@fusion.dev/helix-headless 0.53.0 → 0.53.2

package/custom-elements.json

@@ -705,7 +705,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "Holds the mode and nothing else. The item that was clicked is what\ncoordinates: it walks up to the nearest accordion, reads the mode off it, and\ncloses its peers itself.\n\nCoordination is structural, not a subscription. `helix-before-open` does not bubble\nand is not composed, because overlays nest and an inner one's lifecycle must\nnever read as the outer one's, so an accordion cannot learn that a child\nopened by listening for it.\n\nMembership resolves on demand, so there is no registry to keep in step: an\nitem added a second after render coordinates on its first click.\n\nItems are matched with `instanceof HelixHeadlessAccordionItem`, never by tag\nname. Matching `helix-accordion-item` would tie this tier to the styled one.\n\nThe keyboard listener is the controller's and is attached on connect, so\nnone of it runs on the server, where there is no tree to walk and no events\nto hear.",
+          "description": "Holds the mode and nothing else. The item that was clicked is what\ncoordinates: it walks up to the nearest accordion, reads the mode off it, and\ncloses its peers itself.\n\nCoordination is structural, not a subscription. `helix-before-open` does not bubble\nand is not composed, because overlays nest and an inner one's lifecycle must\nnever read as the outer one's, so an accordion cannot learn that a child\nopened by listening for it.\n\nMembership resolves on demand, so there is no registry to keep in step: an\nitem added a second after render coordinates on its first click.\n\nItems are matched with `instanceof HelixHeadlessAccordionItem`, never by tag\nname, so this tier does not learn about the styled one.\n\nThe keyboard listener is the controller's and is attached on connect, so\nnone of it runs on the server, where there is no tree to walk and no events\nto hear.",
           "name": "HelixHeadlessAccordion",
           "slots": [
             {
@@ -3271,7 +3271,7 @@
             "text": "number"
           },
           "default": "12",
-          "description": "The largest `max` the styled tier can hide past.\n\nThe fold is `:host([max=\"N\"]) ::slotted(:nth-child(n + N+1))`, one selector\nper value, because `An+B` takes a literal and no custom property can be read\ninto it. So the range is finite and this is where it is written down, in the\ntier that has to agree with it: a count that folds and a stylesheet that does\nnot would put a \"+5\" beside five avatars that are still on screen."
+          "description": "The largest `max` a stylesheet can hide past.\n\nThe fold is `:host([max=\"N\"]) ::slotted(:nth-child(n + N+1))`, one selector\nper value, because `An+B` takes a literal and no custom property can be read\ninto it. So the range is finite, and it is written down beside the count that\nhas to agree with it: a count that folds and a stylesheet that does not would\nput a \"+5\" beside five avatars that are still on screen."
         },
         {
           "kind": "class",
@@ -3306,7 +3306,7 @@
               "type": {
                 "text": "number | undefined"
               },
-              "description": "How many avatars are shown before the rest become a count.\n\nUnset shows all of them, which is what no limit means and is more useful\nthan refusing to render.\n\n**Reflected, because the fold is a stylesheet rule and nothing else.** The\nstyled tier hides past the limit with `:host([max=\"N\"]) ::slotted(:nth-child(n + N+1))`,\nso the attribute *is* the mechanism: a `max` that reaches this property and\nnever reaches the host matches no rule, and every avatar is drawn beside a\ncount claiming some are hidden. Markup is the safe path and hides the fault;\na script is the broken one, and React takes it for everyone, assigning a\nmatching property rather than an attribute.\n\n**Optional rather than defaulted, per the rule that a reflected property\ntakes no default**: `max = 0` would stamp `max=\"0\"` onto every host that\nnever asked for a limit, including on the server.",
+              "description": "How many avatars are shown before the rest become a count.\n\nUnset shows all of them, which is what no limit means and is more useful\nthan refusing to render.\n\n**Reflected, because the fold is a stylesheet rule and nothing else.** A\nsheet hides past the limit one selector per value, as `FOLDS_UP_TO`\ndescribes, so the attribute *is* the mechanism: a `max` that reaches this\nproperty and never reaches the host matches no rule, and every avatar is\ndrawn beside a count claiming some are hidden. Markup is the safe path and\nhides the fault; a script is the broken one, and React takes it for\neveryone, assigning a matching property rather than an attribute.\n\n**Optional rather than defaulted, per the rule that a reflected property\ntakes no default**: `max = 0` would stamp `max=\"0\"` onto every host that\nnever asked for a limit, including on the server.",
               "attribute": "max",
               "reflects": true
             },
@@ -3678,7 +3678,7 @@
               "type": {
                 "text": "number | undefined"
               },
-              "description": "How many avatars are shown before the rest become a count.\n\nUnset shows all of them, which is what no limit means and is more useful\nthan refusing to render.\n\n**Reflected, because the fold is a stylesheet rule and nothing else.** The\nstyled tier hides past the limit with `:host([max=\"N\"]) ::slotted(:nth-child(n + N+1))`,\nso the attribute *is* the mechanism: a `max` that reaches this property and\nnever reaches the host matches no rule, and every avatar is drawn beside a\ncount claiming some are hidden. Markup is the safe path and hides the fault;\na script is the broken one, and React takes it for everyone, assigning a\nmatching property rather than an attribute.\n\n**Optional rather than defaulted, per the rule that a reflected property\ntakes no default**: `max = 0` would stamp `max=\"0\"` onto every host that\nnever asked for a limit, including on the server.",
+              "description": "How many avatars are shown before the rest become a count.\n\nUnset shows all of them, which is what no limit means and is more useful\nthan refusing to render.\n\n**Reflected, because the fold is a stylesheet rule and nothing else.** A\nsheet hides past the limit one selector per value, as `FOLDS_UP_TO`\ndescribes, so the attribute *is* the mechanism: a `max` that reaches this\nproperty and never reaches the host matches no rule, and every avatar is\ndrawn beside a count claiming some are hidden. Markup is the safe path and\nhides the fault; a script is the broken one, and React takes it for\neveryone, assigning a matching property rather than an attribute.\n\n**Optional rather than defaulted, per the rule that a reflected property\ntakes no default**: `max = 0` would stamp `max=\"0\"` onto every host that\nnever asked for a limit, including on the server.",
               "fieldName": "max"
             }
           ],
@@ -4839,7 +4839,7 @@
               "name": "--bar-chart-axis"
             },
             {
-              "description": "The color of the secondary axis when one series owns it. It resolves a `--helix-color-series-*` token, so on the headless tier alone it names something that is not there. #138.",
+              "description": "The color of the secondary axis, where one series owns it. Written by the component from that series' `--chart-series-*`, so it promises the series color and nothing more: whatever that hook resolves to is what the axis is drawn in.",
               "name": "--bar-chart-secondary"
             },
             {
@@ -6398,7 +6398,7 @@
               "type": {
                 "text": "string | undefined"
               },
-              "description": "Where the crumb goes, and nothing where it is the current page.\n\n**Optional and reflected, because a stylesheet does select on it.**\n`breadcrumb.css` writes `:host(:not([href])) [part=\"link\"]` to take the\nblock padding off a crumb that is not a link, so the attribute is what\ndecides how tall a crumb is. Unreflected, an `href` set from script reached\nthe property, rendered the anchor, and left the host looking unlinked: the\ncrumb drew at 20px where a markup twin drew at 28.\n\nOptional rather than defaulted, which is the general rule that a reflected\nproperty takes no default: an `href` written onto every host would be a link\nto the current page on every crumb nobody asked to be one.\n\nReading cannot catch this and the picture comparison can. React assigns a\nmatching property instead of an attribute, so every crumb in the React twin\nof `breadcrumbs/links` lost its padding at once.",
+              "description": "Where the crumb goes, and nothing where it is the current page.\n\n**Optional and reflected, because a stylesheet does select on it.** A sheet\nwrites `:host(:not([href])) [part=\"link\"]` to take the block padding off a\ncrumb that is not a link, so the attribute is what decides how tall a crumb\nis. Unreflected, an `href` set from script reached the property, rendered\nthe anchor, and left the host looking unlinked: the crumb drew at 20px\nwhere a markup twin drew at 28.\n\nOptional rather than defaulted, which is the general rule that a reflected\nproperty takes no default: an `href` written onto every host would be a link\nto the current page on every crumb nobody asked to be one.\n\nReading cannot catch this and the picture comparison can. React assigns a\nmatching property instead of an attribute, so every crumb in the React twin\nof `breadcrumbs/links` lost its padding at once.",
               "attribute": "href",
               "reflects": true
             },
@@ -6773,7 +6773,7 @@
               "type": {
                 "text": "string | undefined"
               },
-              "description": "Where the crumb goes, and nothing where it is the current page.\n\n**Optional and reflected, because a stylesheet does select on it.**\n`breadcrumb.css` writes `:host(:not([href])) [part=\"link\"]` to take the\nblock padding off a crumb that is not a link, so the attribute is what\ndecides how tall a crumb is. Unreflected, an `href` set from script reached\nthe property, rendered the anchor, and left the host looking unlinked: the\ncrumb drew at 20px where a markup twin drew at 28.\n\nOptional rather than defaulted, which is the general rule that a reflected\nproperty takes no default: an `href` written onto every host would be a link\nto the current page on every crumb nobody asked to be one.\n\nReading cannot catch this and the picture comparison can. React assigns a\nmatching property instead of an attribute, so every crumb in the React twin\nof `breadcrumbs/links` lost its padding at once.",
+              "description": "Where the crumb goes, and nothing where it is the current page.\n\n**Optional and reflected, because a stylesheet does select on it.** A sheet\nwrites `:host(:not([href])) [part=\"link\"]` to take the block padding off a\ncrumb that is not a link, so the attribute is what decides how tall a crumb\nis. Unreflected, an `href` set from script reached the property, rendered\nthe anchor, and left the host looking unlinked: the crumb drew at 20px\nwhere a markup twin drew at 28.\n\nOptional rather than defaulted, which is the general rule that a reflected\nproperty takes no default: an `href` written onto every host would be a link\nto the current page on every crumb nobody asked to be one.\n\nReading cannot catch this and the picture comparison can. React assigns a\nmatching property instead of an attribute, so every crumb in the React twin\nof `breadcrumbs/links` lost its padding at once.",
               "fieldName": "href"
             },
             {
@@ -7791,7 +7791,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
+              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-headless-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-headless-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
               "attribute": "command",
               "helix": {
                 "references": [
@@ -8481,7 +8481,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
+              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-headless-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-headless-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
               "fieldName": "command"
             },
             {
@@ -9153,7 +9153,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "Regions are slots rather than separate elements, because a header, body, and\nfooter carry no state, no behavior, and no API of their own. Slotting them\ncosts one custom element instead of five and lets the consumer bring their\nown markup, so a title can be a real `<h2>` rather than an emulated one.\n\nThe host is the surface. A wrapper `<div part=\"container\">` would be a second\nbox doing nothing the host cannot: the element itself takes the border, the\nradius, and the grid, and a consumer styles `helix-card` rather than reaching\nthrough `::part`. One node instead of two, on every card on the page.\n\nName the card with `aria-label` or `aria-labelledby`. That is the only\nnaming that survives server rendering: reading a title out of the light DOM\nneeds a DOM, and the server has none.",
+          "description": "Regions are slots rather than separate elements, because a header, body, and\nfooter carry no state, no behavior, and no API of their own. Slotting them\ncosts one custom element instead of five and lets the consumer bring their\nown markup, so a title can be a real `<h2>` rather than an emulated one.\n\nThe host is the surface. A wrapper `<div part=\"container\">` would be a second\nbox doing nothing the host cannot: the element itself takes the border, the\nradius, and the grid, and a consumer styles `Card` rather than reaching\nthrough `::part`. One node instead of two, on every card on the page.\n\nName the card with `aria-label` or `aria-labelledby`. That is the only\nnaming that survives server rendering: reading a title out of the light DOM\nneeds a DOM, and the server has none.",
           "name": "HelixHeadlessCard",
           "slots": [
             {
@@ -9814,7 +9814,7 @@
               "type": {
                 "text": "number | 'auto' | undefined"
               },
-              "description": "How many slides are shown across the track at once. Absent shows one.\n\n**Here rather than on the styled tier, because it decides how many places\nthe track can stop.** From the stylesheet it looks like a width, one rule\nper value setting a custom property that the slides divide the track by.\nBut a carousel showing four of six slides has three positions, not six, and\neverything that counts positions is in this file. Declared up there, `goTo`\nclamps to the last *slide*, so three of the six indicators point at an\noffset the track has already reached: pressing next walks the dots to the\nend while the slides stop moving a third of the way through.\n\n**The number is published to the styles rather than enumerated by them**,\nas `--carousel-per-view` on the track, which is what keeps a width and a\ncount from disagreeing. A sheet can only name a value it wrote a rule for,\nso the styled tier enumerated `per-view=\"2\"` through `5` and called five a\njudgment; above it the pane fell back to one across while this file went on\ncounting positions. Measured in Chromium over eight slides, six across gave\nthree dots, a pane taking the whole track and five slides reachable only by\ndragging. Written from here there is no ceiling to exceed, and a consumer's\nown sheet reads the same declaration the styled one does. `runEnd` publishes\nits correction the same way and for the same reason.\n\n**On the track and not on the host.** A custom property on the host is a\nreflected attribute by another name: it would arrive on upgrade, and a\nconsumer's server markup never had it. Inside the component's own shadow\nroot it serializes with the rest of the template on the server, and the\ntrack's own declaration beats anything a consumer inherits in, so the\nattribute is the one answer to how wide a pane is.\n\nA value the track cannot honor is taken to the nearest one it can, by\n`across`, and both tiers read that rather than this.\n\n**`auto` fits as many as the track can hold**, at `--carousel-slide-min`\napiece, and nothing is written onto the track: the count is the sheet's,\nand the sheet's alone, because it is the one number here CSS can work out\nfor itself. Everything it takes is a length the sheet already has, so the\npanes are laid out at their final width by the first rule that runs, on a\nserver's markup as on the client's, and no script is waited on. That is the\nwhole reason a declared count was ever the only option: a carousel that\nasks how wide its container is can only ask on the client, so the panes go\ndown at one width and move to another when the answer arrives. Written\ninline the count would outrank the sheet that computes it, which is why an\n`auto` track is handed nothing. A count stays for the design that calls for\nexactly three.\n\nThe styled sheet says how it divides, and the headless `auto` example\ncarries the same three lines for a consumer writing their own.\n\n**A value rather than the resting behavior**, which is what the\nresponsiveness policy asks of every axis that a box decides: a splitter\nturns on `orientation=\"auto\"` and a plain one does not turn at all. Fitting\nchanges what the carousel *does*, since how many are across is how many\nplaces the track can stop, so a carousel that had not asked for it would\nfind its position count moving with the window.\n\nNo default, because it reflects. A reflected property with a value writes\nits attribute on upgrade, and the host then carries something a consumer's\nserver markup never had, which React reports as a mismatch. The one-across\ncase is the absent attribute, and the styles read it that way.",
+              "description": "How many slides are shown across the track at once. Absent shows one.\n\n**Here rather than on the styled tier, because it decides how many places\nthe track can stop.** From the stylesheet it looks like a width, one rule\nper value setting a custom property that the slides divide the track by.\nBut a carousel showing four of six slides has three positions, not six, and\neverything that counts positions is in this file. Declared up there, `goTo`\nclamps to the last *slide*, so three of the six indicators point at an\noffset the track has already reached: pressing next walks the dots to the\nend while the slides stop moving a third of the way through.\n\n**The number is published to the styles rather than enumerated by them**,\nas `--carousel-per-view` on the track, which is what keeps a width and a\ncount from disagreeing. A sheet can only name a value it wrote a rule for,\nso a sheet enumerating `per-view=\"2\"` through `5` is a ceiling of five, and\nabove it the pane falls back to one across while this file goes on counting\npositions. Measured in Chromium over eight slides against such a sheet, six\nacross gave three dots, a pane taking the whole track and five slides\nreachable only by dragging. Published from here there is no ceiling to\nexceed, and every sheet reads the same declaration. `runEnd` publishes its\ncorrection the same way and for the same reason.\n\n**On the track and not on the host.** A custom property on the host is a\nreflected attribute by another name: it would arrive on upgrade, and a\nconsumer's server markup never had it. Inside the component's own shadow\nroot it serializes with the rest of the template on the server, and the\ntrack's own declaration beats anything a consumer inherits in, so the\nattribute is the one answer to how wide a pane is.\n\nA value the track cannot honor is taken to the nearest one it can, by\n`across`, and both tiers read that rather than this.\n\n**`auto` fits as many as the track can hold**, at `--carousel-slide-min`\napiece, and nothing is written onto the track: the count is the sheet's,\nand the sheet's alone, because it is the one number here CSS can work out\nfor itself. Everything it takes is a length the sheet already has, so the\npanes are laid out at their final width by the first rule that runs, on a\nserver's markup as on the client's, and no script is waited on. That is the\nwhole reason a declared count was ever the only option: a carousel that\nasks how wide its container is can only ask on the client, so the panes go\ndown at one width and move to another when the answer arrives. Written\ninline the count would outrank the sheet that computes it, which is why an\n`auto` track is handed nothing. A count stays for the design that calls for\nexactly three.\n\nA sheet says how it divides, and the `auto` example here carries the same\nthree lines for a consumer writing their own.\n\n**A value rather than the resting behavior**, which is what the\nresponsiveness policy asks of every axis that a box decides: a splitter\nturns on `orientation=\"auto\"` and a plain one does not turn at all. Fitting\nchanges what the carousel *does*, since how many are across is how many\nplaces the track can stop, so a carousel that had not asked for it would\nfind its position count moving with the window.\n\nNo default, because it reflects. A reflected property with a value writes\nits attribute on upgrade, and the host then carries something a consumer's\nserver markup never had, which React reports as a mismatch. The one-across\ncase is the absent attribute, and the styles read it that way.",
               "attribute": "per-view",
               "reflects": true
             },
@@ -10170,7 +10170,7 @@
                   "text": "string | typeof nothing"
                 }
               },
-              "description": "How far short of a whole pixel the run of slides ends, for the last one.\n\n**The run has to end on a whole pixel, because a right-to-left track cannot\nreach a fraction.** Three across 692 makes the run 450.66 longer than the\ntrack. Chromium measures a right-to-left scroller from its far end and\nfloors that length, so the track stops at -450 where a left-to-right one\nrounds to 451, and two thirds of a pixel of the slide before stays at the\nleading edge. Ending on a whole pixel, the far end is one number both ways\nand the last place rests on it in both, so a swipe to the end stops there\ntoo.\n\n**Here rather than in a stylesheet, because the count is the component's.**\nThe arithmetic wants how many slides there are, and CSS can only ask\n`sibling-count()`, which is Chromium-only, evaluates to 0 through\n`::part()`, and needs a `:last-child` that `::part()` cannot carry either.\nSo the styled tier could correct its run and a consumer writing their own\nseveral-across sheet could not: measured in Chromium at three across, the\nheadless tier rested at 451 against -450 while the styled tier rested at\none number both ways. Published from here, one declaration reading this\ngets both tiers and any consumer to the same place.\n\n**WebKit takes the margin and does not need it**, which is the reason a\ncorrection written in CSS looked like enough. It has no `sibling-count()`\nat all, so nothing reached that engine, and its track rested on one number\nboth ways anyway. Measured with the margin in place, the same widths:\nevery resting place is the number it was without it.\n\n**On the last slide alone, which is what makes that one declaration\nenough.** A sheet writes `margin-inline-end: var(--carousel-run-end, 0px)`\nacross every pane and the fallback answers on all but the last, so nothing\nhas to select a slide by its position in the run.\n\n**A sheet says how wide a pane is and how far apart they sit, and this\nreads both back.** Set neither and the correction is invalid at\ncomputed-value time, which is the fallback rather than a broken margin. The\npercentage resolves where the value is used, on the last pane, so it is the\ntrack's content box and not the track's own containing block: the width the\nrun is actually measured against.\n\nHorizontal only. A vertical track scrolls the other axis, `scrollTop`\ncounts up from zero whatever the writing mode, and there is no far end to\nfloor."
+              "description": "How far short of a whole pixel the run of slides ends, for the last one.\n\n**The run has to end on a whole pixel, because a right-to-left track cannot\nreach a fraction.** Three across 692 makes the run 450.66 longer than the\ntrack. Chromium measures a right-to-left scroller from its far end and\nfloors that length, so the track stops at -450 where a left-to-right one\nrounds to 451, and two thirds of a pixel of the slide before stays at the\nleading edge. Ending on a whole pixel, the far end is one number both ways\nand the last place rests on it in both, so a swipe to the end stops there\ntoo.\n\n**Here rather than in a stylesheet, because the count is the component's.**\nThe arithmetic wants how many slides there are, and CSS can only ask\n`sibling-count()`, which is Chromium-only, evaluates to 0 through\n`::part()`, and needs a `:last-child` that `::part()` cannot carry either.\nSo a sheet shipped beside the component could correct its own run and a\nconsumer writing their own several-across sheet could not: measured in\nChromium at three across, an uncorrected track rested at 451 against -450\nwhere a corrected one rested on one number both ways. Published from here,\none declaration reading this gets every sheet to the same place.\n\n**WebKit takes the margin and does not need it**, which is the reason a\ncorrection written in CSS looked like enough. It has no `sibling-count()`\nat all, so nothing reached that engine, and its track rested on one number\nboth ways anyway. Measured with the margin in place, the same widths:\nevery resting place is the number it was without it.\n\n**On the last slide alone, which is what makes that one declaration\nenough.** A sheet writes `margin-inline-end: var(--carousel-run-end, 0px)`\nacross every pane and the fallback answers on all but the last, so nothing\nhas to select a slide by its position in the run.\n\n**A sheet says how wide a pane is and how far apart they sit, and this\nreads both back.** Set neither and the correction is invalid at\ncomputed-value time, which is the fallback rather than a broken margin. The\npercentage resolves where the value is used, on the last pane, so it is the\ntrack's content box and not the track's own containing block: the width the\nrun is actually measured against.\n\nHorizontal only. A vertical track scrolls the other axis, `scrollTop`\ncounts up from zero whatever the writing mode, and there is no far end to\nfloor."
             },
             {
               "kind": "field",
@@ -10582,7 +10582,7 @@
               "type": {
                 "text": "number | 'auto' | undefined"
               },
-              "description": "How many slides are shown across the track at once. Absent shows one.\n\n**Here rather than on the styled tier, because it decides how many places\nthe track can stop.** From the stylesheet it looks like a width, one rule\nper value setting a custom property that the slides divide the track by.\nBut a carousel showing four of six slides has three positions, not six, and\neverything that counts positions is in this file. Declared up there, `goTo`\nclamps to the last *slide*, so three of the six indicators point at an\noffset the track has already reached: pressing next walks the dots to the\nend while the slides stop moving a third of the way through.\n\n**The number is published to the styles rather than enumerated by them**,\nas `--carousel-per-view` on the track, which is what keeps a width and a\ncount from disagreeing. A sheet can only name a value it wrote a rule for,\nso the styled tier enumerated `per-view=\"2\"` through `5` and called five a\njudgment; above it the pane fell back to one across while this file went on\ncounting positions. Measured in Chromium over eight slides, six across gave\nthree dots, a pane taking the whole track and five slides reachable only by\ndragging. Written from here there is no ceiling to exceed, and a consumer's\nown sheet reads the same declaration the styled one does. `runEnd` publishes\nits correction the same way and for the same reason.\n\n**On the track and not on the host.** A custom property on the host is a\nreflected attribute by another name: it would arrive on upgrade, and a\nconsumer's server markup never had it. Inside the component's own shadow\nroot it serializes with the rest of the template on the server, and the\ntrack's own declaration beats anything a consumer inherits in, so the\nattribute is the one answer to how wide a pane is.\n\nA value the track cannot honor is taken to the nearest one it can, by\n`across`, and both tiers read that rather than this.\n\n**`auto` fits as many as the track can hold**, at `--carousel-slide-min`\napiece, and nothing is written onto the track: the count is the sheet's,\nand the sheet's alone, because it is the one number here CSS can work out\nfor itself. Everything it takes is a length the sheet already has, so the\npanes are laid out at their final width by the first rule that runs, on a\nserver's markup as on the client's, and no script is waited on. That is the\nwhole reason a declared count was ever the only option: a carousel that\nasks how wide its container is can only ask on the client, so the panes go\ndown at one width and move to another when the answer arrives. Written\ninline the count would outrank the sheet that computes it, which is why an\n`auto` track is handed nothing. A count stays for the design that calls for\nexactly three.\n\nThe styled sheet says how it divides, and the headless `auto` example\ncarries the same three lines for a consumer writing their own.\n\n**A value rather than the resting behavior**, which is what the\nresponsiveness policy asks of every axis that a box decides: a splitter\nturns on `orientation=\"auto\"` and a plain one does not turn at all. Fitting\nchanges what the carousel *does*, since how many are across is how many\nplaces the track can stop, so a carousel that had not asked for it would\nfind its position count moving with the window.\n\nNo default, because it reflects. A reflected property with a value writes\nits attribute on upgrade, and the host then carries something a consumer's\nserver markup never had, which React reports as a mismatch. The one-across\ncase is the absent attribute, and the styles read it that way.",
+              "description": "How many slides are shown across the track at once. Absent shows one.\n\n**Here rather than on the styled tier, because it decides how many places\nthe track can stop.** From the stylesheet it looks like a width, one rule\nper value setting a custom property that the slides divide the track by.\nBut a carousel showing four of six slides has three positions, not six, and\neverything that counts positions is in this file. Declared up there, `goTo`\nclamps to the last *slide*, so three of the six indicators point at an\noffset the track has already reached: pressing next walks the dots to the\nend while the slides stop moving a third of the way through.\n\n**The number is published to the styles rather than enumerated by them**,\nas `--carousel-per-view` on the track, which is what keeps a width and a\ncount from disagreeing. A sheet can only name a value it wrote a rule for,\nso a sheet enumerating `per-view=\"2\"` through `5` is a ceiling of five, and\nabove it the pane falls back to one across while this file goes on counting\npositions. Measured in Chromium over eight slides against such a sheet, six\nacross gave three dots, a pane taking the whole track and five slides\nreachable only by dragging. Published from here there is no ceiling to\nexceed, and every sheet reads the same declaration. `runEnd` publishes its\ncorrection the same way and for the same reason.\n\n**On the track and not on the host.** A custom property on the host is a\nreflected attribute by another name: it would arrive on upgrade, and a\nconsumer's server markup never had it. Inside the component's own shadow\nroot it serializes with the rest of the template on the server, and the\ntrack's own declaration beats anything a consumer inherits in, so the\nattribute is the one answer to how wide a pane is.\n\nA value the track cannot honor is taken to the nearest one it can, by\n`across`, and both tiers read that rather than this.\n\n**`auto` fits as many as the track can hold**, at `--carousel-slide-min`\napiece, and nothing is written onto the track: the count is the sheet's,\nand the sheet's alone, because it is the one number here CSS can work out\nfor itself. Everything it takes is a length the sheet already has, so the\npanes are laid out at their final width by the first rule that runs, on a\nserver's markup as on the client's, and no script is waited on. That is the\nwhole reason a declared count was ever the only option: a carousel that\nasks how wide its container is can only ask on the client, so the panes go\ndown at one width and move to another when the answer arrives. Written\ninline the count would outrank the sheet that computes it, which is why an\n`auto` track is handed nothing. A count stays for the design that calls for\nexactly three.\n\nA sheet says how it divides, and the `auto` example here carries the same\nthree lines for a consumer writing their own.\n\n**A value rather than the resting behavior**, which is what the\nresponsiveness policy asks of every axis that a box decides: a splitter\nturns on `orientation=\"auto\"` and a plain one does not turn at all. Fitting\nchanges what the carousel *does*, since how many are across is how many\nplaces the track can stop, so a carousel that had not asked for it would\nfind its position count moving with the window.\n\nNo default, because it reflects. A reflected property with a value writes\nits attribute on upgrade, and the host then carries something a consumer's\nserver markup never had, which React reports as a mismatch. The one-across\ncase is the absent attribute, and the styles read it that way.",
               "fieldName": "perView"
             }
           ],
@@ -10834,7 +10834,7 @@
               "name": "handleInput",
               "privacy": "private",
               "readonly": true,
-              "description": "Records the draft, and stops the textarea speaking in this element's name.\n\n**The stop is not incidental.** `input` is composed by specification, so the\n`<textarea>` in this shadow root sent its own past this host and on through\nwhatever `<helix-chat>` it is slotted into, and neither element declares an\n`input`: a consumer binding one on either received an undeclared event whose\ntarget was a control they cannot reach. What a composer offers is\n`helix-send`, once, when the reader is done."
+              "description": "Records the draft, and stops the textarea speaking in this element's name.\n\n**The stop is not incidental.** `input` is composed by specification, so\nthe `<textarea>` in this shadow root sent its own past this host and on\nthrough whatever `<helix-headless-chat>` it is slotted into, and neither\nelement declares an `input`: a consumer binding one on either received an\nundeclared event whose target was a control they cannot reach. What a\ncomposer offers is `helix-send`, once, when the reader is done."
             },
             {
               "kind": "field",
@@ -12189,7 +12189,7 @@
           ],
           "slots": [
             {
-              "description": "The turns, as `helix-chat-message` children.",
+              "description": "The turns.",
               "name": ""
             },
             {
@@ -12281,7 +12281,7 @@
               "type": {
                 "text": "boolean"
               },
-              "description": "Whether the transcript follows new turns to the bottom.\n\nThe reader's scrolling asks first: `helix-before-pin` is cancelable, so a\nconsumer refusing it takes the decision and assigns their own value.\n`default-pinned` seeds the state this element then owns, and it defaults to\npinned because a conversation opens at its most recent turn.\n\n**Not reflected, and the default is why.** A reflected property that\ndefaults to true writes its attribute on upgrade whether or not the\nconsumer wrote one, so the server serves `<helix-chat label=\"x\">` and the\nclient turns it into `<helix-chat label=\"x\" pinned>`: a hydration mismatch a\nReact consumer reports and does not patch up. Nothing needs the attribute\neither, because the one thing that changes with the pin is the way back, and\nthat is a part inside the shadow tree where a stylesheet can already reach\nit. Measured rather than assumed: `label, pinned` after upgrade against\n`label` from the server.",
+              "description": "Whether the transcript follows new turns to the bottom.\n\nThe reader's scrolling asks first: `helix-before-pin` is cancelable, so a\nconsumer refusing it takes the decision and assigns their own value.\n`default-pinned` seeds the state this element then owns, and it defaults to\npinned because a conversation opens at its most recent turn.\n\n**Not reflected, and the default is why.** A reflected property that\ndefaults to true writes its attribute on upgrade whether or not the\nconsumer wrote one, so the server serves `<helix-headless-chat label=\"x\">`\nand the client turns it into `<helix-headless-chat label=\"x\" pinned>`: a\nhydration mismatch a React consumer reports and does not patch up. Nothing\nneeds the attribute either, because the one thing that changes with the pin\nis the way back, and that is a part inside the shadow tree where a\nstylesheet can already reach it. Measured rather than assumed: `label,\npinned` after upgrade against `label` from the server.",
               "attribute": "pinned"
             },
             {
@@ -13329,7 +13329,7 @@
               "type": {
                 "text": "boolean"
               },
-              "description": "Whether the transcript follows new turns to the bottom.\n\nThe reader's scrolling asks first: `helix-before-pin` is cancelable, so a\nconsumer refusing it takes the decision and assigns their own value.\n`default-pinned` seeds the state this element then owns, and it defaults to\npinned because a conversation opens at its most recent turn.\n\n**Not reflected, and the default is why.** A reflected property that\ndefaults to true writes its attribute on upgrade whether or not the\nconsumer wrote one, so the server serves `<helix-chat label=\"x\">` and the\nclient turns it into `<helix-chat label=\"x\" pinned>`: a hydration mismatch a\nReact consumer reports and does not patch up. Nothing needs the attribute\neither, because the one thing that changes with the pin is the way back, and\nthat is a part inside the shadow tree where a stylesheet can already reach\nit. Measured rather than assumed: `label, pinned` after upgrade against\n`label` from the server.",
+              "description": "Whether the transcript follows new turns to the bottom.\n\nThe reader's scrolling asks first: `helix-before-pin` is cancelable, so a\nconsumer refusing it takes the decision and assigns their own value.\n`default-pinned` seeds the state this element then owns, and it defaults to\npinned because a conversation opens at its most recent turn.\n\n**Not reflected, and the default is why.** A reflected property that\ndefaults to true writes its attribute on upgrade whether or not the\nconsumer wrote one, so the server serves `<helix-headless-chat label=\"x\">`\nand the client turns it into `<helix-headless-chat label=\"x\" pinned>`: a\nhydration mismatch a React consumer reports and does not patch up. Nothing\nneeds the attribute either, because the one thing that changes with the pin\nis the way back, and that is a part inside the shadow tree where a\nstylesheet can already reach it. Measured rather than assumed: `label,\npinned` after upgrade against `label` from the server.",
               "fieldName": "pinned"
             },
             {
@@ -14641,7 +14641,7 @@
               "name": "handleLabelChange",
               "privacy": "private",
               "readonly": true,
-              "description": "Whether there is a label to leave room for, written to the host.\n\nThe row is a flex line with a gap in it, so a label span rendered empty is\nstill a box and still takes that gap: an unlabelled box came out with eight\npixels of nothing after it. The input answers this by not rendering the\nelement, which cannot work here, because a checkbox is named by its own\ncontent and the property is only that slot's fallback. Gating on the\nproperty would unlabel every `<helix-checkbox>Text</helix-checkbox>` ever\nwritten.\n\nSo the slot is asked instead, the same way `Tooltip` asks its own: with\n`flatten`, because here the fallback counts, being where the property\nrenders. Client only, so a server-rendered unlabelled checkbox carries the\ngap for one frame. That is a control with no accessible name, which has a\nworse problem than its spacing."
+              "description": "Whether there is a label to leave room for, written to the host.\n\nThe row is a flex line with a gap in it, so a label span rendered empty is\nstill a box and still takes that gap: an unlabelled box came out with eight\npixels of nothing after it. The input answers this by not rendering the\nelement, which cannot work here, because a checkbox is named by its own\ncontent and the property is only that slot's fallback. Gating on the\nproperty would unlabel every\n`<helix-headless-checkbox>Text</helix-headless-checkbox>` ever written.\n\nSo the slot is asked instead, the same way `Tooltip` asks its own: with\n`flatten`, because here the fallback counts, being where the property\nrenders. Client only, so a server-rendered unlabelled checkbox carries the\ngap for one frame. That is a control with no accessible name, which has a\nworse problem than its spacing."
             },
             {
               "kind": "method",
@@ -15980,7 +15980,7 @@
                 "text": "boolean"
               },
               "default": "false",
-              "description": "Whether to leave the copy button off.\n\nNegative, which is not a style choice. A boolean attribute is true whenever\nit is *present*, so a `copyable` defaulting to true could never be turned\noff from markup: `copyable=\"false\"` is the string \"false\", which is\npresent, which is true. Measured, and it is why this is named the other way\nround. The platform already writes booleans like this for the same reason,\n`novalidate` and `nomodule` and `noreferrer` among them.\n\n**Reflected, because the gutter is a stylesheet rule.** `code-block.css`\nreserves the copy button's width with `:host(:not([no-copy]))`, so the\nattribute is what decides whether the block pays for a control. Unreflected,\nsetting it from script takes the button away (`render` reads the property)\nand leaves the 40px it stood in reserved.\n\nA false boolean reflects as the attribute's absence, so this costs no\n`no-copy=\"\"` on hosts that never asked for one, and the rule about a\nreflected property taking no default is satisfied by the default being\n`false`.",
+              "description": "Whether to leave the copy button off.\n\nNegative, which is not a style choice. A boolean attribute is true whenever\nit is *present*, so a `copyable` defaulting to true could never be turned\noff from markup: `copyable=\"false\"` is the string \"false\", which is\npresent, which is true. Measured, and it is why this is named the other way\nround. The platform already writes booleans like this for the same reason,\n`novalidate` and `nomodule` and `noreferrer` among them.\n\n**Reflected, because the gutter is a stylesheet rule.** A sheet reserves\nthe copy button's width with `:host(:not([no-copy]))`, so the attribute is\nwhat decides whether the block pays for a control. Unreflected, setting it\nfrom script takes the button away (`render` reads the property) and leaves\nthe 40px it stood in reserved.\n\nA false boolean reflects as the attribute's absence, so this costs no\n`no-copy=\"\"` on hosts that never asked for one, and the rule about a\nreflected property taking no default is satisfied by the default being\n`false`.",
               "attribute": "no-copy",
               "reflects": true
             },
@@ -16561,7 +16561,7 @@
                 "text": "boolean"
               },
               "default": "false",
-              "description": "Whether to leave the copy button off.\n\nNegative, which is not a style choice. A boolean attribute is true whenever\nit is *present*, so a `copyable` defaulting to true could never be turned\noff from markup: `copyable=\"false\"` is the string \"false\", which is\npresent, which is true. Measured, and it is why this is named the other way\nround. The platform already writes booleans like this for the same reason,\n`novalidate` and `nomodule` and `noreferrer` among them.\n\n**Reflected, because the gutter is a stylesheet rule.** `code-block.css`\nreserves the copy button's width with `:host(:not([no-copy]))`, so the\nattribute is what decides whether the block pays for a control. Unreflected,\nsetting it from script takes the button away (`render` reads the property)\nand leaves the 40px it stood in reserved.\n\nA false boolean reflects as the attribute's absence, so this costs no\n`no-copy=\"\"` on hosts that never asked for one, and the rule about a\nreflected property taking no default is satisfied by the default being\n`false`.",
+              "description": "Whether to leave the copy button off.\n\nNegative, which is not a style choice. A boolean attribute is true whenever\nit is *present*, so a `copyable` defaulting to true could never be turned\noff from markup: `copyable=\"false\"` is the string \"false\", which is\npresent, which is true. Measured, and it is why this is named the other way\nround. The platform already writes booleans like this for the same reason,\n`novalidate` and `nomodule` and `noreferrer` among them.\n\n**Reflected, because the gutter is a stylesheet rule.** A sheet reserves\nthe copy button's width with `:host(:not([no-copy]))`, so the attribute is\nwhat decides whether the block pays for a control. Unreflected, setting it\nfrom script takes the button away (`render` reads the property) and leaves\nthe 40px it stood in reserved.\n\nA false boolean reflects as the attribute's absence, so this costs no\n`no-copy=\"\"` on hosts that never asked for one, and the rule about a\nreflected property taking no default is satisfied by the default being\n`false`.",
               "fieldName": "noCopy"
             },
             {
@@ -17076,7 +17076,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The picker is `helix-color-picker`, not a second one written here.** It owns the\nsquare, the ramps, the eyedropper and the syntax toggle, and it is worth\nhaving on its own: a settings pane wants a picker flat on the page with\nnothing to open. This element is the field, the overlay, and the form.\nExactly the arrangement Date Input has with `helix-date-picker`.\n\n**Typed as well as picked, and the typing is the reason this is not\n`<input type=\"color\">`.** The native control gives a free picker and takes\nnearly all styling with it, which is a fair trade until a product needs the\npanel to match everything around it. What is kept from the native one is the\npart worth keeping: a real text field, so a reader who has the hex pastes it\nand never opens anything.\n\n**Every syntax is read; the one `format` names is written.** `#f00`,\n`#ff0000`, `rgb(255 0 0)`, `rgba(255,0,0,.5)`, `rgb(255 0 0 / 50%)` and\n`hsl(0 100% 50%)` are all colors, and a field that refuses one because it\nprefers another is arguing with the person using it. What the field *shows* is\nthe value itself, for the reason Date Input gives about its dates: half-typed\ninput does not survive a round trip through a formatter.\n\nThe custom property below is written by this component and never set by a\nconsumer: it carries a value this element resolved and CSS cannot, which is\nwhat lets a stylesheet on this tier paint the swatch without parsing anything\nitself.",
+          "description": "**The picker is `ColorPicker`, not a second one written here.** It owns the\nsquare, the ramps, the eyedropper and the syntax toggle, and it is worth\nhaving on its own: a settings pane wants a picker flat on the page with\nnothing to open. This element is the field, the overlay, and the form.\nExactly the arrangement Date Input has with `DatePicker`.\n\n**Typed as well as picked, and the typing is the reason this is not\n`<input type=\"color\">`.** The native control gives a free picker and takes\nnearly all styling with it, which is a fair trade until a product needs the\npanel to match everything around it. What is kept from the native one is the\npart worth keeping: a real text field, so a reader who has the hex pastes it\nand never opens anything.\n\n**Every syntax is read; the one `format` names is written.** `#f00`,\n`#ff0000`, `rgb(255 0 0)`, `rgba(255,0,0,.5)`, `rgb(255 0 0 / 50%)` and\n`hsl(0 100% 50%)` are all colors, and a field that refuses one because it\nprefers another is arguing with the person using it. What the field *shows* is\nthe value itself, for the reason Date Input gives about its dates: half-typed\ninput does not survive a round trip through a formatter.\n\nThe custom property below is written by this component and never set by a\nconsumer: it carries a value this element resolved and CSS cannot, which is\nwhat lets a stylesheet on this tier paint the swatch without parsing anything\nitself.",
           "name": "HelixHeadlessColorInput",
           "cssProperties": [
             {
@@ -17403,7 +17403,7 @@
               "name": "handleSampleStart",
               "privacy": "private",
               "readonly": true,
-              "description": "The picker is about to take the pointer over, so the panel gets out of the\nway and comes back after.\n\n`helix-color-picker` says what it is doing rather than reaching for a panel it\ncannot know it is in. A floating panel left up during an eyedropper is one\nthe reader cannot dismiss and may well be trying to sample from underneath."
+              "description": "The picker is about to take the pointer over, so the panel gets out of the\nway and comes back after.\n\n`ColorPicker` says what it is doing rather than reaching for a panel it\ncannot know it is in. A floating panel left up during an eyedropper is one\nthe reader cannot dismiss and may well be trying to sample from underneath."
             },
             {
               "kind": "field",
@@ -19298,7 +19298,7 @@
                 "text": "boolean"
               },
               "default": "false",
-              "description": "Whether the form refuses to submit while no color is chosen.\n\n**A color is exactly as requirable as a date**, which is the comparison\nthat settles it: this element extends `HelixFormElement`, carries `name`\nand reports its value, so a consumer writing\n`<helix-color-picker name=\"brand\" required>` inside a form has every reason\nto expect the form to refuse an empty one. Its two siblings in the same\ncomposition position, Date Picker and Time Picker, both check it.\n\nWritten out rather than taken from `MixinField`, for the reason those two\nrecord: the mixin brings a label, a hint and a feedback region, and this is\na panel composed *into* a field that already has all three, so taking it\nwould give one control two of each. What is left is the validity call and\nthe anchor the browser points its bubble at.\n\n**Emptiness is the only fault reported, and bad input deliberately is\nnot.** The value row is typed into a character at a time and every\nkeystroke reaches `requestValue`, so `#0`, `#00` and `#00f` would each be a\nreported fault on the way to `#00ff00`. Date Picker and Time Picker can\nreport bad input because their value moves only on a commit. Color Input\nreports it for this family instead, once, against the field the reader\ntyped into.",
+              "description": "Whether the form refuses to submit while no color is chosen.\n\n**A color is exactly as requirable as a date**, which is the comparison\nthat settles it: this element extends `HelixFormElement`, carries `name`\nand reports its value, so a consumer writing `<helix-headless-color-picker\nname=\"brand\" required>` inside a form has every reason to expect the form\nto refuse an empty one. Its two siblings in the same composition position,\nDate Picker and Time Picker, both check it.\n\nWritten out rather than taken from `MixinField`, for the reason those two\nrecord: the mixin brings a label, a hint and a feedback region, and this is\na panel composed *into* a field that already has all three, so taking it\nwould give one control two of each. What is left is the validity call and\nthe anchor the browser points its bubble at.\n\n**Emptiness is the only fault reported, and bad input deliberately is\nnot.** The value row is typed into a character at a time and every\nkeystroke reaches `requestValue`, so `#0`, `#00` and `#00f` would each be a\nreported fault on the way to `#00ff00`. Date Picker and Time Picker can\nreport bad input because their value moves only on a commit. Color Input\nreports it for this family instead, once, against the field the reader\ntyped into.",
               "attribute": "required",
               "reflects": true
             },
@@ -19464,7 +19464,7 @@
                   "text": "void"
                 }
               },
-              "description": "Reports the color to the form.\n\n**The one thing a form-associated element owes, and this element owed it\nfrom the start without paying.** It extends `HelixFormElement`, takes\n`MixinControllerValue` and carries `name`, so a consumer writing\n`<helix-color-picker name=\"brand\">` inside a form has every reason to\nexpect an entry, and the form carried none: no error, nothing in the\nmarkup to look at, and a field silently missing from the submission. Its\ntwo siblings in the same composition position, Color Input and Slider,\nboth report.\n\n`null` rather than an empty string where there is no color, which is the\ndifference between a control that submits nothing and one that submits a\nblank.\n\nThe verdict goes beside it, because `required` is this element's own claim\nand nothing else can answer it: a picker standing on a page has no field\naround it to check on its behalf. What it is *not* is a second opinion on\neverything else, which is why bad input is left to Color Input. See\n`required`."
+              "description": "Reports the color to the form.\n\n**The one thing a form-associated element owes, and this element owed it\nfrom the start without paying.** It extends `HelixFormElement`, takes\n`MixinControllerValue` and carries `name`, so a consumer writing\n`<helix-headless-color-picker name=\"brand\">` inside a form has every reason\nto expect an entry, and the form carried none: no error, nothing in the\nmarkup to look at, and a field silently missing from the submission. Its\ntwo siblings in the same composition position, Color Input and Slider, both\nreport.\n\n`null` rather than an empty string where there is no color, which is the\ndifference between a control that submits nothing and one that submits a\nblank.\n\nThe verdict goes beside it, because `required` is this element's own claim\nand nothing else can answer it: a picker standing on a page has no field\naround it to check on its behalf. What it is *not* is a second opinion on\neverything else, which is why bad input is left to Color Input. See\n`required`."
             },
             {
               "kind": "method",
@@ -19609,7 +19609,7 @@
                   }
                 }
               ],
-              "description": "A channel's new value, read off the slider that just applied it.\n\n**Both native events, because a slider sends both and at different\nmoments.** It is one of the controls a reader edits continuously, so it\nannounces `input` all the way through a drag and `change` when the drag\nends. Listening only for the commit leaves the picker deaf while a handle\nmoves, and the hue and alpha handles do not move at all.\n\n**Read off the target rather than out of a detail.** Do not bind\n`helix-input` and `helix-change` and take the proposed value out of the\nevent: those are requests, announcing what the slider is *about* to do,\nbefore anything has moved and while a listener may still refuse it. The\nnative pair fires after the slider applied, so the value is on the element\nand cannot disagree with it.\n\nIdempotent, so taking both costs nothing: the same value applied twice is\nthe same color."
+              "description": "A channel's new value, read off the slider that just applied it.\n\n**Both native events, because a slider sends both and at different\nmoments.** It is one of the controls a reader edits continuously, so it\nannounces `input` all the way through a drag and `change` when the drag\nends. Listening only for the commit leaves the picker deaf while a handle\nmoves, and the hue and alpha handles do not move at all.\n\n**Read off the target rather than out of a detail.** Do not bind `Input`\nand `helix-change` and take the proposed value out of the event: those are\nrequests, announcing what the slider is *about* to do, before anything has\nmoved and while a listener may still refuse it. The native pair fires after\nthe slider applied, so the value is on the element and cannot disagree with\nit.\n\nIdempotent, so taking both costs nothing: the same value applied twice is\nthe same color."
             },
             {
               "kind": "method",
@@ -20344,7 +20344,7 @@
                 "text": "boolean"
               },
               "default": "false",
-              "description": "Whether the form refuses to submit while no color is chosen.\n\n**A color is exactly as requirable as a date**, which is the comparison\nthat settles it: this element extends `HelixFormElement`, carries `name`\nand reports its value, so a consumer writing\n`<helix-color-picker name=\"brand\" required>` inside a form has every reason\nto expect the form to refuse an empty one. Its two siblings in the same\ncomposition position, Date Picker and Time Picker, both check it.\n\nWritten out rather than taken from `MixinField`, for the reason those two\nrecord: the mixin brings a label, a hint and a feedback region, and this is\na panel composed *into* a field that already has all three, so taking it\nwould give one control two of each. What is left is the validity call and\nthe anchor the browser points its bubble at.\n\n**Emptiness is the only fault reported, and bad input deliberately is\nnot.** The value row is typed into a character at a time and every\nkeystroke reaches `requestValue`, so `#0`, `#00` and `#00f` would each be a\nreported fault on the way to `#00ff00`. Date Picker and Time Picker can\nreport bad input because their value moves only on a commit. Color Input\nreports it for this family instead, once, against the field the reader\ntyped into.",
+              "description": "Whether the form refuses to submit while no color is chosen.\n\n**A color is exactly as requirable as a date**, which is the comparison\nthat settles it: this element extends `HelixFormElement`, carries `name`\nand reports its value, so a consumer writing `<helix-headless-color-picker\nname=\"brand\" required>` inside a form has every reason to expect the form\nto refuse an empty one. Its two siblings in the same composition position,\nDate Picker and Time Picker, both check it.\n\nWritten out rather than taken from `MixinField`, for the reason those two\nrecord: the mixin brings a label, a hint and a feedback region, and this is\na panel composed *into* a field that already has all three, so taking it\nwould give one control two of each. What is left is the validity call and\nthe anchor the browser points its bubble at.\n\n**Emptiness is the only fault reported, and bad input deliberately is\nnot.** The value row is typed into a character at a time and every\nkeystroke reaches `requestValue`, so `#0`, `#00` and `#00f` would each be a\nreported fault on the way to `#00ff00`. Date Picker and Time Picker can\nreport bad input because their value moves only on a commit. Color Input\nreports it for this family instead, once, against the field the reader\ntyped into.",
               "fieldName": "required"
             },
             {
@@ -20497,7 +20497,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The list is `helix-listbox`, and the filtering is its own.** It owns\nnarrowing, the empty state, the highlight, `aria-activedescendant` and\nscrolling an option into view. What is left for this element is the field,\nthe panel and the traffic between them.\n\n**The text and the value are two things.** What a reader types is a filter:\nit narrows the list and is thrown away. The value is whichever option they\nthen choose, and the field is rewritten to that option's words once they\nhave. The filter is cleared on close, so a reopened control does not start\nfrom the last visit's leftovers.\n\n**Focus stays in the field, which is the whole of what a combobox is.** The\narrows move the list without leaving the text, through the listbox's public\n`navigate`. `role=\"combobox\"` is on the input rather than on a wrapper,\nbecause the thing the pattern describes is the text field.\n\n**With `multiple`, the chosen options are tags in the field**, one object per\nanswer, each with its own remove control. The panel stays open while more\nthan one can be chosen and the filter clears on each choice, so the next pick\nstarts from the whole list.\n\n**`creatable` announces, and does not write.** The options are the consumer's\nlight DOM and nothing here puts an element into it: a component that grew the\nconsumer's markup would be one their framework re-renders over. It offers the\nrow and fires the event; adding the option is the consumer's.",
+          "description": "**The list is `Listbox`, and the filtering is its own.** It owns narrowing,\nthe empty state, the highlight, `aria-activedescendant` and scrolling an\noption into view. What is left for this element is the field, the panel and\nthe traffic between them.\n\n**The text and the value are two things.** What a reader types is a filter:\nit narrows the list and is thrown away. The value is whichever option they\nthen choose, and the field is rewritten to that option's words once they\nhave. The filter is cleared on close, so a reopened control does not start\nfrom the last visit's leftovers.\n\n**Focus stays in the field, which is the whole of what a combobox is.** The\narrows move the list without leaving the text, through the listbox's public\n`navigate`. `role=\"combobox\"` is on the input rather than on a wrapper,\nbecause the thing the pattern describes is the text field.\n\n**With `multiple`, the chosen options are tags in the field**, one object per\nanswer, each with its own remove control. The panel stays open while more\nthan one can be chosen and the filter clears on each choice, so the next pick\nstarts from the whole list.\n\n**`creatable` announces, and does not write.** The options are the consumer's\nlight DOM and nothing here puts an element into it: a component that grew the\nconsumer's markup would be one their framework re-renders over. It offers the\nrow and fires the event; adding the option is the consumer's.",
           "name": "HelixHeadlessCombobox",
           "cssParts": [
             {
@@ -20603,7 +20603,7 @@
           ],
           "slots": [
             {
-              "description": "The options, which are `helix-option` elements.",
+              "description": "The options.",
               "name": ""
             },
             {
@@ -20811,6 +20811,16 @@
               "description": "The chosen options themselves, for their words.\n\nRead off the options rather than stored, because the words are the\nconsumer's content and a copy of them goes stale the moment they change it.\nSelect records the same.\n\nIn the value's own order, which is the order they were chosen. It has to\nbe: a created answer is not in the list at all, so there is no position in\nthe list to give it. Neither order ever reshuffles what is already there,\nso a tag does not move out from under the pointer about to remove it.",
               "readonly": true
             },
+            {
+              "kind": "field",
+              "name": "unresolved",
+              "type": {
+                "text": "boolean"
+              },
+              "privacy": "private",
+              "description": "Whether a value is declared and this render cannot say what it is.\n\n**A placeholder is a statement, so a render that cannot read the words does\nnot make it.** `value` is an attribute and a server has it; the words are\nthe options' and the options are light-DOM children a server has none of.\nThe field then came out empty with \"Search owners\" over it, which says\nnothing is chosen where the consumer had already said what is. Blank, it\nsays nothing, keeps its box, and is filled in by the render after\nhydration, which is the answer `Select.Root.display` reaches for in the same\nposition and for the same reason.\n\nThe placeholder still stands where there is genuinely no value, which is\nwhat it is for and the one case a server can be sure of.",
+              "readonly": true
+            },
             {
               "kind": "field",
               "name": "showing",
@@ -20945,7 +20955,7 @@
               "name": "handleFieldClick",
               "privacy": "private",
               "readonly": true,
-              "description": "Anywhere in the field opens the list, and puts the caret where typing goes.\n\nOn the field rather than on the `<input>`. The caret is a sibling of the\ninput and so is the tag strip, so a handler on the input alone leaves the\none part of the control that visibly says \"this opens\" doing nothing.\nAnything drawn inside the field is the field.\n\nA tag is the exception, because a tag is its own control: its dismiss\nremoves a choice, and opening the list on the way past would leave the\nreader looking at a panel they did not ask for. Read from the composed\npath, since the press lands on a button inside the tag's shadow root and\nthe event has retargeted by the time it arrives here.\n\nIt only ever opens. Closing is Escape, a choice, or leaving, which is what\nthe input already did and what a text field wants: a control that closed\nwhen you clicked into the middle of your own typing would be unusable."
+              "description": "Anywhere in the field opens the list, and puts the caret where typing goes.\n\nOn the field rather than on the `<input>`. The caret is a sibling of the\ninput and so is the tag strip, so a handler on the input alone leaves the\none part of the control that visibly says \"this opens\" doing nothing.\nAnything drawn inside the field is the field.\n\nA tag is the exception, because a tag is its own control: its dismiss\nremoves a choice, and opening the list on the way past would leave the\nreader looking at a panel they did not ask for. Read from the composed\npath, since the press lands on a button inside the tag's shadow root and\nthe event has retargeted by the time it arrives here.\n\nIt only ever opens. Closing is Escape, a choice, or leaving, which is what\nthe input already did and what a text field wants: a control that closed\nwhen you clicked into the middle of your own typing would be unusable.\n\nThe tag is found by identity rather than by tag name. A name matches one\ntier's element and leaves the other's press falling through to the open,\nand a subclass under a third name is missed by both."
             },
             {
               "kind": "method",
@@ -22634,7 +22644,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The list is `helix-listbox`, filtering and all.** Listbox already owns\nnarrowing, the empty state, the highlight over what matched,\n`aria-activedescendant`, scrolling an option into view and the keyboard that\nmoves between them. Its own note says a command palette is that list again\nwith something different opening it, and this is that something: a dialog,\nand a key that summons it.\n\n**The value is not kept.** A palette runs a command and forgets it: nothing\nhere is a form control and nothing is submitted, so what it reports is that\nan option was chosen, once, with `helix-select`. A consumer that wants the\nlast command remembered is a consumer with somewhere better to keep it.\n\n**The hotkey is global and the element does not have to be open to hear it.**\nThe listener is on the document, because the point of a palette is that it\nanswers from wherever the reader is, including from inside another control's\ntext field. `mod` is Command on a Mac and Control elsewhere, resolved here so\na consumer writes one thing.\n\nThe filter is emptied as the palette closes, for the reason Combobox and\nSelect both record: a control reopened still narrowed by the last visit's\nsearch is one whose second use starts from the first one's leftovers, and a\npalette is used far more often than either.",
+          "description": "**The list is `Listbox`, filtering and all.** Listbox already owns narrowing,\nthe empty state, the highlight over what matched, `aria-activedescendant`,\nscrolling an option into view and the keyboard that moves between them. Its\nown note says a command palette is that list again with something different\nopening it, and this is that something: a dialog, and a key that summons it.\n\n**The value is not kept.** A palette runs a command and forgets it: nothing\nhere is a form control and nothing is submitted, so what it reports is that\nan option was chosen, once, with `helix-select`. A consumer that wants the\nlast command remembered is a consumer with somewhere better to keep it.\n\n**The hotkey is global and the element does not have to be open to hear it.**\nThe listener is on the document, because the point of a palette is that it\nanswers from wherever the reader is, including from inside another control's\ntext field. `mod` is Command on a Mac and Control elsewhere, resolved here so\na consumer writes one thing.\n\nThe filter is emptied as the palette closes, for the reason Combobox and\nSelect both record: a control reopened still narrowed by the last visit's\nsearch is one whose second use starts from the first one's leftovers, and a\npalette is used far more often than either.",
           "name": "HelixHeadlessCommand",
           "cssParts": [
             {
@@ -22692,7 +22702,7 @@
           ],
           "slots": [
             {
-              "description": "The commands, which are `helix-option` elements and their groups.",
+              "description": "The commands, and any groups of them.",
               "name": ""
             },
             {
@@ -23666,7 +23676,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
+              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-headless-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-headless-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
               "attribute": "command",
               "helix": {
                 "references": [
@@ -24449,7 +24459,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
+              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-headless-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-headless-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
               "fieldName": "command",
               "inheritedFrom": {
                 "name": "HelixHeadlessButton",
@@ -24735,15 +24745,15 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The schema is markup and the data is a property.** Columns are declared as\n`helix-table-column` children, because a schema is small and something a\nreader should see in the HTML; rows are handed over as an array, because a\ntable that draws ten thousand elements cannot window them.\n\n```html\n<helix-data-table .rows=${rows} sort=\"name:asc\" page-size=\"20\">\n  <helix-table-column field=\"name\" sortable>Name</helix-table-column>\n  <helix-table-column field=\"size\" align=\"end\" sortable .cell=${bytes}>Size</helix-table-column>\n</helix-data-table>\n```\n\n`helix-table` takes rows as markup and is the right component for a table of\ntwelve things a person wrote out.\n\n**Sort and filter apply to the whole set, never the page.** Sorting the drawn\nwindow is correct on the first screen and wrong on every other one.\n\n**The cell values do not server-render.** The headings do, and the row\nelements do, but reading a cell needs the columns, and the columns are markup\nchildren Lit SSR does not hand this element. So a crawler sees a table of the\nright shape with nothing in it. That is the cost of the first paragraph: use\n`helix-table` where the served HTML has to carry the values.\n\n**`selectable` adds a column of checkboxes, keyed on `row-key`.** A row is\nselected by its identity rather than its position, so a selection survives a\nsort, a filter and a page. `selected` is the chosen keys, comma separated, the\nshape Listbox uses for a multiple choice, and like `sort` it is controlled\nwhen set and seeded by `default-selected`. The box in the heading selects every\nrow in the result, across pages, and shows a dash while only some are chosen.\nChoosing one of many, rather than several, is opening a row, and that is a\nlink or a button in a cell rather than a selection.\n\n**A reader can size and order the columns.** `resizable` on a column puts a\nseparator on its edge; `reorderable` on the table puts a grip on each heading\nthat has a `field`, moved by dragging or with the arrow keys. The order lands\nin `order` as fields, and the consumer's own markup is never rewritten: the\ntable assigns the headings to its header in that order instead.\n\n**`csv()` hands back the result as text**: every row the filter lets through,\nin the order on screen, in the columns on screen. Saving it is the page's\nbusiness, the same as any other file.\n\n**There is no `columns` attribute here, and `helix-table` has one.** The\ndifference is what a column is in each: here it is the schema, so a table\nwithout one has nothing to read a row with, and every column is a place to\nwrite a `width`. A `helix-table` holds cells a consumer wrote and can be\nwritten with no column at all, so its `columns` is where such a table is\nsized from. Written on this element the attribute is inert, which is the trap\nworth knowing when moving between the two. What both take is the CSS hook,\n`--data-table-columns` here and `--table-columns` there.",
+          "description": "**The schema is markup and the data is a property.** Columns are declared as\n`TableColumn` children, because a schema is small and something a reader\nshould see in the HTML; rows are handed over as an array, because a table\nthat draws ten thousand elements cannot window them.\n\n```html\n<helix-headless-data-table .rows=${rows} sort=\"name:asc\" page-size=\"20\">\n  <helix-headless-table-column field=\"name\" sortable>Name</helix-headless-table-column>\n  <helix-headless-table-column field=\"size\" align=\"end\" sortable .cell=${bytes}>Size</helix-headless-table-column>\n</helix-headless-data-table>\n```\n\n`Table` takes rows as markup and is the right component for a table of twelve\nthings a person wrote out.\n\n**Sort and filter apply to the whole set, never the page.** Sorting the drawn\nwindow is correct on the first screen and wrong on every other one.\n\n**The cell values do not server-render.** The headings do, and the row\nelements do, but reading a cell needs the columns, and the columns are markup\nchildren Lit SSR does not hand this element. So a crawler sees a table of the\nright shape with nothing in it. That is the cost of the first paragraph: use\n`Table` where the served HTML has to carry the values.\n\n**`selectable` adds a column of checkboxes, keyed on `row-key`.** A row is\nselected by its identity rather than its position, so a selection survives a\nsort, a filter and a page. `selected` is the chosen keys, comma separated, the\nshape Listbox uses for a multiple choice, and like `sort` it is controlled\nwhen set and seeded by `default-selected`. The box in the heading selects every\nrow in the result, across pages, and shows a dash while only some are chosen.\nChoosing one of many, rather than several, is opening a row, and that is a\nlink or a button in a cell rather than a selection.\n\n**A reader can size and order the columns.** `resizable` on a column puts a\nseparator on its edge; `reorderable` on the table puts a grip on each heading\nthat has a `field`, moved by dragging or with the arrow keys. The order lands\nin `order` as fields, and the consumer's own markup is never rewritten: the\ntable assigns the headings to its header in that order instead.\n\n**`csv()` hands back the result as text**: every row the filter lets through,\nin the order on screen, in the columns on screen. Saving it is the page's\nbusiness, the same as any other file.\n\n**There is no `columns` attribute here, and Table has one.** The difference\nis what a column is in each: here it is the schema, so a table without one\nhas nothing to read a row with, and every column is a place to write a\n`width`. A Table holds cells a consumer wrote and can be written with no\ncolumn at all, so its `columns` is where such a table is sized from. Written\non this element the attribute is inert, which is the trap worth knowing when\nmoving between the two.",
           "name": "HelixHeadlessDataTable",
           "cssProperties": [
             {
-              "description": "The edge the column a cell sits under is set to, `start`, `center` or `end`, and unset under a column that sets none. Written by the component onto each cell it draws, read by the sheet that paints one; `helix-table` publishes `--table-cell-align` for the same attribute on the cells a consumer wrote, under its own family's name.",
+              "description": "The edge the column a cell sits under is set to, `start`, `center` or `end`, and unset under a column that sets none. Written by the component onto each cell it draws, read by the sheet that paints one; Table publishes `--table-cell-align` for the same attribute on the cells a consumer wrote, under its own family's name.",
               "name": "--data-table-cell-align"
             },
             {
-              "description": "The column track list this element worked out from the columns declared in it. Written by the component, read by the sheet that paints: a stylesheet takes it to lay the grid out, and setting it does nothing, because the component writes it inline on the very element that reads it and an inline declaration beats any sheet. `--data-table-columns` is the one to set.",
+              "description": "The column track list this element worked out from the columns declared in it. Written by the component, read by the sheet that paints: a stylesheet takes it to lay the grid out, and setting it does nothing, because the component writes it inline on the very element that reads it and an inline declaration beats any sheet. A column's `width` is the one to set.",
               "name": "--data-table-template"
             }
           ],
@@ -24899,7 +24909,7 @@
           ],
           "slots": [
             {
-              "description": "The columns, declared as `helix-table-column` children.",
+              "description": "The columns, declared rather than drawn.",
               "name": ""
             },
             {
@@ -24920,7 +24930,7 @@
               },
               "static": true,
               "default": "{ ...HelixElement.shadowRootOptions, slotAssignment: 'manual', }",
-              "description": "Nothing is assigned by name, so the columns can be put in a header row.\n\n`role=\"columnheader\"` under `role=\"grid\"` with no owning row is invalid, and\nthe columns are ordinary children a consumer writes with no `slot`. Manual\nassignment is the only way to wrap them without editing the consumer's\nmarkup, which is a change a React consumer's hydration reports and does not\npatch up. The same arrangement `helix-table` arrived at, for the same reason."
+              "description": "Nothing is assigned by name, so the columns can be put in a header row.\n\n`role=\"columnheader\"` under `role=\"grid\"` with no owning row is invalid,\nand the columns are ordinary children a consumer writes with no `slot`.\nManual assignment is the only way to wrap them without editing the\nconsumer's markup, which is a change a React consumer's hydration reports\nand does not patch up. The same arrangement `Table` arrived at, for the\nsame reason."
             },
             {
               "kind": "field",
@@ -25102,7 +25112,7 @@
               "type": {
                 "text": "number"
               },
-              "description": "How many rows a page holds. Zero is every row, which is what \"no paging\" means.\n\n**It announces nothing, and that is not an omission.** Nothing in this\nelement ever asks for a different page size: a size selector is\n`helix-pagination-size`, which a consumer puts in the `footer` slot and\nanswers themselves. Routing this through `requestChange` would announce an\nevent the element never has cause to fire.",
+              "description": "How many rows a page holds. Zero is every row, which is what \"no paging\" means.\n\n**It announces nothing, and that is not an omission.** Nothing in this\nelement ever asks for a different page size: a size selector is\n`PaginationSize`, which a consumer puts in the `footer` slot and answers\nthemselves. Routing this through `requestChange` would announce an event\nthe element never has cause to fire.",
               "attribute": "page-size"
             },
             {
@@ -25430,12 +25440,22 @@
               },
               "privacy": "protected",
               "readonly": true,
-              "description": "Emptiness is a fact about the data here, not about slotted children.\n\n**The page rather than the window**, which is the difference between the two\ncounts `view` carries and the reason this reads `available`. A windowed\ntable draws a slice, and a slice can be short of the rows that are there:\nasking whether *this* slice has rows says \"nothing matched\" about a result\nthat matched plenty.\n\nThe mixin's own answer counts children with no `slot`, which for this\nelement is the columns. Left in place, a table with columns and no data\nwould call itself full.",
+              "description": "Emptiness is a fact about the data here, not about slotted children.\n\n**The page rather than the window**, which is the difference between the two\ncounts `view` carries and the reason this reads `available`. A windowed\ntable draws a slice, and a slice can be short of the rows that are there:\nasking whether *this* slice has rows says \"nothing matched\" about a result\nthat matched plenty.\n\nThe mixin's own answer counts children with no `slot`, which for this\nelement is the columns. Left in place, a table with columns and no data\nwould call itself full.\n\n**`sourceKnown` is the half the mixin was keeping**, and overriding without\nit is what put \"Nothing to show\" over eight rows of files on every server\nrender of this family's own preview. `MixinCollection.hasRows` answers true\nuntil the rows have been counted for exactly that reason, and a count is\nwhat this replaced.",
               "inheritedFrom": {
                 "name": "MixinCollection",
                 "module": "src/mixins/mixin-collection/mixin-collection.ts"
               }
             },
+            {
+              "kind": "field",
+              "name": "sourceKnown",
+              "type": {
+                "text": "boolean"
+              },
+              "privacy": "private",
+              "description": "Whether this render can tell a result with no rows from one not yet handed\nover.\n\n**Unknown is not empty, which is what an empty state must never say.** The\nrows are a property, and a property is set by whoever holds the element: a\nconsumer binding it in a template has it on the server, and a consumer\nassigning it from a script has it only once that script has run, which is\nafter the parse a server's markup ends at. A render that cannot see a\ndocument cannot tell the two apart, and a hydrating render must not look,\nits output having to be the server's.\n\nThe same shape as `Heatmap.sourceKnown`, asked of a property rather than of\na document, and `childrenKnown` is already both halves of the question: a\nserver has no tree and a hydrating render may not read one.",
+              "readonly": true
+            },
             {
               "kind": "method",
               "name": "requestSort",
@@ -25770,7 +25790,7 @@
                   }
                 }
               ],
-              "description": "One cell, carrying its column's edge as a fact rather than as a decision.\n\n**A fact published rather than a property chosen**, which is what\n`helix-table` does with the same attribute: what lining up *means*, a\n`justify-content` for a value that fits or a `text-align` for one that\nwraps, belongs to whoever draws the cell. Written here as an inline\n`text-align` it was this tier choosing, in the tier that ships no CSS, and a\nconsumer could not overrule it either, an inline declaration beating any\n`::part(cell)` rule they wrote.\n\nOnly the three edges reach it, `align` being a string anybody can write and\nthis being a style attribute: `edgeOf` is that guard, and `helix-table`\nholds the same value to the same three."
+              "description": "One cell, carrying its column's edge as a fact rather than as a decision.\n\n**A fact published rather than a property chosen**, which is what `Table`\ndoes with the same attribute: what lining up *means*, a `justify-content`\nfor a value that fits or a `text-align` for one that wraps, belongs to\nwhoever draws the cell. Written here as an inline `text-align` it was this\ntier choosing, in the tier that ships no CSS, and a consumer could not\noverrule it either, an inline declaration beating any `::part(cell)` rule\nthey wrote.\n\nOnly the three edges reach it, `align` being a string anybody can write and\nthis being a style attribute: `edgeOf` is that guard, and `Table` holds the\nsame value to the same three."
             },
             {
               "kind": "method",
@@ -26671,7 +26691,7 @@
               "type": {
                 "text": "number"
               },
-              "description": "How many rows a page holds. Zero is every row, which is what \"no paging\" means.\n\n**It announces nothing, and that is not an omission.** Nothing in this\nelement ever asks for a different page size: a size selector is\n`helix-pagination-size`, which a consumer puts in the `footer` slot and\nanswers themselves. Routing this through `requestChange` would announce an\nevent the element never has cause to fire.",
+              "description": "How many rows a page holds. Zero is every row, which is what \"no paging\" means.\n\n**It announces nothing, and that is not an omission.** Nothing in this\nelement ever asks for a different page size: a size selector is\n`PaginationSize`, which a consumer puts in the `footer` slot and answers\nthemselves. Routing this through `requestChange` would announce an event\nthe element never has cause to fire.",
               "fieldName": "pageSize"
             },
             {
@@ -26960,7 +26980,7 @@
               }
             }
           ],
-          "description": "One page of rows, clamped.\n\nA page past the end returns the last page rather than nothing, because the\nusual way to be past the end is a bookmarked link after the result shrank, and\nan empty table is a worse answer than the last page of a smaller one. The\nsame clamp `helix-pagination` already applies to what it reports."
+          "description": "One page of rows, clamped.\n\nA page past the end returns the last page rather than nothing, because the\nusual way to be past the end is a bookmarked link after the result shrank,\nand an empty table is a worse answer than the last page of a smaller one. The\nsame clamp `Pagination` already applies to what it reports."
         },
         {
           "kind": "function",
@@ -27111,7 +27131,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The calendar is `helix-date-picker`, not a second one written here.** It\nalready knows about months, ranges, disabled days, the first day of the week\nand every keyboard rule a grid owes a reader. This element is the field, the\noverlay, and the translation between them.\n\n**Typed as well as picked, and the typing is why it is not\n`<input type=\"date\">`.** The native control gives a free picker and a free\nlocale and takes almost all styling with it. What is kept from it is the part\nworth keeping: a real text field.",
+          "description": "**The calendar is `DatePicker`, not a second one written here.** It already\nknows about months, ranges, disabled days, the first day of the week and\nevery keyboard rule a grid owes a reader. This element is the field, the\noverlay, and the translation between them.\n\n**Typed as well as picked, and the typing is why it is not\n`<input type=\"date\">`.** The native control gives a free picker and a free\nlocale and takes almost all styling with it. What is kept from it is the part\nworth keeping: a real text field.",
           "name": "HelixHeadlessDateInput",
           "cssParts": [
             {
@@ -27604,7 +27624,7 @@
                   "text": "TemplateResult"
                 }
               },
-              "description": "The mark on the button that opens the calendar, behind a seam.\n\nSlot fallback, so a consumer replaces it by slotting; and a method, so the\ntier above replaces it by overriding.\n\n**A named import rather than `<helix-icon>`**, which loses on three counts.\n`helix-icon` indexes the whole icon map at runtime, `icons[this.name]`, and\nno bundler can shake a runtime lookup, so every consumer of any component\nthat renders one ships all 50kB of path data; a named import ships one\npath. The import is checked when the package compiles, where a `name`\nstring is checked only by a later name lookup. And `helix-icon` is not\nregistered on the headless tier at all, which registers `helix-headless-icon`,\nso a headless consumer gets an inert unknown element and no mark whatsoever."
+              "description": "The mark on the button that opens the calendar, behind a seam.\n\nSlot fallback, so a consumer replaces it by slotting; and a method, so the\ntier above replaces it by overriding.\n\n**A named import rather than `<helix-headless-icon>`**, which loses on\nthree counts. `Icon` indexes the whole icon map at runtime,\n`icons[this.name]`, and no bundler can shake a runtime lookup, so every\nconsumer of any component that renders one ships all 50kB of path data; a\nnamed import ships one path. The import is checked when the package\ncompiles, where a `name` string is checked only by a later name lookup. And\n`Icon` is not registered on the headless tier at all, which registers\n`helix-headless-icon`, so a headless consumer gets an inert unknown element\nand no mark whatsoever."
             },
             {
               "kind": "field",
@@ -32026,7 +32046,7 @@
                   "text": "void"
                 }
               },
-              "description": "Says no to a dismissal, by moving.\n\nA persistent dialog that ignores a press on the backdrop is\nindistinguishable from one that did not notice, and a reader's next move is\nto press again harder. A short swell and settle is the smallest thing that\nreads as \"I heard you, and no\": it is the same gesture a native alert makes\non macOS and the one every reader has already been taught.\n\nAn attribute for the length of the animation rather than a class, cleared\non `animationend`, so the styled tier decides what refusing looks like and\nthis decides only when it happens. Removed first so a second press\nrestarts it: without that, pressing twice does nothing the second time,\nwhich is the impression the whole thing exists to avoid.\n\n**`once: true` is not a release when the event cannot arrive.** Under\n`prefers-reduced-motion: reduce` the styled tier sets `animation: none`, so\nno animation runs, `animationend` never fires, and every refused press left\na listener that could not be removed and an attribute that stayed on. The\ncontroller bounds it to one listener, aborted before the next is added.\n\n**And the attribute is cleared at once when no animation is running**, asked\nof the platform rather than of a media query: this tier does not know what\nrefusing looks like, and the same silence arrives from a consumer's\nstylesheet, from reduced motion, and from no styled tier being loaded at\nall.\n\n**`CSSAnimation` specifically, because `getAnimations` returns transitions\ntoo.** A panel that is still running its entrance transition answers the\nbare `length` check with a truthy number while no animation exists to end,\nso a refusal landing inside that window took the listener branch and waited\nfor an `animationend` that was never coming: the attribute stayed on until\nthe next refused press. That is the same bug this method exists to close,\none window narrower. `getAnimations` flushes pending style itself, so no\nreflow is forced for it."
+              "description": "Says no to a dismissal, by moving.\n\nA persistent dialog that ignores a press on the backdrop is\nindistinguishable from one that did not notice, and a reader's next move is\nto press again harder. A short swell and settle is the smallest thing that\nreads as \"I heard you, and no\": it is the same gesture a native alert makes\non macOS and the one every reader has already been taught.\n\nAn attribute for the length of the animation rather than a class, cleared\non `animationend`, so the styled tier decides what refusing looks like and\nthis decides only when it happens. Removed first so a second press\nrestarts it: without that, pressing twice does nothing the second time,\nwhich is the impression the whole thing exists to avoid.\n\n**`once: true` is not a release when the event cannot arrive.** A sheet\nthat sets `animation: none` under `prefers-reduced-motion: reduce` runs\nnone, so `animationend` never fires, and every refused press left a\nlistener that could not be removed and an attribute that stayed on. The\ncontroller bounds it to one listener, aborted before the next is added.\n\n**And the attribute is cleared at once when no animation is running**, asked\nof the platform rather than of a media query: this tier does not know what\nrefusing looks like, and the same silence arrives from a consumer's\nstylesheet, from reduced motion, and from no styled tier being loaded at\nall.\n\n**`CSSAnimation` specifically, because `getAnimations` returns transitions\ntoo.** A panel that is still running its entrance transition answers the\nbare `length` check with a truthy number while no animation exists to end,\nso a refusal landing inside that window took the listener branch and waited\nfor an `animationend` that was never coming: the attribute stayed on until\nthe next refused press. That is the same bug this method exists to close,\none window narrower. `getAnimations` flushes pending style itself, so no\nreflow is forced for it."
             },
             {
               "kind": "field",
@@ -33986,7 +34006,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "Takes `MixinControllerOpen`, so `open`, `defaultOpen` and the\n`helix-before-open` and `helix-before-close` requests behave exactly as they\ndo on an accordion item, and a consumer already knows the API.\n\nModal by default, which is three things together and not one: focus is\ntrapped inside, Escape and an outside press close it, and the page behind is\ncovered. A panel with the look of a modal and none of its behavior is worse\nthan an inline one, because it invites the reader to treat it as blocking.\n\nContent stays mounted when closed, so it is in the server's markup and can\nanimate. A closed `<dialog>` is already out of the tab order and out of the\naccessibility tree, so nothing here writes `inert`. The `role`, the\n`aria-modal` and the scrim come from the element too, and\n`drawer-root.ssr.spec.ts` asserts the panel carries no hand-written `inert`\nand no hand-written `role`.\n\nOpened declaratively with the platform's invoker attributes:\n\n```html\n<helix-button commandfor=\"nav\" command=\"--toggle\">Menu</helix-button>\n<helix-drawer id=\"nav\">…</helix-drawer>\n```",
+          "description": "Takes `MixinControllerOpen`, so `open`, `defaultOpen` and the\n`helix-before-open` and `helix-before-close` requests behave exactly as they\ndo on an accordion item, and a consumer already knows the API.\n\nModal by default, which is three things together and not one: focus is\ntrapped inside, Escape and an outside press close it, and the page behind is\ncovered. A panel with the look of a modal and none of its behavior is worse\nthan an inline one, because it invites the reader to treat it as blocking.\n\nContent stays mounted when closed, so it is in the server's markup and can\nanimate. A closed `<dialog>` is already out of the tab order and out of the\naccessibility tree, so nothing here writes `inert`. The `role`, the\n`aria-modal` and the scrim come from the element too, and\n`drawer-root.ssr.spec.ts` asserts the panel carries no hand-written `inert`\nand no hand-written `role`.\n\nOpened declaratively with the platform's invoker attributes:\n\n```html\n<helix-headless-button commandfor=\"nav\" command=\"--toggle\">Menu</helix-headless-button>\n<helix-headless-drawer id=\"nav\">…</helix-headless-drawer>\n```",
           "name": "HelixHeadlessDrawer",
           "cssParts": [
             {
@@ -34056,7 +34076,7 @@
               "type": {
                 "text": "'start' | 'end' | 'top' | 'bottom' | undefined"
               },
-              "description": "Which edge the drawer comes from.\n\nUnset is `end`, which the styled tier reads as `:host(:not([placement]))`\nbeside `:host([placement=\"end\"])` rather than as a value written here: a\ndefault assigned to the property would reflect onto every host and put a\n`placement=\"end\"` into markup no consumer wrote.",
+              "description": "Which edge the drawer comes from.\n\nUnset is `end`, and it stays unset rather than being written down: a\ndefault assigned to the property would reflect onto every host and put a\n`placement=\"end\"` into markup no consumer wrote. So what this tier\nguarantees is the absence, and a sheet selects on it.",
               "attribute": "placement",
               "reflects": true
             },
@@ -34606,7 +34626,7 @@
               "type": {
                 "text": "'start' | 'end' | 'top' | 'bottom' | undefined"
               },
-              "description": "Which edge the drawer comes from.\n\nUnset is `end`, which the styled tier reads as `:host(:not([placement]))`\nbeside `:host([placement=\"end\"])` rather than as a value written here: a\ndefault assigned to the property would reflect onto every host and put a\n`placement=\"end\"` into markup no consumer wrote.",
+              "description": "Which edge the drawer comes from.\n\nUnset is `end`, and it stays unset rather than being written down: a\ndefault assigned to the property would reflect onto every host and put a\n`placement=\"end\"` into markup no consumer wrote. So what this tier\nguarantees is the absence, and a sheet selects on it.",
               "fieldName": "placement"
             },
             {
@@ -39185,7 +39205,7 @@
               "type": {
                 "text": "HelixHeadlessFilterField[]"
               },
-              "description": "The fields declared inside this, in the order they were written.\n\nWhat a render may know of them: none on a server, whose shim has no tree\nand threw here, and none while hydrating what one sent.",
+              "description": "The fields declared inside this, in the order they were written.\n\nWhat a render may know of them: none on a server, whose shim has no tree\nand threw here, and none while hydrating what one sent.\n\nBy identity rather than by tag. A list of names can only hold the tags this\npackage knows about, so it misses any third one, which is what subclassing\na headless element produces, and it is also how a headless package learns a\nstyled tag. `instanceof` answers for every one of them and proves the type\nthe signature claims.",
               "readonly": true
             },
             {
@@ -40159,7 +40179,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The tokens are the consumer's own elements, slotted.** What a filter narrows\nby belongs to the consumer's model rather than to this element: a facet with a\nprefix, a saved query, a range. So they are slotted rather than rendered from\na list of strings, and removing one is a request this element makes rather\nthan a change it applies. Combobox is the other way round, and that is the\ndifference between the two: its tokens are a view of a selection it owns,\ndrawn from options it was handed.\n\n**`value` is the text being typed, not the tokens**, which is what makes this\na field a consumer can listen to. Every keystroke announces `input` with the\ntext on the element, so a table narrows as the reader types. The tokens need\nno reporting back, being markup the consumer already holds.\n\nBackspace on an empty field asks the last token to go, through the same\n`helix-dismiss` its own remove control fires, so a consumer who handled the\npointer has already handled the keyboard.\n\n**It reads what was declared, and nothing that was not.** Write\n`FilterField`s (usually inside the `FilterMenu` slotted into `start`, so the\npicker and the text work off one set), and Enter reads the box against them:\n`disposition:refunded,settled` and `-disposition:refunded` announce\n`helix-facet`, exactly as ticking those values in the picker would, and the\ntext they were read from goes. Everything else stays where it was typed, so\n`refund tuesday disposition:settled` is a filter and a search at once.\n`grammar.ts` carries the whole rule and is worth reading before relying on it;\nthe short version is that Helix owns the punctuation and the consumer owns\nevery word in it. A field with no `FilterField` in it parses nothing, which is\nthe behavior this had before it could.\n\n**A field whose values cannot be listed belongs here rather than in the\npicker.** `assignee:ada` is a filter and nobody wants to enumerate every name\nin the company, but a menu page for it would have nothing on it to press.\nWritten directly inside this element it is read by the grammar and never\noffered as a page: unslotted, so nothing draws it, which is the same trick\n`FilterField` already plays on its own children.\n\n**It can suggest what to type next, and deciding what to suggest is still the\nconsumer's.** Slot `helix-option` elements into `suggestions` and the field\nshows them, moves through them with the arrows, and lets Enter press one; the\noption fires its own `helix-select` and the consumer decides what choosing\nmeant. Keys before a colon and values after it are the same panel with\ndifferent contents, and `readFilterText` is exported for a consumer narrowing\nthat panel against what a reader has half typed.\n\n**The panel opens when there is something in it and closes when there is\nnot**, rather than on a choice. A reader who picks a key wants its values\nnext, and closing on the choice would shut the panel between two renders. Use\n`selectionStart` and `setSelectionRange` to write a completion back and put\nthe caret after it, or the next parse reads the wrong term.\n\n**Three things here have three different owners, and only one of them is a\ncontrolled property.**\n\n`value` is the text, and it is uncontrolled by default: the reader types, the\nelement applies it, and `input` says so. **Writing `value` does not take that\nover.** It is a seed and a correction, so a consumer who sets it and then\nexpects the field to hold still will watch the next keystroke overwrite them.\nOwnership is taken by refusing `helix-before-change` and writing the value\nyou want, which is the same shape every field in this library uses and is\nworth saying out loud here because the property looks like React's.\n`default-value` seeds the field without claiming anything.\n\n`open` is the suggestions panel, and it is the ordinary controlled pair:\nrefuse `helix-before-open` or `helix-before-close` to take the decision, and\n`default-open` to seed it.\n\nThe tokens are neither, because they are not state. They are the consumer's\nown elements in a slot, so there is nothing for this to own and nothing to\nreport back: a token goes when the consumer removes it from their markup,\nhaving been asked by `helix-dismiss`. **A `FilterChip` naming a declared field\nis asked differently**, and it has to be: the picker holds what is ticked, so\na chip that removed itself without telling it left the value ticked in a list\nnobody could see was wrong. That dismissal is answered as the field being\nfiltered by nothing, so it arrives as `helix-facet` like every other change\nand the chip goes because the filter went.",
+          "description": "**The tokens are the consumer's own elements, slotted.** What a filter narrows\nby belongs to the consumer's model rather than to this element: a facet with a\nprefix, a saved query, a range. So they are slotted rather than rendered from\na list of strings, and removing one is a request this element makes rather\nthan a change it applies. Combobox is the other way round, and that is the\ndifference between the two: its tokens are a view of a selection it owns,\ndrawn from options it was handed.\n\n**`value` is the text being typed, not the tokens**, which is what makes this\na field a consumer can listen to. Every keystroke announces `input` with the\ntext on the element, so a table narrows as the reader types. The tokens need\nno reporting back, being markup the consumer already holds.\n\nBackspace on an empty field asks the last token to go, through the same\n`helix-dismiss` its own remove control fires, so a consumer who handled the\npointer has already handled the keyboard.\n\n**It reads what was declared, and nothing that was not.** Write\n`FilterField`s (usually inside the `FilterMenu` slotted into `start`, so the\npicker and the text work off one set), and Enter reads the box against them:\n`disposition:refunded,settled` and `-disposition:refunded` announce\n`helix-facet`, exactly as ticking those values in the picker would, and the\ntext they were read from goes. Everything else stays where it was typed, so\n`refund tuesday disposition:settled` is a filter and a search at once.\n`grammar.ts` carries the whole rule and is worth reading before relying on it;\nthe short version is that Helix owns the punctuation and the consumer owns\nevery word in it. A field with no `FilterField` in it parses nothing, which is\nthe behavior this had before it could.\n\n**A field whose values cannot be listed belongs here rather than in the\npicker.** `assignee:ada` is a filter and nobody wants to enumerate every name\nin the company, but a menu page for it would have nothing on it to press.\nWritten directly inside this element it is read by the grammar and never\noffered as a page: unslotted, so nothing draws it, which is the same trick\n`FilterField` already plays on its own children.\n\n**It can suggest what to type next, and deciding what to suggest is still the\nconsumer's.** Slot options into `suggestions` and the field shows them, moves\nthrough them with the arrows, and lets Enter press one; the option fires its\nown `helix-select` and the consumer decides what choosing meant. Keys before\na colon and values after it are the same panel with different contents, and\n`readFilterText` is exported for a consumer narrowing that panel against what\na reader has half typed.\n\n**The panel opens when there is something in it and closes when there is\nnot**, rather than on a choice. A reader who picks a key wants its values\nnext, and closing on the choice would shut the panel between two renders. Use\n`selectionStart` and `setSelectionRange` to write a completion back and put\nthe caret after it, or the next parse reads the wrong term.\n\n**Three things here have three different owners, and only one of them is a\ncontrolled property.**\n\n`value` is the text, and it is uncontrolled by default: the reader types, the\nelement applies it, and `input` says so. **Writing `value` does not take that\nover.** It is a seed and a correction, so a consumer who sets it and then\nexpects the field to hold still will watch the next keystroke overwrite them.\nOwnership is taken by refusing `helix-before-change` and writing the value\nyou want, which is the same shape every field in this library uses and is\nworth saying out loud here because the property looks like React's.\n`default-value` seeds the field without claiming anything.\n\n`open` is the suggestions panel, and it is the ordinary controlled pair:\nrefuse `helix-before-open` or `helix-before-close` to take the decision, and\n`default-open` to seed it.\n\nThe tokens are neither, because they are not state. They are the consumer's\nown elements in a slot, so there is nothing for this to own and nothing to\nreport back: a token goes when the consumer removes it from their markup,\nhaving been asked by `helix-dismiss`. **A `FilterChip` naming a declared field\nis asked differently**, and it has to be: the picker holds what is ticked, so\na chip that removed itself without telling it left the value ticked in a list\nnobody could see was wrong. That dismissal is answered as the field being\nfiltered by nothing, so it arrives as `helix-facet` like every other change\nand the chip goes because the filter went.",
           "name": "HelixHeadlessFilterInput",
           "cssParts": [
             {
@@ -42225,7 +42245,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "The timestamp, as anything `Date` parses.\n\nAn ISO string is what an API sends and what `<time datetime>` takes, so it\nis the shape that needs no translation at either end. Left unset this\nrenders nothing rather than the epoch, which is what parsing an empty string\nwould give.\n\n**The element's own text is read as a fallback, and only in a browser.**\n`<helix-format-date elapsed>2026-08-27T01:00:00Z</helix-format-date>` is the\nspelling that reads best in markup, and it cannot be the only one: measured\nagainst Lit's server shim, `textContent`, `children`, `firstChild` and\n`querySelector` are all `undefined` there, so a server has no way to see it.\nAn element rendered on a server needs the attribute; one rendered in a\nbrowser may use either.",
+              "description": "The timestamp, as anything `Date` parses.\n\nAn ISO string is what an API sends and what `<time datetime>` takes, so it\nis the shape that needs no translation at either end. Left unset this\nrenders nothing rather than the epoch, which is what parsing an empty string\nwould give.\n\n**The element's own text is read as a fallback, and only in a browser.**\n`<helix-headless-format-date\nelapsed>2026-08-27T01:00:00Z</helix-headless-format-date>` is the spelling\nthat reads best in markup, and it cannot be the only one: measured against\nLit's server shim, `textContent`, `children`, `firstChild` and\n`querySelector` are all `undefined` there, so a server has no way to see\nit. An element rendered on a server needs the attribute; one rendered in a\nbrowser may use either.",
               "attribute": "date"
             },
             {
@@ -42752,7 +42772,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "The timestamp, as anything `Date` parses.\n\nAn ISO string is what an API sends and what `<time datetime>` takes, so it\nis the shape that needs no translation at either end. Left unset this\nrenders nothing rather than the epoch, which is what parsing an empty string\nwould give.\n\n**The element's own text is read as a fallback, and only in a browser.**\n`<helix-format-date elapsed>2026-08-27T01:00:00Z</helix-format-date>` is the\nspelling that reads best in markup, and it cannot be the only one: measured\nagainst Lit's server shim, `textContent`, `children`, `firstChild` and\n`querySelector` are all `undefined` there, so a server has no way to see it.\nAn element rendered on a server needs the attribute; one rendered in a\nbrowser may use either.",
+              "description": "The timestamp, as anything `Date` parses.\n\nAn ISO string is what an API sends and what `<time datetime>` takes, so it\nis the shape that needs no translation at either end. Left unset this\nrenders nothing rather than the epoch, which is what parsing an empty string\nwould give.\n\n**The element's own text is read as a fallback, and only in a browser.**\n`<helix-headless-format-date\nelapsed>2026-08-27T01:00:00Z</helix-headless-format-date>` is the spelling\nthat reads best in markup, and it cannot be the only one: measured against\nLit's server shim, `textContent`, `children`, `firstChild` and\n`querySelector` are all `undefined` there, so a server has no way to see\nit. An element rendered on a server needs the attribute; one rendered in a\nbrowser may use either.",
               "fieldName": "date"
             },
             {
@@ -43117,26 +43137,6 @@
           "description": "**One family covers both readings.** `rows=\"1\"` is a status strip, `rows=\"7\"`\nis a contribution graph, and a cell that takes a tone covers both: GitHub's\ncontribution graph is five discrete levels rather than a gradient.\n\n**What it cannot be is a bar chart with flat values**, and the data model is\nwhat closes that: chart color is per *series* everywhere else here, so ninety\ndifferently colored cells would be ninety series.\n\n**The data is a real `<table>` named by `for`**, the convention every chart\nin this library uses, and it is the accessible representation. Hide it with\n`data-hidden-visually` and the reading is still there.\n\nA cell takes its tone from its own `data-tone`, or from `thresholds` and\n`tones` bucketing its value. The cell's own wins, so a page that already\nknows a day was degraded does not encode that as a number to be decoded.\n\n**A tone is a name this tier never resolves.** `tone=\"success\"` writes\n`--heatmap-tone: var(--heatmap-tone-success)` and stops, so what a name is\nworth is the answering tier's to say. Reaching for `var(--helix-color-*)`\nhere is the dependency the tiers exist to prevent.",
           "name": "HelixHeadlessHeatmap",
           "cssProperties": [
-            {
-              "description": "How tall a cell is. A row of a tall thin cell is what a status page draws.",
-              "name": "--heatmap-cell-height"
-            },
-            {
-              "description": "How wide a cell is. The strip fits as many as the container allows.",
-              "name": "--heatmap-cell-inline-size"
-            },
-            {
-              "description": "How far a cell's corners are rounded.",
-              "name": "--heatmap-cell-radius"
-            },
-            {
-              "description": "The size a cell wants on both axes, where neither is given its own.",
-              "name": "--heatmap-cell-size"
-            },
-            {
-              "description": "Space between cells. Zero is a continuous field rather than a grid of tiles.",
-              "name": "--heatmap-gap"
-            },
             {
               "description": "How many rows the strip lays out. Written by the component from the data, read by the tier that draws it.",
               "name": "--heatmap-rows"
@@ -43330,7 +43330,7 @@
                 "text": "number"
               },
               "default": "8",
-              "description": "The smallest a cell may get before the strip draws fewer of them, in CSS pixels.\n\n**Set this only to overrule what the strip actually draws.** The arithmetic\nneeds a width in pixels and the width is a length the stylesheet resolves,\nso a number written here that disagrees with the sheet is a promise the\nconsumer has to keep by hand, and the one number a summary is told to trust\nis the one it gets wrong.\n\nThe strip measures itself, so `--heatmap-cell-inline-size` and `--heatmap-gap` are\nenough on their own and these two are the estimate used until the first\nmeasurement lands. Writing the attribute pins the value and stops the\nmeasurement being consulted, which is what a container query or a transition\non the cell size wants, where what is drawn this frame is not what the\narithmetic should plan around.\n\nZero turns the dropping off altogether, which is right for a strip in a\ncontainer that cannot get narrow.",
+              "description": "The smallest a cell may get before the strip draws fewer of them, in CSS pixels.\n\n**Set this only to overrule what the strip actually draws.** The arithmetic\nneeds a width in pixels and the width is a length the stylesheet resolves,\nso a number written here that disagrees with the sheet is a promise the\nconsumer has to keep by hand, and the one number a summary is told to trust\nis the one it gets wrong.\n\nThe strip measures itself, so sizing the cells in a sheet is enough on its\nown and these two are the estimate used until the first measurement lands.\nWriting the attribute pins the value and stops the measurement being\nconsulted, which is what a container query or a transition on the cell size\nwants, where what is drawn this frame is not what the arithmetic should\nplan around.\n\nZero turns the dropping off altogether, which is right for a strip in a\ncontainer that cannot get narrow.",
               "attribute": "min-cell"
             },
             {
@@ -43390,7 +43390,7 @@
               },
               "privacy": "private",
               "default": "undefined",
-              "description": "The cell and gap the strip is drawing, in pixels, once it has been read.\n\n**The arithmetic's inputs are a stylesheet's decisions, so the honest way to\nlearn them is to ask the stylesheet.** A cell's size is a length that only\nthe cascade can resolve: it may be a `rem` against a root font size, a\npercentage of a container, or whatever a container query chose this frame.\nEvery one of those is invisible to a number typed into an attribute, and\neach of `--heatmap-cell-inline-size`, `--heatmap-cell-height` and `--heatmap-gap`\nadded one more way for the two to drift apart silently.\n\nReading it back is not styling and does not breach the tier: this asks what\nwas decided rather than deciding, which is the same direction as the facts\nthis component writes out as custom properties, run backwards."
+              "description": "The cell and gap the strip is drawing, in pixels, once it has been read.\n\n**The arithmetic's inputs are a stylesheet's decisions, so the honest way\nto learn them is to ask the stylesheet.** A cell's size is a length that\nonly the cascade can resolve: it may be a `rem` against a root font size, a\npercentage of a container, or whatever a container query chose this frame.\nEvery one of those is invisible to a number typed into an attribute, and\nevery hook a sheet offers for the cell box or the gap is one more way for\nthe two to drift apart silently.\n\nReading it back is not styling and does not breach the tier: this asks what\nwas decided rather than deciding, which is the same direction as the facts\nthis component writes out as custom properties, run backwards."
             },
             {
               "kind": "field",
@@ -44037,7 +44037,7 @@
                 "text": "number"
               },
               "default": "8",
-              "description": "The smallest a cell may get before the strip draws fewer of them, in CSS pixels.\n\n**Set this only to overrule what the strip actually draws.** The arithmetic\nneeds a width in pixels and the width is a length the stylesheet resolves,\nso a number written here that disagrees with the sheet is a promise the\nconsumer has to keep by hand, and the one number a summary is told to trust\nis the one it gets wrong.\n\nThe strip measures itself, so `--heatmap-cell-inline-size` and `--heatmap-gap` are\nenough on their own and these two are the estimate used until the first\nmeasurement lands. Writing the attribute pins the value and stops the\nmeasurement being consulted, which is what a container query or a transition\non the cell size wants, where what is drawn this frame is not what the\narithmetic should plan around.\n\nZero turns the dropping off altogether, which is right for a strip in a\ncontainer that cannot get narrow.",
+              "description": "The smallest a cell may get before the strip draws fewer of them, in CSS pixels.\n\n**Set this only to overrule what the strip actually draws.** The arithmetic\nneeds a width in pixels and the width is a length the stylesheet resolves,\nso a number written here that disagrees with the sheet is a promise the\nconsumer has to keep by hand, and the one number a summary is told to trust\nis the one it gets wrong.\n\nThe strip measures itself, so sizing the cells in a sheet is enough on its\nown and these two are the estimate used until the first measurement lands.\nWriting the attribute pins the value and stops the measurement being\nconsulted, which is what a container query or a transition on the cell size\nwants, where what is drawn this frame is not what the arithmetic should\nplan around.\n\nZero turns the dropping off altogether, which is right for a strip in a\ncontainer that cannot get narrow.",
               "fieldName": "minCell"
             },
             {
@@ -44255,7 +44255,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
+              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-headless-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-headless-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
               "fieldName": "command",
               "inheritedFrom": {
                 "name": "HelixHeadlessButton",
@@ -44425,7 +44425,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
+              "description": "What to send when the button is activated, and what to send it to.\n\nThe platform's invoker attributes, forwarded onto the inner `<button>`\nrather than left on the host: only a real button is recognized as an\ninvoker, so `commandfor` written on a custom element resolves nothing and\nthe trigger silently does nothing at all. A wrapper has to hand them down,\nthe same way `label` becomes `aria-label` in there.\n\n```html\n<helix-headless-button commandfor=\"confirm\" command=\"--show\">Delete…</helix-headless-button>\n```\n\n`commandFor` spells its attribute out, because Lit lowercases a property\nname to derive one rather than dash-casing it, and what a consumer writes\nis the platform's `commandfor`.",
               "attribute": "command",
               "helix": {
                 "references": [
@@ -45836,7 +45836,7 @@
                 "text": "'error' | 'loaded' | 'loading'"
               },
               "default": "'loading'",
-              "description": "Whether the picture is on its way, here, or never coming.\n\nRead-only to a consumer: it is a report rather than a setting.\n\n**Mirrored to `data-status`, and not reflected as `status`.** It is element\nstate, so reflecting it puts the element's own bookkeeping into the markup\na consumer rendered: React reads the attribute appearing at upgrade as a\nhydration mismatch, and a serialized `<helix-image status=\"loaded\">` pasted\nback declares that a picture has arrived which this element has never\nfetched. The `data-` name is the library's answer to both, being a name a\nre-parse cannot mistake for a declaration, and it is what `data-closed` on\n`Toast.Item` already does. A picture arrives after somebody fetched it,\nwhich puts this on the safe side of hydration: a fact that can be true on\nthe *first* update mirrors to a custom state instead, which is what\n`:state(open)` and `:state(checked)` are.\n\n`loading` writes nothing, because it is where every image starts and the\nresting state of a mirror is the attribute's absence."
+              "description": "Whether the picture is on its way, here, or never coming.\n\nRead-only to a consumer: it is a report rather than a setting.\n\n**Mirrored to `data-status`, and not reflected as `status`.** It is element\nstate, so reflecting it puts the element's own bookkeeping into the markup\na consumer rendered: React reads the attribute appearing at upgrade as a\nhydration mismatch, and a host pasted back carrying `status=\"loaded\"`\ndeclares that a picture has arrived which this element has never fetched.\nThe `data-` name is the library's answer to both, being a name a re-parse\ncannot mistake for a declaration, and it is what `data-closed` on\n`Toast.Item` already does. A picture arrives after somebody fetched it,\nwhich puts this on the safe side of hydration: a fact that can be true on\nthe *first* update mirrors to a custom state instead, which is what\n`:state(open)` and `:state(checked)` are.\n\n`loading` writes nothing, because it is where every image starts and the\nresting state of a mirror is the attribute's absence."
             },
             {
               "kind": "field",
@@ -47371,7 +47371,7 @@
                 "text": "boolean"
               },
               "default": "false",
-              "description": "Whether the field refuses input while still submitting its value.\n\n**The attribute is `readonly`, which is the platform's spelling.** A\ncamelCase property needs an explicit `attribute`, and dash-casing a word\nHTML writes as one costs what an attribute always costs here: nothing\nwarns. `<helix-input readonly>` is what anybody who has written HTML types;\nagainst `read-only` it parses to no property at all, and the field stays\neditable with no error anywhere. Measured on the dash-cased form, `readOnly`\ncame back `false`.",
+              "description": "Whether the field refuses input while still submitting its value.\n\n**The attribute is `readonly`, which is the platform's spelling.** A\ncamelCase property needs an explicit `attribute`, and dash-casing a word\nHTML writes as one costs what an attribute always costs here: nothing\nwarns. `<helix-headless-input readonly>` is what anybody who has written\nHTML types; against `read-only` it parses to no property at all, and the\nfield stays editable with no error anywhere. Measured on the dash-cased\nform, `readOnly` came back `false`.",
               "attribute": "readonly",
               "reflects": true
             },
@@ -47585,7 +47585,7 @@
                 "text": "InputType"
               },
               "default": "'text'",
-              "description": "What the field accepts, which decides the keyboard and the validation.\n\n**Not reflected.** Reflection exists so a selector can reach a value, and\nno stylesheet in this repository selects on `type`. Reflecting it writes\n`type=\"text\"` onto every `<helix-input>` ever rendered, an attribute no\nconsumer's markup has, which is the fault the area rules write up under\n`Button`'s `href`, `target` and `rel`.\n\nNothing reads it from the attribute either. The real `<input>` inside takes\nit as a property binding, which is what makes the keyboard and the\nvalidation follow.",
+              "description": "What the field accepts, which decides the keyboard and the validation.\n\n**Not reflected.** Reflection exists so a selector can reach a value, and\nno stylesheet in this repository selects on `type`. Reflecting it writes\n`type=\"text\"` onto every `<helix-headless-input>` ever rendered, an\nattribute no consumer's markup has, which is the fault the area rules write\nup under `Button`'s `href`, `target` and `rel`.\n\nNothing reads it from the attribute either. The real `<input>` inside takes\nit as a property binding, which is what makes the keyboard and the\nvalidation follow.",
               "attribute": "type"
             },
             {
@@ -48524,7 +48524,7 @@
                 "text": "boolean"
               },
               "default": "false",
-              "description": "Whether the field refuses input while still submitting its value.\n\n**The attribute is `readonly`, which is the platform's spelling.** A\ncamelCase property needs an explicit `attribute`, and dash-casing a word\nHTML writes as one costs what an attribute always costs here: nothing\nwarns. `<helix-input readonly>` is what anybody who has written HTML types;\nagainst `read-only` it parses to no property at all, and the field stays\neditable with no error anywhere. Measured on the dash-cased form, `readOnly`\ncame back `false`.",
+              "description": "Whether the field refuses input while still submitting its value.\n\n**The attribute is `readonly`, which is the platform's spelling.** A\ncamelCase property needs an explicit `attribute`, and dash-casing a word\nHTML writes as one costs what an attribute always costs here: nothing\nwarns. `<helix-headless-input readonly>` is what anybody who has written\nHTML types; against `read-only` it parses to no property at all, and the\nfield stays editable with no error anywhere. Measured on the dash-cased\nform, `readOnly` came back `false`.",
               "fieldName": "readOnly"
             },
             {
@@ -48551,7 +48551,7 @@
                 "text": "InputType"
               },
               "default": "'text'",
-              "description": "What the field accepts, which decides the keyboard and the validation.\n\n**Not reflected.** Reflection exists so a selector can reach a value, and\nno stylesheet in this repository selects on `type`. Reflecting it writes\n`type=\"text\"` onto every `<helix-input>` ever rendered, an attribute no\nconsumer's markup has, which is the fault the area rules write up under\n`Button`'s `href`, `target` and `rel`.\n\nNothing reads it from the attribute either. The real `<input>` inside takes\nit as a property binding, which is what makes the keyboard and the\nvalidation follow.",
+              "description": "What the field accepts, which decides the keyboard and the validation.\n\n**Not reflected.** Reflection exists so a selector can reach a value, and\nno stylesheet in this repository selects on `type`. Reflecting it writes\n`type=\"text\"` onto every `<helix-headless-input>` ever rendered, an\nattribute no consumer's markup has, which is the fault the area rules write\nup under `Button`'s `href`, `target` and `rel`.\n\nNothing reads it from the attribute either. The real `<input>` inside takes\nit as a property binding, which is what makes the keyboard and the\nvalidation follow.",
               "fieldName": "type"
             },
             {
@@ -49968,7 +49968,7 @@
               "type": {
                 "text": "'none' | 'ring' | 'square' | 'triangle' | undefined"
               },
-              "description": "The shape of the mark on each series' last reading.\n\nAbsent is a dot, and absent rather than defaulted because a reflecting\nproperty with a default writes its attribute onto a host the consumer\nrendered, which React reports as a hydration mismatch.\n\n**One per series, in a column at the far edge**, because every series here\nshares an x-axis and therefore ends at the same reading. That column is the\nuseful thing: it is where a reader looks for \"and what is it now\", which is\nthe question a sparkline's single mark answers too.\n\n`none` lets the lines stop on their own, which suits a chart read for its\nshape rather than for where it finished. `square` is the same mark with its\ncorners, for a drawing whose neighbors are square, the columns of a table\nbeside it for instance. `ring` lets the ground through, for a chart drawn\nover something busy. `triangle` points along the axis and reads as \"and\nonward\" rather than as a second claim about the data.\n\nThe shapes come from\n`helix/styled/src/shared/chart/terminus.css`, the same sheet `Sparkline` takes, so\nthe set is identical wherever one appears. **Which is why every value is\nnamed here.** `Sparkline` shipped `square` implemented, drawn, asserted and\nundocumented until the cohort before this one; this family reads the same\nfour out of the same sheet and had left two of them unwritten, so the\nfinding arrived twice. A shared vocabulary wants naming at each door into\nit, because a reader arrives at one door and not at the set.\n\n`ring` is the one that asks something of the drawing. It leaves a hole and a\nline ends at the reading, which is the hole's center, so the `<svg>` is\nmasked with one disc per series and every stroke stops at its own ring's\ninner edge. `mask-composite: intersect` is what makes N holes out of N\nlayers; a single `mask-image` would only ever cut the last one.",
+              "description": "The shape of the mark on each series' last reading.\n\nAbsent is a dot, and absent rather than defaulted because a reflecting\nproperty with a default writes its attribute onto a host the consumer\nrendered, which React reports as a hydration mismatch.\n\n**One per series, in a column at the far edge**, because every series here\nshares an x-axis and therefore ends at the same reading. That column is the\nuseful thing: it is where a reader looks for \"and what is it now\", which is\nthe question a sparkline's single mark answers too.\n\n`none` lets the lines stop on their own, which suits a chart read for its\nshape rather than for where it finished. `square` is the same mark with its\ncorners, for a drawing whose neighbors are square, the columns of a table\nbeside it for instance. `ring` lets the ground through, for a chart drawn\nover something busy. `triangle` points along the axis and reads as \"and\nonward\" rather than as a second claim about the data.\n\nThe shapes come from one shared sheet, the same one `Sparkline` takes, so\nthe set is identical wherever one appears. **Which is why every value is\nnamed here.** `Sparkline` shipped `square` implemented, drawn, asserted and\nundocumented until the cohort before this one; this family reads the same\nfour out of the same sheet and had left two of them unwritten, so the\nfinding arrived twice. A shared vocabulary wants naming at each door into\nit, because a reader arrives at one door and not at the set.\n\n`ring` is the one that asks something of the drawing. It leaves a hole and a\nline ends at the reading, which is the hole's center, so the `<svg>` is\nmasked with one disc per series and every stroke stops at its own ring's\ninner edge. `mask-composite: intersect` is what makes N holes out of N\nlayers; a single `mask-image` would only ever cut the last one.",
               "attribute": "terminus",
               "reflects": true
             },
@@ -50728,7 +50728,7 @@
               "type": {
                 "text": "'none' | 'ring' | 'square' | 'triangle' | undefined"
               },
-              "description": "The shape of the mark on each series' last reading.\n\nAbsent is a dot, and absent rather than defaulted because a reflecting\nproperty with a default writes its attribute onto a host the consumer\nrendered, which React reports as a hydration mismatch.\n\n**One per series, in a column at the far edge**, because every series here\nshares an x-axis and therefore ends at the same reading. That column is the\nuseful thing: it is where a reader looks for \"and what is it now\", which is\nthe question a sparkline's single mark answers too.\n\n`none` lets the lines stop on their own, which suits a chart read for its\nshape rather than for where it finished. `square` is the same mark with its\ncorners, for a drawing whose neighbors are square, the columns of a table\nbeside it for instance. `ring` lets the ground through, for a chart drawn\nover something busy. `triangle` points along the axis and reads as \"and\nonward\" rather than as a second claim about the data.\n\nThe shapes come from\n`helix/styled/src/shared/chart/terminus.css`, the same sheet `Sparkline` takes, so\nthe set is identical wherever one appears. **Which is why every value is\nnamed here.** `Sparkline` shipped `square` implemented, drawn, asserted and\nundocumented until the cohort before this one; this family reads the same\nfour out of the same sheet and had left two of them unwritten, so the\nfinding arrived twice. A shared vocabulary wants naming at each door into\nit, because a reader arrives at one door and not at the set.\n\n`ring` is the one that asks something of the drawing. It leaves a hole and a\nline ends at the reading, which is the hole's center, so the `<svg>` is\nmasked with one disc per series and every stroke stops at its own ring's\ninner edge. `mask-composite: intersect` is what makes N holes out of N\nlayers; a single `mask-image` would only ever cut the last one.",
+              "description": "The shape of the mark on each series' last reading.\n\nAbsent is a dot, and absent rather than defaulted because a reflecting\nproperty with a default writes its attribute onto a host the consumer\nrendered, which React reports as a hydration mismatch.\n\n**One per series, in a column at the far edge**, because every series here\nshares an x-axis and therefore ends at the same reading. That column is the\nuseful thing: it is where a reader looks for \"and what is it now\", which is\nthe question a sparkline's single mark answers too.\n\n`none` lets the lines stop on their own, which suits a chart read for its\nshape rather than for where it finished. `square` is the same mark with its\ncorners, for a drawing whose neighbors are square, the columns of a table\nbeside it for instance. `ring` lets the ground through, for a chart drawn\nover something busy. `triangle` points along the axis and reads as \"and\nonward\" rather than as a second claim about the data.\n\nThe shapes come from one shared sheet, the same one `Sparkline` takes, so\nthe set is identical wherever one appears. **Which is why every value is\nnamed here.** `Sparkline` shipped `square` implemented, drawn, asserted and\nundocumented until the cohort before this one; this family reads the same\nfour out of the same sheet and had left two of them unwritten, so the\nfinding arrived twice. A shared vocabulary wants naming at each door into\nit, because a reader arrives at one door and not at the set.\n\n`ring` is the one that asks something of the drawing. It leaves a hole and a\nline ends at the reading, which is the hole's center, so the `<svg>` is\nmasked with one disc per series and every stroke stops at its own ring's\ninner edge. `mask-composite: intersect` is what makes N holes out of N\nlayers; a single `mask-image` would only ever cut the last one.",
               "fieldName": "terminus"
             },
             {
@@ -50930,7 +50930,7 @@
                   "text": "TemplateResult | typeof nothing"
                 }
               },
-              "description": "The cell before the row's content, which this tier leaves empty.\n\nA railed list puts its number at the block's own start edge, in a track of\nits own ahead of the words, and the native `::marker` cannot reach there:\n`outside` hangs the marker just clear of the row's content box, which is\nbeside the words rather than at the edge of the block. A number that has to\nsit in a track has to be a box in that track.\n\n**The box belongs to the styled tier, which is why this returns `nothing`.**\nRendering it here puts an unstyled empty box in front of every headless row,\nand a row styled with `display: flex` and a `gap` draws that gap beside the\nempty child: measured on this tier's own example, which is exactly such a\nrow, 1384 pixels moved and every label sat half a rem further in. The styled\nrow's own comment records the same fault about a column gap beside a\nzero-width column.\n\nA tier that ships no CSS cannot make a box it renders cost nothing, so it\ndoes not render one. `marker=\"rail\"` is a styled-tier attribute for the same\nreason: *that* a list is numbered is a fact about the content and *where the\nnumbers are drawn* is not. A headless consumer wanting a rail writes the\ncell they were going to style anyway.\n\n`protected` rather than `private`, so the tier above can reach it."
+              "description": "The cell before the row's content, which this tier leaves empty.\n\nA railed list puts its number at the block's own start edge, in a track of\nits own ahead of the words, and the native `::marker` cannot reach there:\n`outside` hangs the marker just clear of the row's content box, which is\nbeside the words rather than at the edge of the block. A number that has to\nsit in a track has to be a box in that track.\n\n**The box belongs to the styled tier, which is why this returns\n`nothing`.** Rendering it here puts an unstyled empty box in front of every\nheadless row, and a row styled with `display: flex` and a `gap` draws that\ngap beside the empty child: measured on this tier's own example, which is\nexactly such a row, 1384 pixels moved and every label sat half a rem\nfurther in.\n\nA tier that ships no CSS cannot make a box it renders cost nothing, so it\ndoes not render one. `marker=\"rail\"` is a styled-tier attribute for the same\nreason: *that* a list is numbered is a fact about the content and *where the\nnumbers are drawn* is not. A headless consumer wanting a rail writes the\ncell they were going to style anyway.\n\n`protected` rather than `private`, so the tier above can reach it."
             },
             {
               "kind": "field",
@@ -51379,7 +51379,7 @@
               "type": {
                 "text": "'1' | 'a' | 'A' | 'i' | 'I' | undefined"
               },
-              "description": "Which markers an ordered list draws, in `<ol type>`'s own five values.\n\nReflected, because this is one of the things a stylesheet selects on and\nthe styled tier maps each to a `list-style-type`. The platform's spellings\nrather than the CSS ones, so `<helix-list kind=\"ordered\" type=\"i\">` is what\na consumer would have written on an `<ol>`.\n\n**`start` and `reversed` are both absent and neither can be done here.**\nThis renders no `<ol>`: the rows are slotted and numbered by the CSS\n`list-item` counter, and that counter is not author-resettable. Measured in\nChrome 148, `counter-reset: list-item 4` moves nothing on a `<div>`, on a\n`<ul>`, or even on a real `<ol>`, while `<ol start=\"5\">` does and a *named*\ncounter with explicit `::before` content does. So the markers would have to\nbe rebuilt out of generated content to gain a start, which is a different\ncomponent. `type` is only `list-style-type` and works.",
+              "description": "Which markers an ordered list draws, in `<ol type>`'s own five values.\n\nReflected, because this is one of the things a stylesheet selects on, each\nvalue mapping to a `list-style-type`. The platform's spellings rather than\nthe CSS ones, so `<helix-headless-list kind=\"ordered\" type=\"i\">` is what a\nconsumer would have written on an `<ol>`.\n\n**`start` and `reversed` are both absent and neither can be done here.**\nThis renders no `<ol>`: the rows are slotted and numbered by the CSS\n`list-item` counter, and that counter is not author-resettable. Measured in\nChrome 148, `counter-reset: list-item 4` moves nothing on a `<div>`, on a\n`<ul>`, or even on a real `<ol>`, while `<ol start=\"5\">` does and a *named*\ncounter with explicit `::before` content does. So the markers would have to\nbe rebuilt out of generated content to gain a start, which is a different\ncomponent. `type` is only `list-style-type` and works.",
               "attribute": "type",
               "reflects": true
             },
@@ -51731,7 +51731,7 @@
               "type": {
                 "text": "'1' | 'a' | 'A' | 'i' | 'I' | undefined"
               },
-              "description": "Which markers an ordered list draws, in `<ol type>`'s own five values.\n\nReflected, because this is one of the things a stylesheet selects on and\nthe styled tier maps each to a `list-style-type`. The platform's spellings\nrather than the CSS ones, so `<helix-list kind=\"ordered\" type=\"i\">` is what\na consumer would have written on an `<ol>`.\n\n**`start` and `reversed` are both absent and neither can be done here.**\nThis renders no `<ol>`: the rows are slotted and numbered by the CSS\n`list-item` counter, and that counter is not author-resettable. Measured in\nChrome 148, `counter-reset: list-item 4` moves nothing on a `<div>`, on a\n`<ul>`, or even on a real `<ol>`, while `<ol start=\"5\">` does and a *named*\ncounter with explicit `::before` content does. So the markers would have to\nbe rebuilt out of generated content to gain a start, which is a different\ncomponent. `type` is only `list-style-type` and works.",
+              "description": "Which markers an ordered list draws, in `<ol type>`'s own five values.\n\nReflected, because this is one of the things a stylesheet selects on, each\nvalue mapping to a `list-style-type`. The platform's spellings rather than\nthe CSS ones, so `<helix-headless-list kind=\"ordered\" type=\"i\">` is what a\nconsumer would have written on an `<ol>`.\n\n**`start` and `reversed` are both absent and neither can be done here.**\nThis renders no `<ol>`: the rows are slotted and numbered by the CSS\n`list-item` counter, and that counter is not author-resettable. Measured in\nChrome 148, `counter-reset: list-item 4` moves nothing on a `<div>`, on a\n`<ul>`, or even on a real `<ol>`, while `<ol start=\"5\">` does and a *named*\ncounter with explicit `::before` content does. So the markers would have to\nbe rebuilt out of generated content to gain a start, which is a different\ncomponent. `type` is only `list-style-type` and works.",
               "fieldName": "type"
             }
           ],
@@ -54222,7 +54222,7 @@
                   "text": "number"
                 }
               },
-              "description": "The bottom edge of the bar this hangs from.\n\nThe trigger's own bottom is the obvious answer and is the wrong one: a\nheader is taller than the word inside it, so a band starting at the word\nstarts partway up the header, over its lower padding and its border.\nMeasured: trigger bottom 44, header bottom 57, thirteen pixels of header\nwith a band drawn across it.\n\n**The first ancestor as wide as the window, which is a fact and not a tag\nname.** A band spans the viewport, so what it should hang from is whatever\nelse does: that is what makes something a bar. Asking for `helix-app-header`\nby name would make this element know about that one, which is the coupling\n`Menu` is careful not to have with `Drilldown`, and would leave a consumer\nwho wrote their own header with a band across it.\n\n`body` ends the walk. It is as wide as the window and is not a bar, and its\nbottom is the foot of the page: without the guard a megamenu on a plain page\nwould put its band down there. Nothing found means nothing to hang from, and\nthe trigger is the honest answer again."
+              "description": "The bottom edge of the bar this hangs from.\n\nThe trigger's own bottom is the obvious answer and is the wrong one: a\nheader is taller than the word inside it, so a band starting at the word\nstarts partway up the header, over its lower padding and its border.\nMeasured: trigger bottom 44, header bottom 57, thirteen pixels of header\nwith a band drawn across it.\n\n**The first ancestor as wide as the window, which is a fact and not a tag\nname.** A band spans the viewport, so what it should hang from is whatever\nelse does: that is what makes something a bar. Asking for `AppHeader` by\nname would make this element know about that one, which is the coupling\n`Menu` is careful not to have with `Drilldown`, and would leave a consumer\nwho wrote their own header with a band across it.\n\n`body` ends the walk. It is as wide as the window and is not a bar, and its\nbottom is the foot of the page: without the guard a megamenu on a plain page\nwould put its band down there. Nothing found means nothing to hang from, and\nthe trigger is the honest answer again."
             },
             {
               "kind": "field",
@@ -54344,7 +54344,7 @@
                   "text": "TemplateResult"
                 }
               },
-              "description": "The control that opens it, behind a seam the styled tier can move.\n\nThe same case `SpeedDial` records: a bare `<button>` is what this tier can\nwrite, `helix-button` lives a tier above, and a render method is how a tier\nthat has one reaches this without reimplementing the panel below it."
+              "description": "The control that opens it, behind a seam the styled tier can move.\n\nThe same case `SpeedDial` records: a bare `<button>` is what this tier can\nwrite, `Button` lives a tier above, and a render method is how a tier that\nhas one reaches this without reimplementing the panel below it."
             },
             {
               "kind": "field",
@@ -58749,7 +58749,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "A notification is a toast that did not go away. `helix-toast` is deliberately\nthe opposite of this: a live region whose items carry a `duration` and whose\nroot never creates or destroys anything. What makes this its own family is the\nstate each item carries rather than the arrangement it sits in: read and\nunread, a timestamp that has to stay correct, and an action a reader can take\nlater.\n\n**The row is a label, a caption under it, and one trailing column that holds\neither the time or the controls.** A feed is scanned rather than read, so the\nthing that happened gets the first line on its own and everything qualifying\nit goes under. The time and the controls occupy the same place because a\nreader wants exactly one of them at a time: the time while they are scanning,\nand what they can do about it once they have stopped on a row.\n\n**The timestamp is `helix-format-date`**, which is where the reading, the\n`<time datetime>` and the shared timer now live. It was implemented here, and\na component that formats a date is not a fact about notifications: Chat wanted\nthe same thing and had its own copy of half of it.\n\n**Read state is reported, not invented.** Marking something read is usually a\nround trip, and a component that flips its own state and then hears the\nrequest failed has told the reader something untrue. `helix-read` is\ncancelable, and a `read` set by the consumer makes this element defer entirely.",
+          "description": "A notification is a toast that did not go away. `Toast` is deliberately the\nopposite of this: a live region whose items carry a `duration` and whose root\nnever creates or destroys anything. What makes this its own family is the\nstate each item carries rather than the arrangement it sits in: read and\nunread, a timestamp that has to stay correct, and an action a reader can take\nlater.\n\n**The row is a label, a caption under it, and one trailing column that holds\neither the time or the controls.** A feed is scanned rather than read, so the\nthing that happened gets the first line on its own and everything qualifying\nit goes under. The time and the controls occupy the same place because a\nreader wants exactly one of them at a time: the time while they are scanning,\nand what they can do about it once they have stopped on a row.\n\n**The timestamp is `FormatDate`**, which is where the reading, the `<time\ndatetime>` and the shared timer now live. It was implemented here, and a\ncomponent that formats a date is not a fact about notifications: Chat wanted\nthe same thing and had its own copy of half of it.\n\n**Read state is reported, not invented.** Marking something read is usually a\nround trip, and a component that flips its own state and then hears the\nrequest failed has told the reader something untrue. `helix-read` is\ncancelable, and a `read` set by the consumer makes this element defer entirely.",
           "name": "HelixHeadlessNotification",
           "cssParts": [
             {
@@ -58836,7 +58836,7 @@
               "type": {
                 "text": "boolean"
               },
-              "description": "Whether the reader has seen it.\n\nSet it and this element defers: a reader's click fires `helix-read` and\nchanges nothing until a new value arrives. `default-read` seeds a state\nthis element then owns.\n\n**Not reflected.** Unread is a visual state, which is why the four mixins\nmirror theirs, but the mirror may not be the property's own attribute:\n`<helix-notification default-read>` comes back from a server without `read`\nand would gain it on the first update, which React reports as a hydration\nmismatch, and which a copy of the serialized element re-parses as a\ndeclaration. `:state(read)` is what the stylesheet reads.\n`hydration.dom.spec.ts` holds it.",
+              "description": "Whether the reader has seen it.\n\nSet it and this element defers: a reader's click fires `helix-read` and\nchanges nothing until a new value arrives. `default-read` seeds a state\nthis element then owns.\n\n**Not reflected.** Unread is a visual state, which is why the four mixins\nmirror theirs, but the mirror may not be the property's own attribute:\n`<helix-headless-notification default-read>` comes back from a server\nwithout `read` and would gain it on the first update, which React reports\nas a hydration mismatch, and which a copy of the serialized element\nre-parses as a declaration. `:state(read)` is what the stylesheet reads.\n`hydration.dom.spec.ts` holds it.",
               "attribute": "read"
             },
             {
@@ -59362,7 +59362,7 @@
               "type": {
                 "text": "boolean"
               },
-              "description": "Whether the reader has seen it.\n\nSet it and this element defers: a reader's click fires `helix-read` and\nchanges nothing until a new value arrives. `default-read` seeds a state\nthis element then owns.\n\n**Not reflected.** Unread is a visual state, which is why the four mixins\nmirror theirs, but the mirror may not be the property's own attribute:\n`<helix-notification default-read>` comes back from a server without `read`\nand would gain it on the first update, which React reports as a hydration\nmismatch, and which a copy of the serialized element re-parses as a\ndeclaration. `:state(read)` is what the stylesheet reads.\n`hydration.dom.spec.ts` holds it.",
+              "description": "Whether the reader has seen it.\n\nSet it and this element defers: a reader's click fires `helix-read` and\nchanges nothing until a new value arrives. `default-read` seeds a state\nthis element then owns.\n\n**Not reflected.** Unread is a visual state, which is why the four mixins\nmirror theirs, but the mirror may not be the property's own attribute:\n`<helix-headless-notification default-read>` comes back from a server\nwithout `read` and would gain it on the first update, which React reports\nas a hydration mismatch, and which a copy of the serialized element\nre-parses as a declaration. `:state(read)` is what the stylesheet reads.\n`hydration.dom.spec.ts` holds it.",
               "fieldName": "read"
             },
             {
@@ -59517,7 +59517,7 @@
           ],
           "slots": [
             {
-              "description": "The notifications, as `helix-notification` children.",
+              "description": "The notifications.",
               "name": ""
             },
             {
@@ -63405,7 +63405,7 @@
               "type": {
                 "text": "boolean"
               },
-              "description": "Whether the markup declared this one chosen, which is what a reset returns to.\n\n`<option selected>` reflects to `defaultSelected` and `option.selected` is\nlive state that does not reflect, and those are two different facts. This\nelement has one `selected` doing both jobs, so the declaration is the\nattribute and the state is the property.\n\n**The attribute, read when it is asked for, and never a snapshot taken in\nthe constructor.** A field initializer runs there, and an element the HTML\nparser created has none of its attributes yet at that moment: measured on\nthis element, on `Radio` and on `Toggle`, `<helix-option value=\"a\"\nselected>` reported `defaultSelected` as `false` both from parsed markup\nand from `createElement` followed by `setAttribute`, which is every way a\nconsumer writes one. The snapshot could only ever be true for an element\nalready in the document when `customElements.define` ran.\n\nNothing overwrites it either, which is what makes reading it late correct\nrather than merely later: `selected` does not reflect, for the reason\n`willUpdate` records, so the listbox's writes never reach the attribute and\nwhat it says is what the markup said.\n\n**The deeper half is not this.** State reflecting under the declaration's\nname is the fault `TreeItem` avoids by mirroring to `data-selected`.\nSplitting them means moving what every stylesheet selects on, which is a\nchange of its own rather than a property being added."
+              "description": "Whether the markup declared this one chosen, which is what a reset returns to.\n\n`<option selected>` reflects to `defaultSelected` and `option.selected` is\nlive state that does not reflect, and those are two different facts. This\nelement has one `selected` doing both jobs, so the declaration is the\nattribute and the state is the property.\n\n**The attribute, read when it is asked for, and never a snapshot taken in\nthe constructor.** A field initializer runs there, and an element the HTML\nparser created has none of its attributes yet at that moment: measured on\nthis element, on `Radio` and on `Toggle`, `<helix-headless-option value=\"a\"\nselected>` reported `defaultSelected` as `false` both from parsed markup\nand from `createElement` followed by `setAttribute`, which is every way a\nconsumer writes one. The snapshot could only ever be true for an element\nalready in the document when `customElements.define` ran.\n\nNothing overwrites it either, which is what makes reading it late correct\nrather than merely later: `selected` does not reflect, for the reason\n`willUpdate` records, so the listbox's writes never reach the attribute and\nwhat it says is what the markup said.\n\n**The deeper half is not this.** State reflecting under the declaration's\nname is the fault `TreeItem` avoids by mirroring to `data-selected`.\nSplitting them means moving what every stylesheet selects on, which is a\nchange of its own rather than a property being added."
             },
             {
               "kind": "field",
@@ -64676,7 +64676,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The component is the truncation.** Given a page and a total it works out\nwhich numbers to show and where the runs break, so this is data-driven rather\nthan composed: `<helix-pagination page=\"4\" pages=\"12\">` rather than twelve\nslotted children a consumer keeps in step with their own query.\n\n**The width does not change as you page.** The run around the current page is\n`siblings` either side, and grows to absorb a gap slot at a boundary, so the\ncontrol holds the same number of slots from the first page to the last.\nAnything that lets the count vary moves the buttons under the pointer between\none click and the next.\n\n**A gap is only drawn where it hides more than one page, and a gap is a\ncontrol.** It goes to the middle of the run it stands over, which is the\nbinary-search step: any page in a set of any size is a handful of presses.\n\n**`href` is what makes these links.** Set it and every page renders an `<a>`\nwith `{page}` filled in, so the output is crawlable, a middle click opens a\ntab, and the address bar is the state. Left unset they are buttons, which is\nright for a table paging in place. Either way the click announces\n`helix-navigate` first, and `preventDefault()` stops the navigation as well\nas the change, which is how a router takes over.\n\n**`helix-navigate` and not `helix-change`.** Moving to page four chooses\nnothing; it moves the view over a result that has not changed.\n\n**The count and the page-size control are their own elements.**\n`Pagination.Summary` and `Pagination.Size` sit beside this one, because a\nslot can only offer one place: at one end of one row, in one order.",
+          "description": "**The component is the truncation.** Given a page and a total it works out\nwhich numbers to show and where the runs break, so this is data-driven rather\nthan composed: `<helix-headless-pagination page=\"4\" pages=\"12\">` rather than\ntwelve slotted children a consumer keeps in step with their own query.\n\n**The width does not change as you page.** The run around the current page is\n`siblings` either side, and grows to absorb a gap slot at a boundary, so the\ncontrol holds the same number of slots from the first page to the last.\nAnything that lets the count vary moves the buttons under the pointer between\none click and the next.\n\n**A gap is only drawn where it hides more than one page, and a gap is a\ncontrol.** It goes to the middle of the run it stands over, which is the\nbinary-search step: any page in a set of any size is a handful of presses.\n\n**`href` is what makes these links.** Set it and every page renders an `<a>`\nwith `{page}` filled in, so the output is crawlable, a middle click opens a\ntab, and the address bar is the state. Left unset they are buttons, which is\nright for a table paging in place. Either way the click announces\n`helix-navigate` first, and `preventDefault()` stops the navigation as well\nas the change, which is how a router takes over.\n\n**`helix-navigate` and not `helix-change`.** Moving to page four chooses\nnothing; it moves the view over a result that has not changed.\n\n**The count and the page-size control are their own elements.**\n`Pagination.Summary` and `Pagination.Size` sit beside this one, because a\nslot can only offer one place: at one end of one row, in one order.",
           "name": "HelixHeadlessPagination",
           "cssParts": [
             {
@@ -67019,7 +67019,7 @@
                 "text": "PinType"
               },
               "default": "'numeric'",
-              "description": "What a box accepts, which decides the keyboard a phone offers.\n\n`numeric` and `alphanumeric` also refuse anything else on the way in,\nbecause a code that is digits is digits: letting a letter into a box a\nreader cannot submit is a mistake they have to find and undo themselves.\n`text` refuses nothing and only trims, so it is the mode for a code with a\nhyphen in it.\n\n**`alphanumeric` is ASCII letters and digits**, which is narrower than the\nword is in general and is the right narrowness here. A code of this kind\ncomes off a keypad, an SMS or an authenticator, and none of them emit `é`;\na filter that accepted it would let through a character the service on the\nother end is going to reject. A consumer whose codes really do carry\nmarked letters wants `text`, which judges nothing.\n\n**Not reflected**: nothing selects on it, so reflecting only writes\n`type=\"numeric\"` onto every `<helix-pin-input>` ever rendered, an attribute\nno consumer's markup had. The area rules carry the sentence \"reflect only\nwhat CSS selects on\". The real inputs take it as a binding, which is what\nmakes the keyboard follow.",
+              "description": "What a box accepts, which decides the keyboard a phone offers.\n\n`numeric` and `alphanumeric` also refuse anything else on the way in,\nbecause a code that is digits is digits: letting a letter into a box a\nreader cannot submit is a mistake they have to find and undo themselves.\n`text` refuses nothing and only trims, so it is the mode for a code with a\nhyphen in it.\n\n**`alphanumeric` is ASCII letters and digits**, which is narrower than the\nword is in general and is the right narrowness here. A code of this kind\ncomes off a keypad, an SMS or an authenticator, and none of them emit `é`;\na filter that accepted it would let through a character the service on the\nother end is going to reject. A consumer whose codes really do carry\nmarked letters wants `text`, which judges nothing.\n\n**Not reflected**: nothing selects on it, so reflecting only writes\n`type=\"numeric\"` onto every `<helix-headless-pin-input>` ever rendered, an\nattribute no consumer's markup had. The area rules carry the sentence\n\"reflect only what CSS selects on\". The real inputs take it as a binding,\nwhich is what makes the keyboard follow.",
               "attribute": "type"
             },
             {
@@ -68170,7 +68170,7 @@
                 "text": "PinType"
               },
               "default": "'numeric'",
-              "description": "What a box accepts, which decides the keyboard a phone offers.\n\n`numeric` and `alphanumeric` also refuse anything else on the way in,\nbecause a code that is digits is digits: letting a letter into a box a\nreader cannot submit is a mistake they have to find and undo themselves.\n`text` refuses nothing and only trims, so it is the mode for a code with a\nhyphen in it.\n\n**`alphanumeric` is ASCII letters and digits**, which is narrower than the\nword is in general and is the right narrowness here. A code of this kind\ncomes off a keypad, an SMS or an authenticator, and none of them emit `é`;\na filter that accepted it would let through a character the service on the\nother end is going to reject. A consumer whose codes really do carry\nmarked letters wants `text`, which judges nothing.\n\n**Not reflected**: nothing selects on it, so reflecting only writes\n`type=\"numeric\"` onto every `<helix-pin-input>` ever rendered, an attribute\nno consumer's markup had. The area rules carry the sentence \"reflect only\nwhat CSS selects on\". The real inputs take it as a binding, which is what\nmakes the keyboard follow.",
+              "description": "What a box accepts, which decides the keyboard a phone offers.\n\n`numeric` and `alphanumeric` also refuse anything else on the way in,\nbecause a code that is digits is digits: letting a letter into a box a\nreader cannot submit is a mistake they have to find and undo themselves.\n`text` refuses nothing and only trims, so it is the mode for a code with a\nhyphen in it.\n\n**`alphanumeric` is ASCII letters and digits**, which is narrower than the\nword is in general and is the right narrowness here. A code of this kind\ncomes off a keypad, an SMS or an authenticator, and none of them emit `é`;\na filter that accepted it would let through a character the service on the\nother end is going to reject. A consumer whose codes really do carry\nmarked letters wants `text`, which judges nothing.\n\n**Not reflected**: nothing selects on it, so reflecting only writes\n`type=\"numeric\"` onto every `<helix-headless-pin-input>` ever rendered, an\nattribute no consumer's markup had. The area rules carry the sentence\n\"reflect only what CSS selects on\". The real inputs take it as a binding,\nwhich is what makes the keyboard follow.",
               "fieldName": "type"
             },
             {
@@ -68434,7 +68434,7 @@
               "name": "panel",
               "privacy": "private",
               "readonly": true,
-              "default": "new PopoverController(this, { anchor: () => this.trigger, /* The host *and* whatever invoked it, never the panel alone. An outside press is measured against this, and the styled host is `display: contents` with no box of its own, so a boundary of the panel alone would close on the very press that opened it. A `--toggle` button outside the host is part of the control by any reading a reader would give it: pressing it while the panel is open should close it once, not close it on `pointerdown` and reopen it on the `click` that follows. */ boundary: (): (HTMLElement | undefined)[] => [this, this.invoker], offset: () => this.offset, onDismiss: () => this.hide(), open: () => this.open, placement: () => this.placement, surface: () => this.partNamed('popover'), })"
+              "default": "new PopoverController(this, { anchor: () => this.trigger, /* The host *and* whatever invoked it, never the panel alone. An outside press is measured against this, and the host is not a box to rely on, a sheet being free to give it none, so a boundary of the panel alone would close on the very press that opened it. A `--toggle` button outside the host is part of the control by any reading a reader would give it: pressing it while the panel is open should close it once, not close it on `pointerdown` and reopen it on the `click` that follows. */ boundary: (): (HTMLElement | undefined)[] => [this, this.invoker], offset: () => this.offset, onDismiss: () => this.hide(), open: () => this.open, placement: () => this.placement, surface: () => this.partNamed('popover'), })"
             },
             {
               "kind": "field",
@@ -68443,7 +68443,7 @@
                 "text": "HTMLElement | undefined"
               },
               "privacy": "private",
-              "description": "The invoker that last opened the panel, when a command did.\n\nKept because the `source` a command hands the handlers is needed twice\nover. It is the anchor: the styled host is a boxless `display: contents`\nelement, so a popover opened by a command outside it would otherwise open\nagainst the window origin. And the outside-press dismissal has to know the\ninvoker is part of the control, the press firing on `pointerdown` and the\ncommand arriving on the `click` after it, or a `--toggle` button outside\nthe host closes the panel and immediately reopens it, for ever."
+              "description": "The invoker that last opened the panel, when a command did.\n\nKept because the `source` a command hands the handlers is needed twice\nover. It is the anchor: the host is not a box to position against, a sheet\nbeing free to give it none, so a popover opened by a command outside it\nwould otherwise open against the window origin. And the outside-press\ndismissal has to know the invoker is part of the control, the press firing\non `pointerdown` and the command arriving on the `click` after it, or a\n`--toggle` button outside the host closes the panel and immediately reopens\nit, for ever."
             },
             {
               "kind": "field",
@@ -71895,7 +71895,7 @@
               "type": {
                 "text": "boolean"
               },
-              "description": "Whether the markup declared this one chosen, which is what a reset returns to.\n\n`<option selected>` reflects to `defaultSelected` and `option.selected` is\nlive state that does not reflect, and those are two different facts. This\nelement has one `selected` doing both jobs, so the declaration is the\nattribute and the state is the property.\n\n**The attribute, read when it is asked for, and never a snapshot taken in\nthe constructor.** A field initializer runs there, and an element the HTML\nparser created has none of its attributes yet at that moment: measured on\nthis element, on `Option` and on `Toggle`, `<helix-radio value=\"a\"\nselected>` reported `defaultSelected` as `false` both from parsed markup\nand from `createElement` followed by `setAttribute`, which is every way a\nconsumer writes one. The snapshot could only ever be true for an element\nalready in the document when `customElements.define` ran.\n\nNothing overwrites it either, which is what makes reading it late correct\nrather than merely later: `selected` does not reflect, for the reason\n`willUpdate` records, so the group's writes never reach the attribute and\nwhat it says is what the markup said.\n\n**The deeper half is not this.** State reflecting under the declaration's\nname is the fault `TreeItem` avoids by mirroring to `data-selected`.\nSplitting them means moving what every stylesheet selects on, which is a\nchange of its own rather than a property being added."
+              "description": "Whether the markup declared this one chosen, which is what a reset returns to.\n\n`<option selected>` reflects to `defaultSelected` and `option.selected` is\nlive state that does not reflect, and those are two different facts. This\nelement has one `selected` doing both jobs, so the declaration is the\nattribute and the state is the property.\n\n**The attribute, read when it is asked for, and never a snapshot taken in\nthe constructor.** A field initializer runs there, and an element the HTML\nparser created has none of its attributes yet at that moment: measured on\nthis element, on `Option` and on `Toggle`, `<helix-headless-radio value=\"a\"\nselected>` reported `defaultSelected` as `false` both from parsed markup\nand from `createElement` followed by `setAttribute`, which is every way a\nconsumer writes one. The snapshot could only ever be true for an element\nalready in the document when `customElements.define` ran.\n\nNothing overwrites it either, which is what makes reading it late correct\nrather than merely later: `selected` does not reflect, for the reason\n`willUpdate` records, so the group's writes never reach the attribute and\nwhat it says is what the markup said.\n\n**The deeper half is not this.** State reflecting under the declaration's\nname is the fault `TreeItem` avoids by mirroring to `data-selected`.\nSplitting them means moving what every stylesheet selects on, which is a\nchange of its own rather than a property being added."
             },
             {
               "kind": "field",
@@ -75607,10 +75607,10 @@
               "kind": "field",
               "name": "display",
               "type": {
-                "text": "string"
+                "text": "string | typeof nothing"
               },
               "privacy": "private",
-              "description": "What the trigger says: the chosen options' words, or the placeholder.\n\nRead off the options rather than stored, because the words are the\nconsumer's content and storing a copy is storing something that goes stale\nthe moment they change it.\n\n**Off this element's own children, not the listbox's**, which reach the\nlistbox through a slot and so only after a render has made it: read there,\nthe first render showed the placeholder over a chosen value and waited on a\n`slotchange`, which a declarative shadow root fires before anything listens,\nso a hydrated select kept \"Choose\" for good. Nothing on a server or while\nhydrating what one sent, which drew the placeholder too.",
+              "description": "What the trigger says: the chosen options' words, or the placeholder.\n\nRead off the options rather than stored, because the words are the\nconsumer's content and storing a copy is storing something that goes stale\nthe moment they change it.\n\n**Off this element's own children, not the listbox's**, which reach the\nlistbox through a slot and so only after a render has made it: read there,\nthe first render showed the placeholder over a chosen value and waited on a\n`slotchange`, which a declarative shadow root fires before anything listens,\nso a hydrated select kept \"Choose\" for good. Nothing on a server or while\nhydrating what one sent, which drew the placeholder too.\n\n**And the placeholder is a statement, so a render that cannot read the\nwords does not make it.** \"Choose\" means nothing is chosen. `value` is an\nattribute and a server has it, the words are the children's and a server\nhas none, so a `<helix-select default-value=\"two\">` served \"Choose\" over an\nanswer the consumer had already given: false to a reader, and false to a\ncrawler, on every select in the library's own examples. Where a value is\ndeclared and the word for it cannot be read, the trigger says nothing and\nkeeps its box, which is the answer this library reaches for everywhere it\ncannot tell empty from unknown. `firstUpdated` asks for the render that\nfills it in.\n\nThe placeholder still stands where there is genuinely no value, which is\nwhat it is for and the one case a server can be sure of.\n\n**`nothing` and never `''`, which is the difference between a word that\narrives and one that never does.** Lit's hydration primes a child part with\nwhatever the render computed, and `''` is a primitive, so `_commitText`\ntakes the branch that assumes the server left a text node to overwrite.\nThere is none, the write lands on the closing part marker's comment data,\nand the trigger stays empty for the life of the page: measured by\n`check:spec:dsd`, which caught all nine of these and showed a blank select\nbeside a client's \"Option Two\". `nothing` primes the sentinel instead, so\nthe commit creates the text node it needs.",
               "readonly": true
             },
             {
@@ -80415,7 +80415,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**A menu, and deliberately not a new one.** Its rows are commands, which is\nwhat `role=\"menu\"` and `menuitem` exist for, so this takes them and takes the\nbehavior with them: the arrow keys are `KeyboardNavigationController`, the\npanel is `PopoverController`, and open and closed is `MixinControllerOpen`.\n`Menu` is built from the same pieces. What is not shared is what actually\ndiffers: a dial has no submenus, no headers, no checkable rows and no long\nlist to type ahead through, and shipping it as an appearance of `Menu` would\nmean defining what a dial with a submenu in it does.\n\n**The trigger is a control the consumer slots**, and the floating action\nbutton is the usual one: `helix-icon-button` with `shape=\"circle\"` and\n`raised`. There is no `helix-fab`, because it would re-declare `emphasis`,\n`tone`, `label` and a size around a control that already has all four. A\nround raised button is a shape, and where it sits is `data-dock` in the base\nlayer.\n\n**Nothing is reserved from the keyboard.** A menu that drills has to keep the\nhorizontal pair for going in and coming back; a dial goes nowhere, so all four\narrows move through the actions and a reader who guesses either axis is\nanswered. That is the controller's own house position, and this is a component\nwith no reason to argue with it.",
+          "description": "**A menu, and deliberately not a new one.** Its rows are commands, which is\nwhat `role=\"menu\"` and `menuitem` exist for, so this takes them and takes the\nbehavior with them: the arrow keys are `KeyboardNavigationController`, the\npanel is `PopoverController`, and open and closed is `MixinControllerOpen`.\n`Menu` is built from the same pieces. What is not shared is what actually\ndiffers: a dial has no submenus, no headers, no checkable rows and no long\nlist to type ahead through, and shipping it as an appearance of `Menu` would\nmean defining what a dial with a submenu in it does.\n\n**The trigger is a control the consumer slots**, and the floating action\nbutton is the usual one: `IconButton` with `shape=\"circle\"` and `raised`.\nThere is no `helix-fab`, because it would re-declare `emphasis`, `tone`,\n`label` and a size around a control that already has all four. A round raised\nbutton is a shape, and where it sits is `data-dock` in the base layer.\n\n**Nothing is reserved from the keyboard.** A menu that drills has to keep the\nhorizontal pair for going in and coming back; a dial goes nowhere, so all four\narrows move through the actions and a reader who guesses either axis is\nanswered. That is the controller's own house position, and this is a component\nwith no reason to argue with it.",
           "name": "HelixHeadlessSpeedDial",
           "cssParts": [
             {
@@ -80551,7 +80551,7 @@
                   "text": "TemplateResult"
                 }
               },
-              "description": "The control that opens it, behind a seam the styled tier can move.\n\n**This is the case the tier rule blocks and a render method unblocks.** A\nbare `<button>` is what the headless tier can write: it must not reach for\n`helix-button`, which lives a tier above and would make `headless` depend\non `styled`. So the base draws the plain control and a tier that has buttons\nof its own overrides this to use one. Written inline, the only way\nto change it would be to reimplement the template, panel and all."
+              "description": "The control that opens it, behind a seam the styled tier can move.\n\n**This is the case the tier rule blocks and a render method unblocks.** A\nbare `<button>` is what the headless tier can write: it must not reach for\n`Button`, which lives a tier above and would make `headless` depend on\n`styled`. So the base draws the plain control and a tier that has buttons\nof its own overrides this to use one. Written inline, the only way to\nchange it would be to reimplement the template, panel and all."
             },
             {
               "kind": "field",
@@ -82777,7 +82777,7 @@
               "type": {
                 "text": "'column-reverse' | 'row' | 'row-reverse' | undefined"
               },
-              "description": "Which way the label and the value are arranged.\n\nAbsent is a column, the label over the number, which is how a figure is read\non a dashboard: the words say what is coming and the number answers.\n`column-reverse` puts the number first, which is right where the number is\nthe headline and the label is a caption under it.\n\nThe row directions set them side by side, for a figure in a line of text or\na narrow strip where a stacked pair would be two lines of almost nothing.\n\nA property rather than a `helix-flex` around it, because the unit has to\nstay on the value's baseline whichever way the pair goes, and only this\nelement knows which of its children the unit belongs to.\n\n**Optional rather than defaulted to `column`, which is not a style\npreference.** A reflecting property with a default writes its attribute onto\nthe host the moment the element upgrades, and the host is markup the\nconsumer rendered: React compares what it rendered against what is there and\nreports a hydration mismatch. The absent attribute is the column, which the\nstylesheet reads as the base case, and the server spec asserts the host\ncarries nothing nobody wrote.",
+              "description": "Which way the label and the value are arranged.\n\nAbsent is a column, the label over the number, which is how a figure is read\non a dashboard: the words say what is coming and the number answers.\n`column-reverse` puts the number first, which is right where the number is\nthe headline and the label is a caption under it.\n\nThe row directions set them side by side, for a figure in a line of text or\na narrow strip where a stacked pair would be two lines of almost nothing.\n\nA property rather than a `Flex` around it, because the unit has to stay on\nthe value's baseline whichever way the pair goes, and only this element\nknows which of its children the unit belongs to.\n\n**Optional rather than defaulted to `column`, which is not a style\npreference.** A reflecting property with a default writes its attribute onto\nthe host the moment the element upgrades, and the host is markup the\nconsumer rendered: React compares what it rendered against what is there and\nreports a hydration mismatch. The absent attribute is the column, which the\nstylesheet reads as the base case, and the server spec asserts the host\ncarries nothing nobody wrote.",
               "attribute": "direction",
               "reflects": true
             },
@@ -83112,7 +83112,7 @@
               "type": {
                 "text": "'column-reverse' | 'row' | 'row-reverse' | undefined"
               },
-              "description": "Which way the label and the value are arranged.\n\nAbsent is a column, the label over the number, which is how a figure is read\non a dashboard: the words say what is coming and the number answers.\n`column-reverse` puts the number first, which is right where the number is\nthe headline and the label is a caption under it.\n\nThe row directions set them side by side, for a figure in a line of text or\na narrow strip where a stacked pair would be two lines of almost nothing.\n\nA property rather than a `helix-flex` around it, because the unit has to\nstay on the value's baseline whichever way the pair goes, and only this\nelement knows which of its children the unit belongs to.\n\n**Optional rather than defaulted to `column`, which is not a style\npreference.** A reflecting property with a default writes its attribute onto\nthe host the moment the element upgrades, and the host is markup the\nconsumer rendered: React compares what it rendered against what is there and\nreports a hydration mismatch. The absent attribute is the column, which the\nstylesheet reads as the base case, and the server spec asserts the host\ncarries nothing nobody wrote.",
+              "description": "Which way the label and the value are arranged.\n\nAbsent is a column, the label over the number, which is how a figure is read\non a dashboard: the words say what is coming and the number answers.\n`column-reverse` puts the number first, which is right where the number is\nthe headline and the label is a caption under it.\n\nThe row directions set them side by side, for a figure in a line of text or\na narrow strip where a stacked pair would be two lines of almost nothing.\n\nA property rather than a `Flex` around it, because the unit has to stay on\nthe value's baseline whichever way the pair goes, and only this element\nknows which of its children the unit belongs to.\n\n**Optional rather than defaulted to `column`, which is not a style\npreference.** A reflecting property with a default writes its attribute onto\nthe host the moment the element upgrades, and the host is markup the\nconsumer rendered: React compares what it rendered against what is there and\nreports a hydration mismatch. The absent attribute is the column, which the\nstylesheet reads as the base case, and the server spec asserts the host\ncarries nothing nobody wrote.",
               "fieldName": "direction"
             }
           ],
@@ -83211,7 +83211,7 @@
           "name": "HelixHeadlessSteps",
           "slots": [
             {
-              "description": "The steps, which are `helix-step` elements.",
+              "description": "The steps.",
               "name": ""
             }
           ],
@@ -83867,7 +83867,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The name is a button, and only the name.** The marker is `aria-hidden`\ndecoration, so wrapping it would add hit area and no meaning. It is\n`helix-button` rather than a `<button>` this element resets: the hover wash,\nthe focus ring and the disabled treatment are decisions the button carries.\n\n**It says nothing about where a reader may go.** Every step is pressable and\nannounces itself with `helix-select`; refusing the ones ahead is `disabled`,\nwhich the consumer sets, because only they know whether a form can be\nskipped.\n\n**The marker is a slot whose fallback is the state.** A finished stage shows\na check, the current one a target, one not yet reached an empty ring. Exactly\none is rendered, chosen from the step's own two booleans, rather than three\nwith two hidden in CSS: the accessibility tree would otherwise carry marks\nthat are not there.",
+          "description": "**The name is a button, and only the name.** The marker is `aria-hidden`\ndecoration, so wrapping it would add hit area and no meaning. It is `Button`\nrather than a `<button>` this element resets: the hover wash, the focus ring\nand the disabled treatment are decisions the button carries.\n\n**It says nothing about where a reader may go.** Every step is pressable and\nannounces itself with `helix-select`; refusing the ones ahead is `disabled`,\nwhich the consumer sets, because only they know whether a form can be\nskipped.\n\n**The marker is a slot whose fallback is the state.** A finished stage shows\na check, the current one a target, one not yet reached an empty ring. Exactly\none is rendered, chosen from the step's own two booleans, rather than three\nwith two hidden in CSS: the accessibility tree would otherwise carry marks\nthat are not there.",
           "name": "HelixHeadlessStep",
           "cssParts": [
             {
@@ -83883,7 +83883,7 @@
               "name": "description"
             },
             {
-              "description": "The step's name, which is a `helix-button`.",
+              "description": "The step's name, which is a control.",
               "name": "label"
             },
             {
@@ -83913,7 +83913,7 @@
               "name": ""
             },
             {
-              "description": "What the stage holds. Always rendered here, a tier with no stylesheet having nothing to fold it with; the styled tier opens it on the current step down a column, and a row has no space under a step to open it into.",
+              "description": "What the stage holds. Always rendered here, a tier with no stylesheet having nothing to fold it with, so whether it is folded away is the sheet's answer and not this tier's.",
               "name": "content"
             },
             {
@@ -86076,11 +86076,11 @@
               }
             }
           ],
-          "description": "The edge a column names, where it names one of the three.\n\n**`align` is a string anybody can write**, the union on the property being a\ncompile-time claim and nothing else, and both tables put the value straight\ninto CSS: `helix-table` into a rule in each row's shadow root, `helix-data-table`\ninto the cell it draws. One guard, here with the property, rather than one in\neach table and only one of them written."
+          "description": "The edge a column names, where it names one of the three.\n\n**`align` is a string anybody can write**, the union on the property being a\ncompile-time claim and nothing else, and both tables put the value straight\ninto CSS: `Table` into a rule in each row's shadow root, `DataTable` into the\ncell it draws. One guard, here with the property, rather than one in each\ntable and only one of them written."
         },
         {
           "kind": "class",
-          "description": "**The column is the source of both its heading and its track.** Add one and\nthe table has another track; remove one and it does not. That is what ties\nthe two together, and a table-level track list on its own does not: a list\nthat is accidentally the headings is invalid CSS a grid ignores in silence,\ndrawing the whole table as one column. `helix-table`'s `columns` is dealt out\na track at a time against the columns declared under it, so a heading among\nthe widths costs its own column a width and no more, and a list longer or\nshorter than the columns is trimmed or shared rather than being a second\ncount of them. A column's own `width` wins over the track in its place.\n\n**It is also where every per-column fact goes**, which is the reason\n`helix-data-table` needs no column config object. `field` names the key the\ncell reads, `sortable` says a reader may reorder by it, `align` says which\nedge the values line up against, and `cell` takes a function for anything the\nfield alone cannot express. The schema is markup and the data is a property:\na schema is small, declarative, and something a reader should be able to see\nin the HTML, and data is large and windowed.\n\n**The same element serves both tables**, the markup one and the data one. It\nis identified by a flag rather than by `instanceof`, so a `helix-table-column`\nworks inside either without the two roots sharing anything else.\n\n**It generates no box of its own**, the same as a row: the heading is a grid\nitem of the table, so it lines up with the cells under it rather than sitting\nin a wrapper that has to be kept the same width.\n\n**`hidden` takes the column out of the grid**, its heading and its track\nboth, so the columns after it keep their own widths and every row keeps its\ncells under their headings. `hidden=\"until-found\"` keeps its track, the\nheading being left drawn, as every Helix element leaves one with that value.\n`helix-data-table` draws the cells, so it leaves out the ones under a hidden\ncolumn. A `helix-table`'s cells are markup the consumer wrote, so hide them\nwith the column, as removing a column means removing its cells: a row with a\ncell more than the table has tracks wraps into the next.\n\n**A heading that does something holds a control that does it.** The host is\nthe column header, and a header is not focusable, so a sortable heading is a\n`<button>` inside it: the pattern for a sortable table, and the only way a\nreader on a keyboard reaches the sort at all. A column a reader may resize\ncarries a `separator` on its end edge, moved by dragging it or with the arrow\nkeys, and a column in a table that lets a reader reorder its columns carries\na grip the table moves.",
+          "description": "**The column is the source of both its heading and its track.** Add one and\nthe table has another track; remove one and it does not. That is what ties\nthe two together, and a table-level track list on its own does not: a list\nthat is accidentally the headings is invalid CSS a grid ignores in silence,\ndrawing the whole table as one column. `Table`'s `columns` is dealt out a\ntrack at a time against the columns declared under it, so a heading among the\nwidths costs its own column a width and no more, and a list longer or shorter\nthan the columns is trimmed or shared rather than being a second count of\nthem. A column's own `width` wins over the track in its place.\n\n**It is also where every per-column fact goes**, which is the reason\n`DataTable` needs no column config object. `field` names the key the cell\nreads, `sortable` says a reader may reorder by it, `align` says which edge\nthe values line up against, and `cell` takes a function for anything the\nfield alone cannot express. The schema is markup and the data is a property:\na schema is small, declarative, and something a reader should be able to see\nin the HTML, and data is large and windowed.\n\n**The same element serves both tables**, the markup one and the data one. It\nis identified by a flag rather than by `instanceof`, so a `TableColumn` works\ninside either without the two roots sharing anything else.\n\n**It generates no box of its own**, the same as a row: the heading is a grid\nitem of the table, so it lines up with the cells under it rather than sitting\nin a wrapper that has to be kept the same width.\n\n**`hidden` takes the column out of the grid**, its heading and its track\nboth, so the columns after it keep their own widths and every row keeps its\ncells under their headings. `hidden=\"until-found\"` keeps its track, the\nheading being left drawn, as every Helix element leaves one with that value.\n`DataTable` draws the cells, so it leaves out the ones under a hidden column.\nA `Table`'s cells are markup the consumer wrote, so hide them with the\ncolumn, as removing a column means removing its cells: a row with a cell more\nthan the table has tracks wraps into the next.\n\n**A heading that does something holds a control that does it.** The host is\nthe column header, and a header is not focusable, so a sortable heading is a\n`<button>` inside it: the pattern for a sortable table, and the only way a\nreader on a keyboard reaches the sort at all. A column a reader may resize\ncarries a `separator` on its end edge, moved by dragging it or with the arrow\nkeys, and a column in a table that lets a reader reorder its columns carries\na grip the table moves.",
           "name": "HelixHeadlessTableColumn",
           "cssParts": [
             {
@@ -86117,7 +86117,7 @@
               "type": {
                 "text": "'center' | 'end' | 'start' | undefined"
               },
-              "description": "Which edge the values line up against.\n\n`end` is what a column of numbers wants, so digits of the same magnitude sit\nabove each other. Left unset the cell inherits the table's own alignment,\nwhich is `start` for text in either writing direction. `helix-data-table`\naligns the cells it draws. A `helix-table`'s cells are markup the consumer\nwrote, so the table writes nothing onto them: it publishes the edge to each\ncell under this column as `--table-cell-align`, which `@fusion.dev/helix`\nreads. The heading is this element's own, and `@fusion.dev/helix` sets it to\nthe same edge in either table. `end` is the inline end, so in a\nright-to-left table it is the left.\n\n**Reflected, because the heading is drawn by a stylesheet and has no other\nway to learn it.** Set as a property, which is how React sets it, an\nunreflected `align` left no attribute to select on, so the heading stayed at\nthe start over a column of digits set to the end. It is the author's\ndeclaration and nothing moves it afterwards, so an attribute cannot come to\ndisagree with the element, which is the reason `open` records a custom\nstate instead. No default, so an unset column writes nothing onto its host.",
+              "description": "Which edge the values line up against.\n\n`end` is what a column of numbers wants, so digits of the same magnitude\nsit above each other. Left unset the cell inherits the table's own\nalignment, which is `start` for text in either writing direction.\n`DataTable` aligns the cells it draws. A `Table`'s cells are markup the\nconsumer wrote, so the table writes nothing onto them: it publishes the\nedge to each cell under this column as `--table-cell-align`, which\n`@fusion.dev/helix` reads. The heading is this element's own, and\n`@fusion.dev/helix` sets it to the same edge in either table. `end` is the\ninline end, so in a right-to-left table it is the left.\n\n**Reflected, because the heading is drawn by a stylesheet and has no other\nway to learn it.** Set as a property, which is how React sets it, an\nunreflected `align` left no attribute to select on, so the heading stayed at\nthe start over a column of digits set to the end. It is the author's\ndeclaration and nothing moves it afterwards, so an attribute cannot come to\ndisagree with the element, which is the reason `open` records a custom\nstate instead. No default, so an unset column writes nothing onto its host.",
               "attribute": "align",
               "reflects": true
             },
@@ -86177,7 +86177,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "The track this column takes, as CSS.\n\nA single track value, so anything valid in `grid-template-columns` works:\n`16rem`, `auto`, `minmax(0, 2fr)`. Left unset the column takes the track\nin its place in a `helix-table`'s `columns`, where that table sets one, and\notherwise shares what is left over equally, as `minmax(max-content, 1fr)`.\nSet here, it wins over that list, which is what lets a reader resize a\ncolumn in a table that set one.\n\n**The floor is `max-content`, never `0` and never `min-content`.** A zero\nfloor lets a track shrink under its own content: at 375 the columns squeeze\nto 75px and `eu-central-1` wraps across lines, which reads as corrupt data\nrather than as a table that ran out of room. A `min-content` floor still\nbreaks it, a hyphen being a wrap opportunity.\n\nThe overflow has somewhere to go: `@fusion.dev/helix` scrolls the table's\nown box, so a long value pushes past the container and the table scrolls\ninstead of the value being destroyed. Measured at 375: tracks of 192, 89 and\n102 against a 343 box, so the table scrolls 384 and the page does not move.\n\nThe cost is that a column of prose asks for its longest paragraph on one\nline, so a column holding sentences rather than values wants\n`width=\"minmax(0, 1fr)\"` to opt back into wrapping. A headless consumer owns\nthe overflow either way, that being paint, and gets the honest track list.",
+              "description": "The track this column takes, as CSS.\n\nA single track value, so anything valid in `grid-template-columns` works:\n`16rem`, `auto`, `minmax(0, 2fr)`. Left unset the column takes the track in\nits place in a `Table`'s `columns`, where that table sets one, and\notherwise shares what is left over equally, as `minmax(max-content, 1fr)`.\nSet here, it wins over that list, which is what lets a reader resize a\ncolumn in a table that set one.\n\n**The floor is `max-content`, never `0` and never `min-content`.** A zero\nfloor lets a track shrink under its own content: at 375 the columns squeeze\nto 75px and `eu-central-1` wraps across lines, which reads as corrupt data\nrather than as a table that ran out of room. A `min-content` floor still\nbreaks it, a hyphen being a wrap opportunity.\n\nThe overflow has somewhere to go: `@fusion.dev/helix` scrolls the table's\nown box, so a long value pushes past the container and the table scrolls\ninstead of the value being destroyed. Measured at 375: tracks of 192, 89 and\n102 against a 343 box, so the table scrolls 384 and the page does not move.\n\nThe cost is that a column of prose asks for its longest paragraph on one\nline, so a column holding sentences rather than values wants\n`width=\"minmax(0, 1fr)\"` to opt back into wrapping. A headless consumer owns\nthe overflow either way, that being paint, and gets the honest track list.",
               "attribute": "width"
             },
             {
@@ -86725,7 +86725,7 @@
               "type": {
                 "text": "'center' | 'end' | 'start' | undefined"
               },
-              "description": "Which edge the values line up against.\n\n`end` is what a column of numbers wants, so digits of the same magnitude sit\nabove each other. Left unset the cell inherits the table's own alignment,\nwhich is `start` for text in either writing direction. `helix-data-table`\naligns the cells it draws. A `helix-table`'s cells are markup the consumer\nwrote, so the table writes nothing onto them: it publishes the edge to each\ncell under this column as `--table-cell-align`, which `@fusion.dev/helix`\nreads. The heading is this element's own, and `@fusion.dev/helix` sets it to\nthe same edge in either table. `end` is the inline end, so in a\nright-to-left table it is the left.\n\n**Reflected, because the heading is drawn by a stylesheet and has no other\nway to learn it.** Set as a property, which is how React sets it, an\nunreflected `align` left no attribute to select on, so the heading stayed at\nthe start over a column of digits set to the end. It is the author's\ndeclaration and nothing moves it afterwards, so an attribute cannot come to\ndisagree with the element, which is the reason `open` records a custom\nstate instead. No default, so an unset column writes nothing onto its host.",
+              "description": "Which edge the values line up against.\n\n`end` is what a column of numbers wants, so digits of the same magnitude\nsit above each other. Left unset the cell inherits the table's own\nalignment, which is `start` for text in either writing direction.\n`DataTable` aligns the cells it draws. A `Table`'s cells are markup the\nconsumer wrote, so the table writes nothing onto them: it publishes the\nedge to each cell under this column as `--table-cell-align`, which\n`@fusion.dev/helix` reads. The heading is this element's own, and\n`@fusion.dev/helix` sets it to the same edge in either table. `end` is the\ninline end, so in a right-to-left table it is the left.\n\n**Reflected, because the heading is drawn by a stylesheet and has no other\nway to learn it.** Set as a property, which is how React sets it, an\nunreflected `align` left no attribute to select on, so the heading stayed at\nthe start over a column of digits set to the end. It is the author's\ndeclaration and nothing moves it afterwards, so an attribute cannot come to\ndisagree with the element, which is the reason `open` records a custom\nstate instead. No default, so an unset column writes nothing onto its host.",
               "fieldName": "align"
             },
             {
@@ -86761,7 +86761,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "The track this column takes, as CSS.\n\nA single track value, so anything valid in `grid-template-columns` works:\n`16rem`, `auto`, `minmax(0, 2fr)`. Left unset the column takes the track\nin its place in a `helix-table`'s `columns`, where that table sets one, and\notherwise shares what is left over equally, as `minmax(max-content, 1fr)`.\nSet here, it wins over that list, which is what lets a reader resize a\ncolumn in a table that set one.\n\n**The floor is `max-content`, never `0` and never `min-content`.** A zero\nfloor lets a track shrink under its own content: at 375 the columns squeeze\nto 75px and `eu-central-1` wraps across lines, which reads as corrupt data\nrather than as a table that ran out of room. A `min-content` floor still\nbreaks it, a hyphen being a wrap opportunity.\n\nThe overflow has somewhere to go: `@fusion.dev/helix` scrolls the table's\nown box, so a long value pushes past the container and the table scrolls\ninstead of the value being destroyed. Measured at 375: tracks of 192, 89 and\n102 against a 343 box, so the table scrolls 384 and the page does not move.\n\nThe cost is that a column of prose asks for its longest paragraph on one\nline, so a column holding sentences rather than values wants\n`width=\"minmax(0, 1fr)\"` to opt back into wrapping. A headless consumer owns\nthe overflow either way, that being paint, and gets the honest track list.",
+              "description": "The track this column takes, as CSS.\n\nA single track value, so anything valid in `grid-template-columns` works:\n`16rem`, `auto`, `minmax(0, 2fr)`. Left unset the column takes the track in\nits place in a `Table`'s `columns`, where that table sets one, and\notherwise shares what is left over equally, as `minmax(max-content, 1fr)`.\nSet here, it wins over that list, which is what lets a reader resize a\ncolumn in a table that set one.\n\n**The floor is `max-content`, never `0` and never `min-content`.** A zero\nfloor lets a track shrink under its own content: at 375 the columns squeeze\nto 75px and `eu-central-1` wraps across lines, which reads as corrupt data\nrather than as a table that ran out of room. A `min-content` floor still\nbreaks it, a hyphen being a wrap opportunity.\n\nThe overflow has somewhere to go: `@fusion.dev/helix` scrolls the table's\nown box, so a long value pushes past the container and the table scrolls\ninstead of the value being destroyed. Measured at 375: tracks of 192, 89 and\n102 against a 343 box, so the table scrolls 384 and the page does not move.\n\nThe cost is that a column of prose asks for its longest paragraph on one\nline, so a column holding sentences rather than values wants\n`width=\"minmax(0, 1fr)\"` to opt back into wrapping. A headless consumer owns\nthe overflow either way, that being paint, and gets the honest track list.",
               "fieldName": "width"
             }
           ],
@@ -86890,7 +86890,7 @@
           "name": "HelixHeadlessTable",
           "cssProperties": [
             {
-              "description": "The column track list this element worked out from the columns declared in it. Written by the component, read by the sheet that paints: a stylesheet takes it to lay the grid out, and setting it does nothing, because the component writes it inline on the very element that reads it and an inline declaration beats any sheet. `--table-columns` is the one to set.",
+              "description": "The column track list this element worked out from the columns declared in it. Written by the component, read by the sheet that paints: a stylesheet takes it to lay the grid out, and setting it does nothing, because the component writes it inline on the very element that reads it and an inline declaration beats any sheet. `columns` is the one to set.",
               "name": "--table-template"
             }
           ],
@@ -86926,7 +86926,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "The column widths as one track list, written once on the table.\n\nWith `helix-table-column` children, each track sizes the column in its\nplace, a hidden column counted, so hiding one takes its track with it and\nthe rest keep theirs. The columns decide how many tracks there are and this\nlist only how wide: a track past the last column is dropped, and a column\npast the last track shares what is left. With no columns declared it is\nthe whole track list. Tracks count as the grid counts them, so\n`minmax(0, 1fr)` is one and `repeat(2, 8rem)` is two, and one the grid\nwould refuse leaves its column at the default rather than taking the table\ndown to one column.\n\n**A column's own `width` wins over its track here.** The width is written\non the column it sizes, which is the more particular statement, and it is\nwhere a resize lands: a reader dragging a column's edge writes `width`, so\nwere this list to win, the drag would move the separator and leave the\ntrack where the list put it. This is the table's default for its columns,\nand each column may say otherwise.\n\n**It is also the only track list a server can send.** The shim has no tree\nto walk, so a render there knows of no columns and this attribute is the\nwhole answer: a table given one lays its grid out in the first paint, and a\ntable sized only by its columns is one track wide until it upgrades.\n\n**`helix-data-table` has none of this, and should not.** Its columns are\nits schema rather than an option, so there is always a column to write a\n`width` on and a list here would be a second way to say what that says.\nThis table's cells are markup, and a table written as rows of cells with no\n`helix-table-column` in it has nowhere else to be sized from. Both families\ntake the same CSS hook, `--table-columns` and `--data-table-columns`, which\nwins over whichever list the element worked out.",
+              "description": "The column widths as one track list, written once on the table.\n\nWith `TableColumn` children, each track sizes the column in its place, a\nhidden column counted, so hiding one takes its track with it and the rest\nkeep theirs. The columns decide how many tracks there are and this list\nonly how wide: a track past the last column is dropped, and a column past\nthe last track shares what is left. With no columns declared it is the\nwhole track list. Tracks count as the grid counts them, so `minmax(0, 1fr)`\nis one and `repeat(2, 8rem)` is two, and one the grid would refuse leaves\nits column at the default rather than taking the table down to one column.\n\n**A column's own `width` wins over its track here.** The width is written\non the column it sizes, which is the more particular statement, and it is\nwhere a resize lands: a reader dragging a column's edge writes `width`, so\nwere this list to win, the drag would move the separator and leave the\ntrack where the list put it. This is the table's default for its columns,\nand each column may say otherwise.\n\n**It is also the only track list a server can send.** The shim has no tree\nto walk, so a render there knows of no columns and this attribute is the\nwhole answer: a table given one lays its grid out in the first paint, and a\ntable sized only by its columns is one track wide until it upgrades.\n\n**Data Table has none of this, and should not.** Its columns are its schema\nrather than an option, so there is always a column to write a `width` on\nand a list here would be a second way to say what that says. This table's\ncells are markup, and a table written as rows of cells with no column\ndeclared in it has nowhere else to be sized from.",
               "attribute": "columns"
             },
             {
@@ -87501,7 +87501,7 @@
                 "text": "string"
               },
               "default": "''",
-              "description": "The column widths as one track list, written once on the table.\n\nWith `helix-table-column` children, each track sizes the column in its\nplace, a hidden column counted, so hiding one takes its track with it and\nthe rest keep theirs. The columns decide how many tracks there are and this\nlist only how wide: a track past the last column is dropped, and a column\npast the last track shares what is left. With no columns declared it is\nthe whole track list. Tracks count as the grid counts them, so\n`minmax(0, 1fr)` is one and `repeat(2, 8rem)` is two, and one the grid\nwould refuse leaves its column at the default rather than taking the table\ndown to one column.\n\n**A column's own `width` wins over its track here.** The width is written\non the column it sizes, which is the more particular statement, and it is\nwhere a resize lands: a reader dragging a column's edge writes `width`, so\nwere this list to win, the drag would move the separator and leave the\ntrack where the list put it. This is the table's default for its columns,\nand each column may say otherwise.\n\n**It is also the only track list a server can send.** The shim has no tree\nto walk, so a render there knows of no columns and this attribute is the\nwhole answer: a table given one lays its grid out in the first paint, and a\ntable sized only by its columns is one track wide until it upgrades.\n\n**`helix-data-table` has none of this, and should not.** Its columns are\nits schema rather than an option, so there is always a column to write a\n`width` on and a list here would be a second way to say what that says.\nThis table's cells are markup, and a table written as rows of cells with no\n`helix-table-column` in it has nowhere else to be sized from. Both families\ntake the same CSS hook, `--table-columns` and `--data-table-columns`, which\nwins over whichever list the element worked out.",
+              "description": "The column widths as one track list, written once on the table.\n\nWith `TableColumn` children, each track sizes the column in its place, a\nhidden column counted, so hiding one takes its track with it and the rest\nkeep theirs. The columns decide how many tracks there are and this list\nonly how wide: a track past the last column is dropped, and a column past\nthe last track shares what is left. With no columns declared it is the\nwhole track list. Tracks count as the grid counts them, so `minmax(0, 1fr)`\nis one and `repeat(2, 8rem)` is two, and one the grid would refuse leaves\nits column at the default rather than taking the table down to one column.\n\n**A column's own `width` wins over its track here.** The width is written\non the column it sizes, which is the more particular statement, and it is\nwhere a resize lands: a reader dragging a column's edge writes `width`, so\nwere this list to win, the drag would move the separator and leave the\ntrack where the list put it. This is the table's default for its columns,\nand each column may say otherwise.\n\n**It is also the only track list a server can send.** The shim has no tree\nto walk, so a render there knows of no columns and this attribute is the\nwhole answer: a table given one lays its grid out in the first paint, and a\ntable sized only by its columns is one track wide until it upgrades.\n\n**Data Table has none of this, and should not.** Its columns are its schema\nrather than an option, so there is always a column to write a `width` on\nand a list here would be a second way to say what that says. This table's\ncells are markup, and a table written as rows of cells with no column\ndeclared in it has nowhere else to be sized from.",
               "fieldName": "columns"
             },
             {
@@ -87625,7 +87625,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "`display: contents` in the styled tier, so the row's cells become items of\nthe table's own grid and every column lines up across every row. The row\nstill exists as an element, which is what a `<table>` gives away when its\ncells are laid out by the table algorithm.",
+          "description": "The row renders its cells straight through and lays out none of them, so a\nsheet can take them into the table's own grid and every column lines up\nacross every row. The row still exists as an element, which is what a\n`<table>` gives away when its cells are laid out by the table algorithm.",
           "name": "HelixHeadlessTableRow",
           "slots": [
             {
@@ -91807,7 +91807,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "**The list is `helix-time-picker`, not a second one written here.** It is worth\nhaving on its own, flat in a settings pane with nothing to open, and this\nelement is the field, the overlay and the form. Exactly the arrangement Input\nDate has with `helix-date-picker` and Color Input with `helix-color-picker`.\n\n**Typed as well as picked, and that is the division the list depends on.** A\nlist at a step offers the times a product actually has, and the field is how a\nreader says `09:07`. Neither is a worse version of the other: without the\nfield a step is a cage, and without the list every common time is typed out.\n\n**The value is `HH:mm` on a twenty-four hour clock, always**, because that is\nwhat a form submits, what `min` and `max` compare against, and what sorts as a\nstring. What the field *shows* is the reader's own convention while it is\nidle, and the value itself once the caret is in it: the same separation Input\nDate draws, and for the same reason, which is that `2:30 PM` cannot be typed\nback and `Intl` has no parser to make it so.",
+          "description": "**The list is `TimePicker`, not a second one written here.** It is worth\nhaving on its own, flat in a settings pane with nothing to open, and this\nelement is the field, the overlay and the form. Exactly the arrangement Input\nDate has with `DatePicker` and Color Input with `ColorPicker`.\n\n**Typed as well as picked, and that is the division the list depends on.** A\nlist at a step offers the times a product actually has, and the field is how a\nreader says `09:07`. Neither is a worse version of the other: without the\nfield a step is a cage, and without the list every common time is typed out.\n\n**The value is `HH:mm` on a twenty-four hour clock, always**, because that is\nwhat a form submits, what `min` and `max` compare against, and what sorts as a\nstring. What the field *shows* is the reader's own convention while it is\nidle, and the value itself once the caret is in it: the same separation Input\nDate draws, and for the same reason, which is that `2:30 PM` cannot be typed\nback and `Intl` has no parser to make it so.",
           "name": "HelixHeadlessTimeInput",
           "cssParts": [
             {
@@ -92063,7 +92063,7 @@
               "type": {
                 "text": "'multiple' | 'range' | undefined"
               },
-              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end, `multiple` holds any\nnumber, and both are carried as one comma-separated value, which is what a\nform submits. `helix-time-picker` carries the longer note on the interaction and on\nwhy a wheel cannot do either.",
+              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end, `multiple` holds any\nnumber, and both are carried as one comma-separated value, which is what a\nform submits. `TimePicker` carries the longer note on the interaction and\non why a wheel cannot do either.",
               "attribute": "mode",
               "reflects": true
             },
@@ -92237,7 +92237,7 @@
               "type": {
                 "text": "'list' | 'wheels' | undefined"
               },
-              "description": "Whether the panel is one list of times or a column each.\n\n`list` offers whole times, which is what a product with slots or opening\nhours has. `wheels` splits them into an hour, a minute and a meridiem where\nthe locale uses one, which is the only way to reach `23:45` without\nscrolling most of a day. `helix-time-picker` carries the longer note on the trade.\n\n**Optional rather than defaulted, which is not a style preference.** A\nreflecting property with a default writes its attribute onto every host the\nmoment the element upgrades, including the ones a consumer rendered and did\nnot touch, and React reads that as a hydration mismatch. The absent\nattribute *is* the list, which the stylesheet reads as the base rule, so\nonly the departure has to be named.",
+              "description": "Whether the panel is one list of times or a column each.\n\n`list` offers whole times, which is what a product with slots or opening\nhours has. `wheels` splits them into an hour, a minute and a meridiem where\nthe locale uses one, which is the only way to reach `23:45` without\nscrolling most of a day. `TimePicker` carries the longer note on the trade.\n\n**Optional rather than defaulted, which is not a style preference.** A\nreflecting property with a default writes its attribute onto every host the\nmoment the element upgrades, including the ones a consumer rendered and did\nnot touch, and React reads that as a hydration mismatch. The absent\nattribute *is* the list, which the stylesheet reads as the base rule, so\nonly the departure has to be named.",
               "attribute": "variant",
               "reflects": true
             },
@@ -92280,7 +92280,7 @@
                   "text": "TemplateResult"
                 }
               },
-              "description": "The mark on the button that opens the clock, behind a seam.\n\nSlot fallback, so a consumer replaces it by slotting; and a method, so the\ntier above replaces it by overriding.\n\n**A named import rather than `<helix-icon>`**, on three counts.\n`helix-icon` indexes the whole icon map at runtime, `icons[this.name]`, and\nno bundler can shake a runtime lookup, so every consumer of any component\nthat renders one ships all 50kB of path data; a named import ships one\npath. The import is checked when the package compiles, where a `name`\nstring is checked only by a later name lookup. And `helix-icon` is not\nregistered on the headless tier at all, which registers `helix-headless-icon`,\nso a headless consumer gets an inert unknown element and no mark whatsoever."
+              "description": "The mark on the button that opens the clock, behind a seam.\n\nSlot fallback, so a consumer replaces it by slotting; and a method, so the\ntier above replaces it by overriding.\n\n**A named import rather than `<helix-headless-icon>`**, on three counts.\n`Icon` indexes the whole icon map at runtime, `icons[this.name]`, and no\nbundler can shake a runtime lookup, so every consumer of any component that\nrenders one ships all 50kB of path data; a named import ships one path. The\nimport is checked when the package compiles, where a `name` string is\nchecked only by a later name lookup. And `Icon` is not registered on the\nheadless tier at all, which registers `helix-headless-icon`, so a headless\nconsumer gets an inert unknown element and no mark whatsoever."
             },
             {
               "kind": "field",
@@ -93772,7 +93772,7 @@
               "type": {
                 "text": "'multiple' | 'range' | undefined"
               },
-              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end, `multiple` holds any\nnumber, and both are carried as one comma-separated value, which is what a\nform submits. `helix-time-picker` carries the longer note on the interaction and on\nwhy a wheel cannot do either.",
+              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end, `multiple` holds any\nnumber, and both are carried as one comma-separated value, which is what a\nform submits. `TimePicker` carries the longer note on the interaction and\non why a wheel cannot do either.",
               "fieldName": "mode"
             },
             {
@@ -93816,7 +93816,7 @@
               "type": {
                 "text": "'list' | 'wheels' | undefined"
               },
-              "description": "Whether the panel is one list of times or a column each.\n\n`list` offers whole times, which is what a product with slots or opening\nhours has. `wheels` splits them into an hour, a minute and a meridiem where\nthe locale uses one, which is the only way to reach `23:45` without\nscrolling most of a day. `helix-time-picker` carries the longer note on the trade.\n\n**Optional rather than defaulted, which is not a style preference.** A\nreflecting property with a default writes its attribute onto every host the\nmoment the element upgrades, including the ones a consumer rendered and did\nnot touch, and React reads that as a hydration mismatch. The absent\nattribute *is* the list, which the stylesheet reads as the base rule, so\nonly the departure has to be named.",
+              "description": "Whether the panel is one list of times or a column each.\n\n`list` offers whole times, which is what a product with slots or opening\nhours has. `wheels` splits them into an hour, a minute and a meridiem where\nthe locale uses one, which is the only way to reach `23:45` without\nscrolling most of a day. `TimePicker` carries the longer note on the trade.\n\n**Optional rather than defaulted, which is not a style preference.** A\nreflecting property with a default writes its attribute onto every host the\nmoment the element upgrades, including the ones a consumer rendered and did\nnot touch, and React reads that as a hydration mismatch. The absent\nattribute *is* the list, which the stylesheet reads as the base rule, so\nonly the departure has to be named.",
               "fieldName": "variant"
             },
             {
@@ -94207,7 +94207,7 @@
         },
         {
           "kind": "class",
-          "description": "**A list rather than three wheels**, and the reason is what a list already\nis. `helix-listbox` brings the keyboard, the typeahead, the scroll into view\nand the announcement of a set of options, and a time at a step is a set of\noptions. Three little wheels for hours, minutes and a meridiem are three\ncontrols to build, three to test, and a tab stop each for something a reader\nanswers once.\n\n**A step is what makes it a list at all.** Fifteen minutes is ninety-six rows,\nwhich a reader scrolls or types into; one minute is one thousand four hundred\nand forty, which is a listbox in name only. So the list is for the times a\nproduct actually offers, and arbitrary times are the field's job in Input\nTime, the same division Date Input draws between typing a date and using the\ncalendar.\n\n**The value is `HH:mm` on a twenty-four hour clock, always.** That is what a\nform submits and what `min` and `max` compare against, and it sorts as a\nstring, which is the whole reason the shape is that one. What each row *says*\nis the reader's own convention, from `Intl`.\n\n**The chosen row is marked in the markup, not worked out on upgrade.** The\nlistbox is handed the answer as `.value`, and lit-ssr serializes no property\nbinding, so a served `value=\"09:30\"` shipped ninety-six rows with nothing on\nany of them saying which one it was: the mark arrived when the listbox\nupgraded and ran `place`. So each option is given `selected` as well, which\nis an attribute binding and does serve, and which the listbox then agrees\nwith rather than fights, `place` writing the same answer from the same value.\n`selected` is the declaration a consumer writes on `Option` anywhere else in\nthe library, and Date Picker expresses its own choice as markup for the same\nreason: a picker whose answer is invisible until a script runs is a picker\nwhose answer is not served at all. Both readings do it, the list and the\nwheels, a wheel being a listbox like the list is.",
+          "description": "**A list rather than three wheels**, and the reason is what a list already\nis. `Listbox` brings the keyboard, the typeahead, the scroll into view and\nthe announcement of a set of options, and a time at a step is a set of\noptions. Three little wheels for hours, minutes and a meridiem are three\ncontrols to build, three to test, and a tab stop each for something a reader\nanswers once.\n\n**A step is what makes it a list at all.** Fifteen minutes is ninety-six rows,\nwhich a reader scrolls or types into; one minute is one thousand four hundred\nand forty, which is a listbox in name only. So the list is for the times a\nproduct actually offers, and arbitrary times are the field's job in Input\nTime, the same division Date Input draws between typing a date and using the\ncalendar.\n\n**The value is `HH:mm` on a twenty-four hour clock, always.** That is what a\nform submits and what `min` and `max` compare against, and it sorts as a\nstring, which is the whole reason the shape is that one. What each row *says*\nis the reader's own convention, from `Intl`.\n\n**The chosen row is marked in the markup, not worked out on upgrade.** The\nlistbox is handed the answer as `.value`, and lit-ssr serializes no property\nbinding, so a served `value=\"09:30\"` shipped ninety-six rows with nothing on\nany of them saying which one it was: the mark arrived when the listbox\nupgraded and ran `place`. So each option is given `selected` as well, which\nis an attribute binding and does serve, and which the listbox then agrees\nwith rather than fights, `place` writing the same answer from the same value.\n`selected` is the declaration a consumer writes on `Option` anywhere else in\nthe library, and Date Picker expresses its own choice as markup for the same\nreason: a picker whose answer is invisible until a script runs is a picker\nwhose answer is not served at all. Both readings do it, the list and the\nwheels, a wheel being a listbox like the list is.",
           "name": "HelixHeadlessTimePicker",
           "cssParts": [
             {
@@ -94403,7 +94403,7 @@
               "type": {
                 "text": "'multiple' | 'range' | undefined"
               },
-              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end with everything between\nthem shown as part of the answer, `multiple` holds any number, and both are\ncarried as one comma-separated value: the same shape `helix-date-picker` uses,\nbecause it is the same question asked of a different unit.\n\n**A range is two picks, and a third starts over.** That is the calendar's\nbehavior and a reader who has met one has met both. The alternative, a\nseparate \"from\" and \"to\" control, is two answers to what a product asks as\none, and it makes the between part impossible to show.\n\n**Neither works on wheels and the variant falls back.** A range on\nwheels is two sets of wheels, which is two controls; a multiple is a set,\nand a wheel has exactly one value by construction. Rather than draw\nsomething that cannot say what it holds, `wheels` renders as a list\nwhenever a mode is set.",
+              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end with everything between\nthem shown as part of the answer, `multiple` holds any number, and both are\ncarried as one comma-separated value: the same shape `DatePicker` uses,\nbecause it is the same question asked of a different unit.\n\n**A range is two picks, and a third starts over.** That is the calendar's\nbehavior and a reader who has met one has met both. The alternative, a\nseparate \"from\" and \"to\" control, is two answers to what a product asks as\none, and it makes the between part impossible to show.\n\n**Neither works on wheels and the variant falls back.** A range on\nwheels is two sets of wheels, which is two controls; a multiple is a set,\nand a wheel has exactly one value by construction. Rather than draw\nsomething that cannot say what it holds, `wheels` renders as a list\nwhenever a mode is set.",
               "attribute": "mode",
               "reflects": true
             },
@@ -94672,7 +94672,7 @@
                   "text": "Promise<void>"
                 }
               },
-              "description": "Reveals once the wheels, or the list and its rows, have rendered themselves.\n\n`updated` is this element's own moment, and each wheel is a `helix-listbox`\nthat was created during this very render: its first update has been\nscheduled and has not run, so it has no rows and no height and `reveal`\nmeasures zero and gives up. Measured on a standalone picker, every column\nsat at `scrollTop: 0` with the answer nine rows down, which reads as a wheel\nshowing the wrong time rather than as one that failed to scroll.\n\nWhatever opens a panel calls `reveal` again afterwards, for the separate\nreason that method documents. Flat on a page there is no panel and no\nsecond call.\n\nSafe to run on every update: the listbox stops as soon as the chosen row is\nwithin a pixel of center, so a call with nothing to do costs two rect reads\nand changes nothing. It cannot fight a reader mid-scroll either, since\nscrolling a wheel changes no property and so schedules no update."
+              "description": "Reveals once the wheels, or the list and its rows, have rendered themselves.\n\n`updated` is this element's own moment, and each wheel is a `Listbox` that\nwas created during this very render: its first update has been scheduled\nand has not run, so it has no rows and no height and `reveal` measures zero\nand gives up. Measured on a standalone picker, every column sat at\n`scrollTop: 0` with the answer nine rows down, which reads as a wheel\nshowing the wrong time rather than as one that failed to scroll.\n\nWhatever opens a panel calls `reveal` again afterwards, for the separate\nreason that method documents. Flat on a page there is no panel and no\nsecond call.\n\nSafe to run on every update: the listbox stops as soon as the chosen row is\nwithin a pixel of center, so a call with nothing to do costs two rect reads\nand changes nothing. It cannot fight a reader mid-scroll either, since\nscrolling a wheel changes no property and so schedules no update."
             },
             {
               "kind": "field",
@@ -95473,7 +95473,7 @@
               "type": {
                 "text": "'multiple' | 'range' | undefined"
               },
-              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end with everything between\nthem shown as part of the answer, `multiple` holds any number, and both are\ncarried as one comma-separated value: the same shape `helix-date-picker` uses,\nbecause it is the same question asked of a different unit.\n\n**A range is two picks, and a third starts over.** That is the calendar's\nbehavior and a reader who has met one has met both. The alternative, a\nseparate \"from\" and \"to\" control, is two answers to what a product asks as\none, and it makes the between part impossible to show.\n\n**Neither works on wheels and the variant falls back.** A range on\nwheels is two sets of wheels, which is two controls; a multiple is a set,\nand a wheel has exactly one value by construction. Rather than draw\nsomething that cannot say what it holds, `wheels` renders as a list\nwhenever a mode is set.",
+              "description": "How many times can be held at once.\n\nAbsent is one. `range` holds a start and an end with everything between\nthem shown as part of the answer, `multiple` holds any number, and both are\ncarried as one comma-separated value: the same shape `DatePicker` uses,\nbecause it is the same question asked of a different unit.\n\n**A range is two picks, and a third starts over.** That is the calendar's\nbehavior and a reader who has met one has met both. The alternative, a\nseparate \"from\" and \"to\" control, is two answers to what a product asks as\none, and it makes the between part impossible to show.\n\n**Neither works on wheels and the variant falls back.** A range on\nwheels is two sets of wheels, which is two controls; a multiple is a set,\nand a wheel has exactly one value by construction. Rather than draw\nsomething that cannot say what it holds, `wheels` renders as a list\nwhenever a mode is set.",
               "fieldName": "mode"
             },
             {
@@ -97337,7 +97337,7 @@
           ],
           "slots": [
             {
-              "description": "The messages, as `helix-toast-item` elements.",
+              "description": "The messages.",
               "name": ""
             }
           ],
@@ -97358,7 +97358,7 @@
               "type": {
                 "text": "| 'bottom-center'\n    | 'bottom-end'\n    | 'bottom-start'\n    | 'top-center'\n    | 'top-end'\n    | 'top-start' | undefined"
               },
-              "description": "Which corner or edge the stack sits at.\n\nNamed here rather than on each message, because where a page puts its\ntoasts is a fact about the page: a stack whose members disagreed about\nwhere they were would not be a stack.\n\n**Optional, and bottom-end is the absence of it.** Reflecting *and*\ndefaulting is the pair that writes onto a host: every toast region anybody\nrendered would gain a `placement=\"bottom-end\"` nobody typed, which a React\nconsumer's server markup does not have. The styled sheet carries the resting\narrangement in its own `:host` rule and the selectors describe only the\ndepartures, which is what `orientation` does.",
+              "description": "Which corner or edge the stack sits at.\n\nNamed here rather than on each message, because where a page puts its\ntoasts is a fact about the page: a stack whose members disagreed about\nwhere they were would not be a stack.\n\n**Optional, and bottom-end is the absence of it.** Reflecting *and*\ndefaulting is the pair that writes onto a host: every toast region anybody\nrendered would gain a `placement=\"bottom-end\"` nobody typed, which a React\nconsumer's server markup does not have. A sheet carries the resting\narrangement in its own `:host` rule and the selectors describe only the\ndepartures, which is what `orientation` does.",
               "attribute": "placement",
               "reflects": true
             },
@@ -97712,7 +97712,7 @@
               "type": {
                 "text": "| 'bottom-center'\n    | 'bottom-end'\n    | 'bottom-start'\n    | 'top-center'\n    | 'top-end'\n    | 'top-start' | undefined"
               },
-              "description": "Which corner or edge the stack sits at.\n\nNamed here rather than on each message, because where a page puts its\ntoasts is a fact about the page: a stack whose members disagreed about\nwhere they were would not be a stack.\n\n**Optional, and bottom-end is the absence of it.** Reflecting *and*\ndefaulting is the pair that writes onto a host: every toast region anybody\nrendered would gain a `placement=\"bottom-end\"` nobody typed, which a React\nconsumer's server markup does not have. The styled sheet carries the resting\narrangement in its own `:host` rule and the selectors describe only the\ndepartures, which is what `orientation` does.",
+              "description": "Which corner or edge the stack sits at.\n\nNamed here rather than on each message, because where a page puts its\ntoasts is a fact about the page: a stack whose members disagreed about\nwhere they were would not be a stack.\n\n**Optional, and bottom-end is the absence of it.** Reflecting *and*\ndefaulting is the pair that writes onto a host: every toast region anybody\nrendered would gain a `placement=\"bottom-end\"` nobody typed, which a React\nconsumer's server markup does not have. A sheet carries the resting\narrangement in its own `:host` rule and the selectors describe only the\ndepartures, which is what `orientation` does.",
               "fieldName": "placement"
             }
           ],
@@ -99030,7 +99030,7 @@
               "type": {
                 "text": "boolean"
               },
-              "description": "Whether the markup declared this one chosen, which is what a reset returns to.\n\n`<option selected>` reflects to `defaultSelected` and `option.selected` is\nlive state that does not reflect, and those are two different facts. This\nelement has one `selected` doing both jobs, so the declaration is the\nattribute and the state is the property.\n\n**The attribute, read when it is asked for, and never a snapshot taken in\nthe constructor.** A field initializer runs there, and an element the HTML\nparser created has none of its attributes yet at that moment: measured on\nthis element, on `Option` and on `Radio`, `<helix-toggle value=\"light\"\nselected>` reported `defaultSelected` as `false` both from parsed markup\nand from `createElement` followed by `setAttribute`, which is every way a\nconsumer writes one. The snapshot could only ever be true for an element\nalready in the document when `customElements.define` ran.\n\nNothing overwrites it either, which is what makes reading it late correct\nrather than merely later: `selected` does not reflect, for the reason\n`willUpdate` records, so the group's writes never reach the attribute and\nwhat it says is what the markup said.\n\n**The deeper half is not this.** State reflecting under the declaration's\nname is the fault `TreeItem` avoids by mirroring to `data-selected`.\nSplitting them means moving what every stylesheet selects on, which is a\nchange of its own rather than a property being added."
+              "description": "Whether the markup declared this one chosen, which is what a reset returns to.\n\n`<option selected>` reflects to `defaultSelected` and `option.selected` is\nlive state that does not reflect, and those are two different facts. This\nelement has one `selected` doing both jobs, so the declaration is the\nattribute and the state is the property.\n\n**The attribute, read when it is asked for, and never a snapshot taken in\nthe constructor.** A field initializer runs there, and an element the HTML\nparser created has none of its attributes yet at that moment: measured on\nthis element, on `Option` and on `Radio`, `<helix-headless-toggle\nvalue=\"light\" selected>` reported `defaultSelected` as `false` both from\nparsed markup and from `createElement` followed by `setAttribute`, which is\nevery way a consumer writes one. The snapshot could only ever be true for\nan element already in the document when `customElements.define` ran.\n\nNothing overwrites it either, which is what makes reading it late correct\nrather than merely later: `selected` does not reflect, for the reason\n`willUpdate` records, so the group's writes never reach the attribute and\nwhat it says is what the markup said.\n\n**The deeper half is not this.** State reflecting under the declaration's\nname is the fault `TreeItem` avoids by mirroring to `data-selected`.\nSplitting them means moving what every stylesheet selects on, which is a\nchange of its own rather than a property being added."
             },
             {
               "kind": "field",
@@ -100141,7 +100141,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "The trigger is slotted and stays the consumer's own element. The bubble is\nrendered here, in the top layer through the popover API, so it clears any\nancestor with `overflow: hidden` without a `z-index` anywhere. Position comes\nfrom `FloatingController`, which flips to another side when the preferred one\ndoes not fit.\n\n**Two forms, and they are not equivalent for a screen reader.** Wrapping is\nthe convenience form and is right where the bubble decorates a control that\nalready says what it is: nothing here can describe a slotted trigger, an\nIDREF being unable to cross a shadow boundary.\n\n`for` is the form that works. The tooltip sits *beside* the trigger in the\nconsumer's own tree and names it by id, so an IDREF between them is legal in\nthe ordinary way:\n\n```html\n<helix-button id=\"save\" aria-describedby=\"save-tip\">Save</helix-button>\n<helix-tooltip id=\"save-tip\" for=\"save\" content=\"Save your work\"></helix-tooltip>\n```\n\nThe reference points at *this host*, not at the bubble in its shadow root:\nthe description is computed from the flat tree, which descends into shadow\ncontent on its own.\n\n**The consumer writes that `aria-describedby`, and nothing here writes it for\nthem.** Setting it at runtime is a hydration mismatch React will not patch\nup. `for` audits itself on first open and warns once where it is missing.\n\nIt shows on hover and on focus, and hides on leave, on blur, on Escape, and\nonce the trigger has been activated: the hint is spent when the control has\nbeen used.",
+          "description": "The trigger is slotted and stays the consumer's own element. The bubble is\nrendered here, in the top layer through the popover API, so it clears any\nancestor with `overflow: hidden` without a `z-index` anywhere. Position comes\nfrom `FloatingController`, which flips to another side when the preferred one\ndoes not fit.\n\n**Two forms, and they are not equivalent for a screen reader.** Wrapping is\nthe convenience form and is right where the bubble decorates a control that\nalready says what it is: nothing here can describe a slotted trigger, an\nIDREF being unable to cross a shadow boundary.\n\n`for` is the form that works. The tooltip sits *beside* the trigger in the\nconsumer's own tree and names it by id, so an IDREF between them is legal in\nthe ordinary way:\n\n```html\n<helix-headless-button id=\"save\" aria-describedby=\"save-tip\">Save</helix-headless-button>\n<helix-headless-tooltip id=\"save-tip\" for=\"save\" content=\"Save your work\"></helix-headless-tooltip>\n```\n\nThe reference points at *this host*, not at the bubble in its shadow root:\nthe description is computed from the flat tree, which descends into shadow\ncontent on its own.\n\n**The consumer writes that `aria-describedby`, and nothing here writes it for\nthem.** Setting it at runtime is a hydration mismatch React will not patch\nup. `for` audits itself on first open and warns once where it is missing.\n\nIt shows on hover and on focus, and hides on leave, on blur, on Escape, and\nonce the trigger has been activated: the hint is spent when the control has\nbeen used.",
           "name": "HelixHeadlessTooltip",
           "cssParts": [
             {
@@ -102017,7 +102017,7 @@
       "declarations": [
         {
           "kind": "class",
-          "description": "A stop names a target with `for` and holds whatever the consumer puts in it.\nIt owns no content of its own, which is the whole shape of this family: the\ntour abstracts the three things that are genuinely hard, and the words are\nthe consumer's.\n\nWhat it does own:\n\n- **Where the panel goes.** Anchored to the target through the same floating\n  pass every other overlay here uses.\n- **That it is on top.** A popover, so it is in the top layer. A tour points\n  at things inside dialogs, inside scrollers, inside anything with a stacking\n  context, and a panel in the ordinary flow loses to all three.\n- **When it shows.** Which is `current`, and which the root decides.\n\nWhat it does not own is the heading, the paragraph, the picture and the\nlayout. Slot a `helix-card`, or a heading and a line of prose, or anything\nelse. Fixed `title` and `body` parts would be a component deciding what a\ntour is allowed to say.\n\n**The buttons are slot fallback, so the common case costs nothing and the\nuncommon one is possible.** Write nothing and get Back, Next and Skip already\nwired. Slot `actions` and they are yours, driven by `next()`, `back()` and\n`skip()` on the root, which are public for exactly that.\n\n**`for` is an id, the same as `Tooltip`'s.** The target is elsewhere on the\npage, and an id is the one reference a consumer can write in markup without\nholding an element. Resolved against the root node, so a stop inside a shadow\ntree points at a target in that tree.",
+          "description": "A stop names a target with `for` and holds whatever the consumer puts in it.\nIt owns no content of its own, which is the whole shape of this family: the\ntour abstracts the three things that are genuinely hard, and the words are\nthe consumer's.\n\nWhat it does own:\n\n- **Where the panel goes.** Anchored to the target through the same floating\n  pass every other overlay here uses.\n- **That it is on top.** A popover, so it is in the top layer. A tour points\n  at things inside dialogs, inside scrollers, inside anything with a stacking\n  context, and a panel in the ordinary flow loses to all three.\n- **When it shows.** Which is `current`, and which the root decides.\n\nWhat it does not own is the heading, the paragraph, the picture and the\nlayout. Slot a `Card`, or a heading and a line of prose, or anything else.\nFixed `title` and `body` parts would be a component deciding what a tour is\nallowed to say.\n\n**The buttons are slot fallback, so the common case costs nothing and the\nuncommon one is possible.** Write nothing and get Back, Next and Skip already\nwired. Slot `actions` and they are yours, driven by `next()`, `back()` and\n`skip()` on the root, which are public for exactly that.\n\n**`for` is an id, the same as `Tooltip`'s.** The target is elsewhere on the\npage, and an id is the one reference a consumer can write in markup without\nholding an element. Resolved against the root node, so a stop inside a shadow\ntree points at a target in that tree.",
           "name": "HelixHeadlessTourStop",
           "cssParts": [
             {
@@ -103615,7 +103615,7 @@
                 "text": "HelixHeadlessTreeItem[]"
               },
               "privacy": "private",
-              "description": "Every item, in the order a reader moving down would meet them.\n\nSelected by tag, never by `'*'` and a filter: this is the array a single\narrow press rebuilds, and `'*'` makes it every descendant element in the\ntree, labels and icons and whatever a consumer slotted into a row included.\n\nThe two tag names rather than `instanceof` on everything: a styled\n`helix-tree-item` extends the base, so both names are needed, and the\n`instanceof` stays as the type guard because a consumer can write either\ntag and neither selector proves what the element is.",
+              "description": "Every item, in the order a reader moving down would meet them.\n\n**By identity rather than by tag**, which is the same reason Splitter\nwrites out at `findOuter`: a tag list names one tier's element from the\nother's package, and misses any third tag, which is exactly what\nsubclassing an item produces. The item then never enters the array a single\narrow press rebuilds, so it is silently unreachable from the keyboard.\n\n`'*'` and a filter rather than a tag: the walk is one native pass and the\n`instanceof` is what narrows it, so the labels, icons and whatever a\nconsumer slotted into a row are gone by the time the array exists.",
               "readonly": true
             },
             {
@@ -105636,7 +105636,7 @@
                 "text": "HelixHeadlessVideoPlaylistItem[]"
               },
               "privacy": "private",
-              "description": "The playlist's entries, in the order it holds them.",
+              "description": "The playlist's entries, in the order it holds them.\n\nBy identity rather than by tag, the same reason the playlist itself reads\nthem that way: a tag list names one tier's element from the other's package\nand misses any third one.",
               "readonly": true
             },
             {
@@ -106106,7 +106106,7 @@
               "name": ""
             },
             {
-              "description": "The `helix-video` they play in.",
+              "description": "The player the entries take turns in.",
               "name": "player"
             }
           ],
@@ -106148,7 +106148,7 @@
                 "text": "| (HTMLElement & {\n        label: string\n        play: () => void\n        poster: string\n        src: string\n        updateComplete?: Promise<unknown>\n      })\n    | null"
               },
               "privacy": "private",
-              "description": "The player, whichever tag the consumer slotted.\n\nAsked of the slot rather than matched by tag name: the headless tier's\nplayer is `helix-headless-video` and the styled tier's is `helix-video`, and\na consumer bringing a third under either name should still be driven. What\nis needed of it is `src`, which is the whole of the contract.",
+              "description": "The player, whichever tag the consumer slotted.\n\nAsked of the slot rather than matched by tag name: each tier registers its\nown player under its own name, and a consumer bringing a third should still\nbe driven. What is needed of it is `src`, which is the whole of the\ncontract.",
               "readonly": true
             },
             {