from the consts (negative-tested: corrupt -> error, regen ->
green). The appendix's own counts pass I1, so I1 doubles as a second guard.
- Linked from README (E2 canon row + two 'I want to' shortcuts), CANON.md
(doctrine here, enumerated lists there), architecture/morfo.md (pick a
part's archetype), and the component-audit minimum-package list (STUMBLES
#8).
docs:check 0 errors; npm run check unchanged in my files.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Cleared the error-severity flags that were mature components missing a
documented exception, not real gaps:
- 4 service components (format-date, format-number, relative-time, trans)
get their E-2.2 exception: they render text via $format/$langs with no
visual recipe — no {name}.css by design.
- alert-dialog gets its R-1.1 exception: it composes Dialog (soma reuses the
Dialog runtime, data-dialog-* markers reach the DOM, dialog.css paints the
chrome) — a [data-alert-dialog] root rule would style nothing.
- card gets its missing A-1.3 texts.label (+ langs entry, registered) —
card had no label at all.
C10 (maturity tiers) reframed after the classification exercise showed the
audit VERDICT is the maturity oracle, and 'experimental' is nearly empty
once exceptions are documented: stable = PASS (82); preview = NEEDS-WORK on
docs/demo only (E-2.3/F-1.x/D-*) — functionally sound but underdocumented,
which includes mature components like button/table/listbox (a README gap is
not immaturity); experimental/WIP = a real FUNCTIONAL gap or foreign track.
C10 folds into C8 (document the tier + exceptions when writing READMEs).
Real gaps registered, NOT silenced: float-panel A-3.7 (28 mutating keyboard
actions, 6 events — a morfo-completeness question), card-group (no morfo),
natural-time-picker A-3.1; announce/clipboard R-1.5 (live-region /
composed-focus) await their READMEs in C8 to host the exception.
npm run check at the 59 baseline; alert-dialog -> PASS.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two real defects the audit's error-severity rules caught:
- onion-menu shipped no default export (E-1.3), unlike every sibling compound;
add 'export default OnionMenu'.
- table + tree-grid referenced 'var(--color-surface-subtle)' — a token that
does not exist (the surface ladder is default/raised/muted/overlay). With no
fallback it resolved to guaranteed-invalid, so the table header + striped rows
rendered with NO background (a silent visual bug R-2.6 flagged). Repointed both
to '--color-surface-muted' (the existing subtle-surface semantic, used for
readonly/muted contexts); verified in the browser the header now paints muted.
A dedicated 'subtle' surface tier is a theme-canon decision, deferred; chronos
(WIP, untouchable) keeps the same reference for its author.
npm run check at the 59 baseline; tree-grid R-2.6 clears.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Accumulated test debt (the eidos suite had been red across several feature
landings — THEME-SYS-1 named z-index, palette-first static donor library,
spin-field extraction, icon stroke-map refinement — none of them regressions;
baselined against HEAD). Each triaged stale-vs-bug by confirming current
behavior in code before touching a line.
Stale tests (code correct/improved, assertion not updated):
- z-index '76' -> 'var(--z-index-overlay-tooltip)' (the code is guarded by
THEME-SYS-1 in contracts.test.ts — the raw integer is forbidden).
- frost '80%' -> 'var(--depth-overlay-translucency, 80%)' (parameterized).
- C7 size-bundle coordinate ('--control-height-md' -> '--size-md-control-height').
- icon visual-stroke ratios 1.5->1.125, 1->0.923 (stroke map md 2->1.5, xxl
icon 48->52px; the arithmetic follows).
- delegate/CSS-only purple-9: the slim foundation legitimately emits the
role-referenced donor scales (purple=primary) as overridable :root defaults;
the real delegation guarantee is 'no baked [data-theme] block', which the
assertion now checks.
Test over-flagging (fixed the test, not the code):
- CSS-only shared layer: spin-field (no wrapper/index) excluded from the
wrapper/barrel/namespace contracts (CSS_ONLY_LAYERS).
- soma-runtime vars: 'var(--x, fallback)' is runtime-optional (the STUMBLES #7
soma->eidos surface) — only bare 'var(--x)' requires a recipe declaration.
- orphan check made corpus-wide: composition means a sibling consumes a
component's tokens (date-picker/month-grid consume calendar-control-*).
- natural-time-picker fixed-tone sky colors documented-exempt (Words-rail
doctrine).
- dialog drift-detector converted to a synthetic fixture (dialog.css no longer
ships the dead rule it self-tested against).
Real code fixes:
- declare '--pin-input-cell-readonly-bg' (I consumed it in C2b without
declaring it).
- remove 5 truly-dead tokens (drawer content-bg + 4 color-field
format-select/segment — 0 consumers anywhere in eidos/soma/web).
- clamp time-picker's 'size' to PopoverSize before forwarding (xs->sm):
TimePickerSize has xs, PopoverSize doesn't — a latent type bug my recipe
change surfaced via a complexity-budget flip. The fix dropped check 61->59,
also clearing navigation-menu-content of the same class.
card-group flagged as a REAL gap (KNOWN_MISSING_MORFO): it ships parts + chrome
but no morfo (morfo-first violation) — tracked, not silenced.
Verified: vitest src/uix/eidos 23 files / 289 tests green; npm run check 59
(better than the 61 baseline); component:audit 0 BROKEN.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The ⚠️ 'canon + guard, not consumed' asterisk is retired: the C7 sweep
made all 34 size-bearing recipes consume the bundle coordinate, so the
row is ✅ (canon + enforcement + consumption). The footnote also drops
the stale control/compact/dense archetypes (superseded by the 2026-06-17
1:1 override) — control text follows the typographic scale 1:1.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The size bundle (--size-{k}-control-height / -font-size / -icon-size,
theming/reference §5) was emitted + guarded but consumed by ONE recipe
(the toggle pilot). This sweep points every recipe at the canonical
size COORDINATE instead of the raw primitive of the same coordinate:
366 refs across 34 recipes (--control-height-{k} -> --size-{k}-control-
height, and the font-size / icon-size coordinates likewise). Retuning a
size's bundle now reaches every consumer; a recipe that maps a key to a
DIFFERENT size's coordinate stays visible as a deliberate deviation.
The bundle vars are pure aliases of the primitives, so this is
computed-value identical — verified in a live browser: the accordion
chain --accordion-trigger-min-height-md = --size-md-control-height =
--control-height-md all resolve to calc(36px * 1 * 1). The generated
diff is 366-for-366 pure name swaps, nothing else moved. base/xxxl have
no size bundle and correctly stay on the typographic primitive.
Guard: recipe-css-contract forbids the raw size-coordinate primitive in
recipe token values (var(--control-height-{k}) etc.) — the drift can't
creep back. This is the ⚠️->✓ that theming/notes' comparison table flags
as the size canon's remaining asterisk (canon + guard, now + consumption).
component:audit 0 BROKEN; npm run check at the 61-error baseline; the
size-bundle guard + palette-cascade guard both green.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two user findings from the Knob build exercise (an agent building a new
component from the docs alone — STUMBLES.md).
Quarantine: the superseded fossils no longer share shelf space with the
live corpus. docs/old-deprecated/ (with an index README explaining what
lands there and pointing readers at docs/README.md) now holds the
executed audits and fix plans: fable_audit, fable-eidos-audit,
inherit_audit + inherit_fix_plan, ARCHETYPE_COHERENCE_AUDIT_2026-06-19
(still citable — the component-guide banner and the eidos components
README repoint to it), COMPONENT_COHERENCE_AUDIT. pendiente.md (a live
pending list, not a fossil) moved to docs/process/. docs-check treats
the folder as sealed chronicle (I1/I2 exempt; I6 skips its internal
links, as its README promises). Root-level *.md is now: README, CLAUDE,
AGENTS + the user's own working files.
STUMBLES fixes applied on the spot (the doc-class ones):
- #2 kind drift: the REAL enum is 'public' | 'private' | 'virtual'
(MorfoPartKind, 671/9/14 uses) — morfo.md omitted 'private', the
checklist invented 'internal' (0 uses). Both fixed; I2 gains the
phantom-'internal' guard. A-2.1's row now says what the audit script
actually checks (kebab only — the archetype may vary, the Toggle
provider-IS-trigger doctrine).
- #6: the langs catalog SHAPE (flat keys, per-language leaves, named
export, index registration) is now shown in morfo.md instead of only
its location.
- #8: component-audit s0 defines the minimum brief package as an
explicit 8-file list.
The engineering-class stumbles are registered as plan batches S1-S6
(generated vocabularies appendix, continuous-gesture trigger doctrine,
part-absent condition, Gesture.rotate, the soma->eidos CSS-var
contract, minor frictions). docs:check 0 errors, 11-warn baseline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
palette-{slot} is now RESERVED recipe vocabulary: normalizeRecipeTokens
(render-css) appends the full per-scale color:{scale} cascade for every
donor scale in PALETTE_SCALES to any token named palette-track/element/
border/solid/solid-hover/text/contrast — author-declared color:*
declarations win, only absent scales are appended. This is the
structural end of the hand-maintained-subset era: a recipe opts into
the per-instance palette by naming the token, and the generator
guarantees all scales (the 31-vs-33 drift class cannot recur at this
layer). Applies to app-config recipes too, since the hook sits where
every RecipeTokenSet is normalized.
Button's seven manual spreads are gone from base.ts (the helpers moved
into the generator); its generated output is byte-identical. Toggle —
which already exposed palette-* tokens — universalized itself: +330
generated lines, and <Toggle color="teal"> / color="steel" verified
resolving in a live browser (steel and fuchsia are exactly the two
scales the hand-kept lists used to miss). eidos-lint now classifies
donor-scale values on data-color as sanctioned eidos-only vocabulary
(the TSC color:* axis extension) instead of invalid — the morfo enum
keeps declaring the semantic roles, per the pilot's deliberate
runtime-open design. Guard: recipe-css-contract pins the LAST scale of
PALETTE_SCALES for every palette-bearing recipe in the generated css.
Also: the stale '31 physical color scales' test TITLE says 33 (its body
already asserted 33). Recorded in the plan: the eidos suite carries 15
pre-existing failures in 6 files (baselined against HEAD before this
change) — their triage is its own batch. npm run check at the 61-error
baseline; lint toggle baseline green again.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The 30 eventful morfos that never declared expression: ALL have a
per-component pack in src/uix/sema/components/ — so the classification
(book-deviations D.4: pack / family-default / delegated / none) is
unambiguous: expression: 'pack' across the board, placed after scope
per the switch/button idiom. The sema surface of the contract is now
explicit on all 82 eventful morfos instead of inferred from pack-file
existence. morfo:vocabulary clean (the one naming warn is media-player,
foreign WIP); npm run check at the 61-error baseline. Additional
persistence/a11ySemantic annotations deliberately NOT mass-applied —
per D.9 the canonical table is reference, authors declare per real
consumer (the six that need them already do).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Every interactive morfo now either anchors its W3C APG pattern or
documents that none exists. Real patterns (20): the picker family +
command + search-field anchor combobox (field/input + popup, the
pattern's own popup taxonomy); date-field mirrors time-field's
spinbutton (per-segment contract); rating-group -> radio;
navigation-menu -> disclosure (the APG's own site-nav guidance);
splitter -> windowsplitter; tag-group/tags-input/virtual-grid -> grid;
clipboard/file-upload -> button (the trigger IS the widget);
announce -> alert (live region); form -> the form landmark;
image-adjustments -> slider; table -> table.
The A-1.4 valve (documented exception) is wired like the R-x rules,
and five components whose widget has NO APG pattern say so in their
README instead of faking an anchor: cropper (2-D manipulation),
pagination (nav + composed Buttons), stepper, timeline (display list),
image-picker (media picker). Remaining warns are the four expected:
card/drag-drop/virtual-list host their note when C8 writes their
READMEs, media-player is foreign WIP.
Catalog: 83 PASS / 47 NEEDS-WORK / 0 BROKEN (was 75/50/5 at the
audit's start). npm run check at the 61-error baseline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Two classes. Real fixes: 8 numeric font-weights tokenized to
var(--font-weight-medium/semibold/regular) in command,
date-range-picker, link-preview, listbox, menubar, textarea,
time-range-picker x2 — same rendered weight, but the foundation now
propagates (verified in a live browser: the token resolves to 500).
Honest annotations: the proportional/identity literals get their
/* literal: <reason> */ valve instead of a fake token — avatar
initials/badge scale with the avatar diameter (physically
proportional), the calc(1em - (md - sm)) family is the Field
label-one-step-down doctrine with a token-derived delta, 1em is
context identity (fab glyph, password toggle), the x0.85 captions are
proportional to their field token, and combobox's -1px is an optical
adjustment on a token-derived size. component:audit R-2.7: 16/16
clean; npm run check at the 61-error baseline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The R-1.2/R-1.3/R-1.4 sweep, same method as C2a (evidence first): each
offender is either a REAL missing state or a shared-layer/composed case
that must not duplicate rules.
Real fixes, all verified in a live browser (computed-style probes, with
the transition-mid-flight trap from the demo guide s13 accounted for):
switch/toggle/rating-group get the family invalid tint
(--color-threat-element, the shipped checkbox/pin-input precedent);
toggle gets its readonly cursor; pin-input cells get the readonly
muted-surface treatment mirroring date-field segments; the SHARED
spin-field layer gains the readonly treatment it lacked (one fix covers
css-field + number-field); natural-time-picker gets the disabled
opacity/pointer rule mirroring date-field.
Exception valve wired into R-1.3/R-1.4 (was only R-1.1/1.2/1.5/E-2.2)
and cited exceptions added: css-field/number-field (states live in the
shared spin-field layer), dialog (data-disabled sits on the VIRTUAL
provider — defaultElement none — and the pressable surfaces are
composed Buttons), color-picker (color-field owns the invalid rules),
date-range-field (the runtime overlays data-date-field-input on the
same elements, per its own recipe header), time-picker (composed
time-field chrome). natural-time-picker keeps its two warns for its C8
pass rather than guessing paint targets.
All 10 audited components now PASS with 0 R-1.x pending; npm run check
at the 61-error baseline. css-field/number-field READMEs fixed on disk
but unstaged (unrelated local modifications).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The audit's focus check grepped only for :focus-visible, blind to how
the framework actually shows focus: the universal archetype ring
(archetypes.css [data-archetype]:focus-visible with its documented
exclusions), the field shell's :focus-within, and the canonical
state-attr ring ([data-focused] + focus-ring tokens, soma-written — the
spin-field/search-field pattern). Blindly adding :focus-visible rules
to the 10 flagged recipes would have recreated the exact 'double
border' anti-pattern archetypes.css documents against. The check now
accepts the four sanctioned evidences (textarea and search-field pass
on their own css; date-picker passes via exception) and the remaining
non-ring cases carry cited R-1.5 exceptions: css-field/number-field
(ring lives in the shared spin-field layer), dialog (content panel is
the documented foundation exclusion; close/trigger are composed
Buttons), date-picker (foundation ring + content exclusion),
metrics/timeline (no focusable part — interactive only by events).
clipboard/announce host their exception when C8 writes their READMEs.
css-field/number-field READMEs fixed on disk but left unstaged (they
carry unrelated local modifications).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The 5 BROKEN components (cascade, gradient-builder, menu-dial, motion,
qr-code) are BROKEN no more: cascade/motion/menu-dial now PASS,
gradient-builder/qr-code drop to NEEDS-WORK with only demo-phase (D-*)
and C8 items left.
Framework-level piece: component-audit gains the documented-exception
valve the checklist already used for A-2.3/R-1.7 — a greppable
'R-x.y exception: reason' line in the component README turns the rule
into a PASS that reports the reason. Wired for R-1.1, R-1.2, R-1.5 and
E-2.2; the checklist rows say the same. This separates deliberate
design (cascade and motion deliberately ship NO recipe — they ride the
foundation stagger + state presets; menu-dial's focus/disabled states
live in the composed Fab/Button recipes) from plain omission, which
stays an error.
Mechanical fixes: texts.label + langs entries for cascade/motion (new
files, registered) and gradient-builder (label added to its existing
entry); menu-dial's missing default export. README contract sections
(Baseline/Comparativa/Decisiones/Gaps/Passive justification + Audit
exceptions) added to cascade, motion, menu-dial and qr-code — mostly
re-heading content those docs already argued; comparativas grounded in
M3 speed dial/MUI SpeedDial/PrimeVue, Framer Motion/AnimatePresence/
Svelte transitions, ark-ui/qr-code-styling per component.
Verified: component:audit 130 -> 78 PASS / 52 NEEDS-WORK / 0 BROKEN;
npm run check at the 61-error pre-existing baseline (0 own).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Extending the I1 count guard to color roles and role slots surfaced
four more stale counts the palette fix predicted: the RFCs still said
'13 slots' and enumerated borderHover8 (the slot was retired 2026-07-02,
zero consumers — COLOR_ROLE_SLOTS is 12), and two component READMEs
(css-field, float-panel) still said '8 roles' from before tertiary made
it 9. The guard itself had the same disease it polices: extractConstArray
counted the quoted 'borderHover' inside the comment documenting its own
retirement, so expected=13 and the stale mentions passed — comments are
now stripped before extraction. css-field's README is fixed on disk but
left unstaged (it carries unrelated local modifications).
docs-check now verifies families=8, intents=6, archetypes=26,
palette=33, roles=9, slots=12 against their consts on every run.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
User-caught drift, and it was live in the UI, not just prose: the demo
harness's hand-copied PALETTE_GROUPS enumerated 31 scales — fuchsia and
steel (the two non-Radix scales) never appeared in any demo's
PalettePicker even though PALETTE_SCALES (the const, 33) and the recipe
cascade (paletteScaleDecls maps the const) resolve them fine. Fixed the
two missing entries, typed the groups as PaletteScale[], and added a
dev-time completeness assertion so growing the const without grouping
throws at module load.
Prose swept (the numbers were inherited verbatim from the pre-book
sources, which contradicted their own '33-scale library' headers):
decisions.md color-model row, demo-authoring x4 (incl. the '31-scale
palette' lede), rfc-color-engine x4 (ground-truth + templates
open-question). Where the count was incidental the text now points at
PALETTE_SCALES instead of restating a number.
Guard: docs-check I1 gains the palette count (EN + ES patterns,
~-approximation and type/space-scale exclusions; s-mark added to the
section-number skip). The check line now prints palette=33. This was
the exact class I1 already covered for families/intents/archetypes and
had not been extended to the palette.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Consolidated audit (2026-07-02): 130 components -> 75 PASS / 50
NEEDS-WORK / 5 BROKEN; palette-31 wired in 7 recipes, size-bundle in 1,
focus-visible missing in 11, 17 literal-typography recipes, 41/82
eventful morfos without expression:, 33 without APG anchor, 28 without
README. Demos (D-*) explicitly deferred to the next phase (component
info web). Batches C1-C10 with per-batch verification.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
CLAUDE.md 981 -> 350 lines. The 12 session hand-offs (2026-05-08 ->
2026-06-05, ~513 L) moved verbatim to docs/process/handoffs-claude-md.md
(chronicle frontmatter). Reference Documents now point into the book
tree, and the long-stale authority claim is fixed: the semantic
vocabulary source is docs/CANON.md (+ decisions/book-deviations.md),
with the Spanish guide correctly framed as a historical seed at its new
decisions/ home — the correction the guia's own banner had been
announcing since Fase 6. The four sema/eidos doctrine sections (~170 L
of copied tables, cascade diagrams and examples — the copy class
docs:check bans in the corpus) compact to the operational MUSTs
(semaSelector mandatory, packs never replace intent primitives,
channels split by owner, types-over-lint) each with its pointer to
CANON.md / architecture/sema.md / theming/channels.md /
architecture/eidos.md. Two head refs swept (eidos-motion ->
theming/motion.md, COMPONENT_GUIDE s4 -> guides/component-guide.md).
Approved by the user after diff review. docs:check 0 errors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The global sweep caught the last corpus links still pointing at
pre-book paths: morfo chapter's s2.bis cross-ref (src/uix/README ->
overview.md), authoring's E1 stratum row (active_architecture ->
docs/architecture/), the related: frontmatter of decisions.md and
canon/recipe-contract.md (THEMING/THEMING_GUIDE -> theming/reference +
guide), and component-audit's typographic-vertebration pointer (eidos
README -> architecture/eidos.md with the translated heading). README
map completed as the book TOC: every chapter now has a row —
theming/motion.md (the model) and theming/changelog.md (the chronicle)
were the two missing. Remaining old-path references live only in
chronicle docs (theming/changelog, process/) and resolve via stubs;
the 11-warn baseline is all foreign targets (MOTION_SERVICE_RFC demo
routes, deleted THEMING_AUDIT/PENDIENTES).
F7.6 complete. The book tree is fully populated: architecture/ (8) +
canon/ (2) + theming/ (7) + rfcs/ (7) + guides/ (4) + decisions/ (2).
Next and last: F7.7 CLAUDE.md thinning — diff to the user first.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
LIBRO_VARIACIONES_Y_EXTENSIONES -> decisions/book-deviations.md and
GUIA_IMPLEMENTACION_SEMAUIX -> decisions/guia-semantica-historica.md,
both moved AS-IS in Spanish: the deviations registry is a logbook of the
author's literal decisions ('transcrita literal') and carries proposed
doctrinal text destined for the Spanish book — translating it would
destroy that function (s G calls itself bitacora); the guia was already
status: historical (Fase 6) and the plan exempts it explicitly. English
frontmatter added to both; internal cross-links repointed (CANON,
theming/reference, book-deviations D.11). docs-check's I2 phantom-field
exemption follows the moved file (it matched by the LIBRO_VARIACIONES
filename; now also matches decisions/book-deviations.md). Corpus swept:
CANON x3, README (E2/E3 strata + tables — also fixed the pre-Fase-6
leftover row still calling the guia 'authoritative for any new wrapper'
and the unswept eidos/TSC.md stratum mention), building-a-component D.4,
completion-checklist G-1.1 + header, theming/reference, architecture
x5 (active-architecture, eidos, morfo, overview, sema x2).
docs:check 0 errors, 11-warn baseline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
src/uix/soma/COMPONENT_GUIDE.md (1449 L, English body) moved to
docs/guides/component-guide.md. The only Spanish passage — the
archetype-canon banner (2026-06-19) — translated in the move; the rest
is verbatim. Banner's ARCHETYPE_COHERENCE_AUDIT link reaches into src
(audit fossil, stays with the code); the checklist cross-link is now
the sibling ./completion-checklist.md and the 'soma/README s9' mention
now names architecture/soma.md (its post-F7.2 home). Corpus swept:
authoring + README E4 stratum rows (also stale THEMING_GUIDE name),
building-a-component phases 0/2, architecture/{soma-architecture s4/
s13-tree/s17, overview, morfo, active-architecture}.
F7.5 complete: docs/guides/ = component-guide, completion-checklist,
demo-authoring, component-audit. docs:check 0 errors, warns back at
the 11 baseline.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Three English guides moved verbatim (copy + link touch-ups, no
translation): COMPONENT_COMPLETION_CHECKLIST -> guides/
completion-checklist.md, COMPONENT_AUDIT_GUIDE -> guides/
component-audit.md, DEMO_AUTHORING_GUIDE -> guides/demo-authoring.md.
Tooling moved with its doc in the same pass: docs-check I5 now parses
the checklist at the new path (verified: 0 errors, rule tables found)
and component-audit.ts's two path strings updated. One reconciliation
in the audit guide's s5, applied as link-don't-copy: the copied v1
6-tab template list (stale vs the v2 9-tab guide it links, drawer
canary vs button) is replaced by a pointer to demo-authoring.md; the
domain canaries in s7 stay. demo-authoring's harness links now reach
into web/routes/uix/lib/ (the modules stay with the code); the deleted
pendiente_color_demos.md reference is marked TODO(reconcile). Corpus
swept: building-a-component (frontmatter related: + phases 6/8),
authoring, architecture/soma, theming/reference s5-parity,
testing-and-tooling, getting-started, README (E4 rows + component-audit
row added + two stale F7.3 paths in 'I want to' fixed). Remaining
warns = links to component-guide.md, which lands in F7.5 (2/2).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
COLOR_ENGINE_RFC (619 L, the largest RFC) -> rfcs/rfc-color-engine.md,
Spanish to English, same section numbering (s0 TL;DR, s6 generator +
s6.1 isomorphism + s6.2 deriveScheme/buildScheme, s7 wide-gamut
strategies, s8 APCA, s10 frozen invariants, s11 phases with 4-bis
implemented, s13 bloat-doc absorption, s14 comparison). Its
THEMING_AUDIT citation notes the audit is retired; the s25/s26 canon
citations point into theming/reference.md. decisions.md row repointed +
Status aligned (implemented through phase 4-bis); the index's naming
note rewritten — the deferred eidos-RFC rename is now DONE via stubs,
while the arts DESIGN_* docs keep their legacy names in place (their
citations were not swept).
F7.4 complete: 7 RFCs live at docs/rfcs/rfc-{scaling, structure, depth,
shape, typography, color-model, color-engine}.md; MOTION_SERVICE_RFC
stays in src (foreign/concurrent). docs:check 0 errors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
TYPOGRAPHY_ENGINE_RFC (191 L) -> rfcs/rfc-typography.md and COLOR_MODEL_RFC
(155 L) -> rfcs/rfc-color-model.md, Spanish to English. color-model is a
RESOLVED RFC kept as the historical record of the rejected single-anchor
proposal — the resolution header travels with it, canonical model stays
theming reference s25. typography's stub notes the measurement arm
(lib/canvas-text, SText) stays in src. decisions.md rows repointed;
typography's Status aligned with the RFC's own closed-phases state.
Theming reference s21/s25 links swept. docs:check 0 errors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
DEPTH_ENGINE_RFC (149 L) -> rfcs/rfc-depth.md and SHAPE_ENGINE_RFC (162 L)
-> rfcs/rfc-shape.md, Spanish to English. Both stubs summarize the shipped
state (phases 1-5) and keep the by-path citations resolving — including
the '// (DEPTH_ENGINE_RFC s5)' comments across src/uix/eidos/lib/*.ts that
decisions.md documents as load-bearing. decisions.md index rows repointed
and their Status column aligned with the status the RFCs themselves
declare (both said implemented; the rows still said Propuesta). Theming
reference s29/s30 canonical-guide links swept. docs:check 0 errors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
SCALING_RFC (92 L) -> rfcs/rfc-scaling.md and STRUCTURE_ENGINE_RFC (113 L)
-> rfcs/rfc-structure.md, Spanish to English, same section numbering. The
legacy-name -> rfc-* rename that decisions.md declared deferred 'until
those citations are swept in the same pass' happens here: stubs at the old
paths keep every by-path citation resolving. Corpus links swept (theming
reference s23/s31, decisions.md index rows, channels.md capstone list —
the channels.md sibling links to rfc-color-engine/typography/depth/shape
land later in this batch). Chronicle docs (theming/changelog,
inherit_audit) keep old paths by doctrine — the stubs resolve them.
docs:check 0 errors.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
dev