* feat: Give the Button Theme Roles for Its Label, Height, Fill and States
The Button drew a fixed 500-weight label at 40px and 36px, painted its
primary fill in the inverted surface its checkbox and switch share, focused
in the primary ink, and only dimmed when disabled, so a theme could not
draw Click UI's 400-weight 32px buttons or its #302e32 primary.
Add controlFontWeight, buttonHeight and buttonHeightSm appearance roles and
button-primary and button-primary-hover color roles, which follow the
inverted surface for a theme that names only that. The Button reads them,
focuses in focus-control, presses in the pressed roles and composes the
shared disabled recipe; the defaults reproduce today's button exactly.
Tailwind Merge now knows the new utilities, so a caller's own height or
weight still wins.
SelectTrigger hid its outline on every focus and drew nothing instead; it
now draws keyboard focus in focus-control, like the shared field.
The ClickHouse theme takes Click UI's label weight, button height and
primary fills, and the drift guard pins them and measures the Button.
* fix: Focus the Shared Inputs in the Focus-Control Role and Repair the Button Specs
SecretInput, MultiSearch, InputWithDropDown and InputCombobox drew keyboard
focus in the decorative ring-primary, about 2.3:1 on the dark default canvas.
They now use focus-control, and a guard holds that role at 3:1 on every
control canvas in each theme. The Scrim spec follows the Button's focus role,
and the button-roles scenario opens the Data & Privacy tab by its current name.
* test: Reach the Shared Input Focus Rings by Keyboard
The focus-role guard read component source, which a dead string or a ring on a
non-focusable wrapper would satisfy. Each control is now reached by Tab and
the focused element, or the combobox's ring host while it holds keyboard
focus, is asserted to paint the focus-control ring.
* feat: Give Theme-Sized Controls Their Own Padding and Gap Roles
The theme variants of Button and Chip, CheckboxButton and every composer
control read spaceNormal and spaceCompact, which also pad message rows,
steer bubbles and the composer send button, so a theme could not give its
controls Click UI's button spacing without reshaping the chat.
Add controlPaddingX and controlGap appearance roles, which those controls
read instead. Their defaults are the shared spacing values, a theme that
names only spaceNormal or spaceCompact keeps its controls on them, and the
preset falls back to the shared roles for a stylesheet that predates these.
The ClickHouse theme takes Click UI's button.basic.space.x and space.gap,
and the drift guard pins both values and measures the themed Button's
padding and gap.
* test: Reach the Temporary Chat Chip Through a Sent Message
The header status chip only appears once a temporary chat has its first message, so the scenarios send one, and measure the real message bubble instead of a probe for the shared spacing.
* fix: Let the Control Spacing Roles Replace the Base Gap
Tailwind Merge could not classify px-theme-control-x or
gap-theme-control-gap, so a theme-sized Button kept the base gap-2 beside
the role and a caller's own padding beside both, leaving the winner to
stylesheet order. Register both utilities with their groups.
* fix: Alias the Stock Control Spacing to the Shared Spacing Roles
A stylesheet that set only --theme-space-normal or --theme-space-compact
stopped reaching Button, Chip and the composer controls once the stock CSS
declared the control roles as literals, so the preset's fallback never ran.
The stock declarations now read the shared roles, with the same defaults.
* test: Assert the Control Spacing Utilities Declare Their Own Properties
* 🎒 feat: Carry Project Context Across Conversations
* fix: fail a project attach whose upload hold cannot be released and keep file deletes independent of project cleanup
- Attaching a project file now fails, with nothing attached, when the file's
temporary upload hold cannot be released, instead of attaching a file that
can still expire.
- A failed project cleanup after a file delete is logged rather than failing a
delete that already happened, and a user-wide delete clears that user's
project references without reading every file first.
- Both resize handles share the orientation-aware grip, so a primary handle in
a vertical group renders a rotated grip like the alternate one.
* fix: treat an empty tenant id as no tenant in the project file picker
* fix: attach nothing when a project file is deleted while the attach is in flight
* fix: localize project file association failures instead of showing server text
* test: expect a localized reason for a rejected project file association
* fix: scope project file listing to the caller's tenant and name the file limit when attaching an existing file
- Listing a project's attached files now returns not found when the project
belongs to another tenant, matching the available-file listing and the
chat context resolver.
- Choosing an existing file for a full project now names the project file
limit instead of the generic attach error, like the upload path.
* fix: treat an empty tenant id as no tenant in every project tenant check
A caller whose tenant id is an empty string was refused its own tenantless
project when listing attached or available files, because the handlers
compared '' with null. The project handlers and the chat context resolver now
share one comparison that treats an empty tenant id as no tenant, the rule the
project file queries already follow.
* test: wait for the workspace layout swap and the mobile drawer before acting in the long-metadata spec
The desktop title was measured while the phone layout could still be the one
rendered, and the sidebar row menu was opened before the mobile drawer
committed its open state, which closes a menu opened ahead of it. The spec
now waits for the workspace's phone-only sidebar toggle to leave and for the
drawer to drop inert, the same drawer signal the agents helpers use.
* fix: keep the agent context fingerprint of a turn outside any project unchanged
Every turn now passes a project context key into the agent context
fingerprint, and a turn outside any project passed the no-project sentinel,
which changed the digest of every existing no-project checkpoint. An agent
event conversation written before this change then cold-started and reloaded
its history on its next event. The fingerprint now hashes the no-project key
as absent, so those checkpoints keep matching.
* test: load the long-metadata spec at phone width before driving the mobile drawer
The spec crossed back from desktop to phone width without a reload, which
leaves the drawer on the desktop toggle path whose delayed focus handoff to
the close button dismissed the row menu roughly 200ms after it opened. The
spec now reloads at phone width, as a phone arrives, and waits for the
drawer's own focus handoff before opening the menu.
* fix: Paint Tool Output Panes With the Code Body Role
The output panes of code execution, file authoring and file reads painted
surface-chat in light and surface-primary-alt in dark. surface-code-body is
that same pair in the default, dark and high-contrast themes, and a theme
that styles its code blocks now styles these panes with them.
* fix: Draw the MCP OAuth Logo and QR Tiles From Fixed Roles
The server logo tile and the QR backdrop were bg-white. They take
surface-fixed and surface-qr, both white in every bundled theme, so a theme
can own them without the QR losing its quiet zone.
* fix: Give TextareaAutosize an Invalid State and Use It for Tool Arguments
The tool-approval edit field flagged invalid JSON with border-red-500, a raw
colour the primitive's owner rule rejects in any semantic form. TextareaAutosize
now draws its border in border-destructive when aria-invalid is set, the
attribute the Skills description fields already pass. ToolApproval sets
aria-invalid and points aria-describedby at the Invalid JSON message, so a
screen reader announces the error as well.
* docs: Record the Image Preview, Artifact Scrollbar and Azure Chip Colours
Adds the image preview's media overlays, the sandboxed artifact document's
scrollbar and the Azure file-source gradient to the theme allowlist.
* test: Cover the Tool Output Pane, Invalid Tool Arguments and OAuth QR Tile
Scenarios for the code-body pane behind a file-authoring card, the
announced invalid state of the tool-approval edit field, and the MCP OAuth
QR backdrop staying white in both modes.
* test: Retry the Output Pane Colour and Set Up the Approval and OAuth Scenarios at Desktop Size
The pane is read with a retrying assertion against the resolved role, since
the card re-renders as highlighting lands. The approval agent and the MCP
Settings panel are reached through side panels the mobile layout keeps in its
drawer, so those two scenarios run at desktop size on the mobile project.
* feat: Give the Table Theme Roles for Header Text, Row Rules and Density
The shared table could not take Click UI's shape: its header had no fill of
its own, column names were fixed to text-secondary, rows had no rule and
cells were fixed at 1rem. The header now always carries surface-secondary,
which every consumer already painted, and three roles are added: a
table-header-text color that falls back to a theme's secondary text, a
tableCellSpaceY length that sets cell padding and header height, and a
tableRowStroke width that is 0px by default. Tailwind Merge learns the
new padding and height utilities so a caller's own padding still wins.
The ClickHouse theme sets Click UI's header title color, small-size cell
space and 1px row stroke, and the drift guard pins them.
* fix: Carry the Column-Header Role Into Both Data Tables
Both exported data tables set text-secondary on the header cell or its
label, which beat the new role, so a theme never reached their column
names; they now read table-header-text. The legacy table's row titles keep
text-secondary explicitly, since they are row data rendered through
TableHead. The token falls back to --text-secondary for a stylesheet that
predates it.
* refactor: Give TableHeader a filled Variant Instead of a Caller Override
* feat: Route the Data Tables Through Table Density and a Header Fill Role
The density role reached no real table: every data table passed its own
py-1 or py-2. TableCell and TableRowHeader gain compact and dense sizes,
and TableHead sm, compact and row sizes, all derived from the table cell
space (half and a quarter of it), so the three data tables now read the
role while drawing exactly what they drew before. The virtualized table's
rows default to 32px plus half the cell space, 40px as before.
The virtualized table's self-sticking header cells now paint a
table-header-fill role that defaults to each theme's dialog surface, so
Shared Links and Archived Chats look the same by default and take Click
UI's header fill under the ClickHouse theme.
ClickHouse's cell space becomes Click UI's md value, 1rem: its sizes halve
the same way, so the compact tables land on sm, 0.5rem, the size Click UI
tables default to.
* fix: Re-render Mounted Table Rows When the Themed Row Height Changes
The memoized row compared only its data, selection and cells version, so a
theme switch left mounted rows on their old inline height and the
virtualizer measured that stale geometry. It now compares the height too.
* fix: Size Both Virtualizers From the Themed Row Geometry
The legacy table estimated every row at a fixed 48px while its title cells
now follow the cell space, and the dense row budgeted 32px of content that
is really 2rem of controls. A useTableRowHeight hook gives both the dense
and the titled height from the root size, the cell space and the row rule,
48px and 40px as before, and the legacy virtualizer re-measures when it
changes.
* fix: Measure the Legacy Table's Rendered Rows
The legacy table can be given columns without a title, whose rows are
shorter than a titled row, so the estimate alone could not match them.
Each rendered row now reports its size to the virtualizer, and the titled
height is only the first guess for rows not yet rendered.
* test: Compare Archived Header Space Apart From the Theme's Text Line
ClickHouse now draws text-sm on its own 1.5 line height, so its header is
a pixel taller than the default while keeping the same cell space. Compare
the space around the text line instead of the whole header height.
* fix: Keep the Row Rule Inside the Titled Row Estimate
The title cell draws the rule as its own bottom border inside a fixed
border-box height, so adding the stroke again overestimated every
unmounted titled row by the stroke width. Only the content-sized dense
and compact rows grow by it.
* test: Read the Table Defaults From the Published Stock Stylesheet
The table roles' stock values and the no-rule stroke now live in the
package's defaults.css, which replaces controls.css, and the stock
palette counts the two new table color roles.
* fix: Draw Controls at the Theme Control Radius
Dropdown gains a shape prop in Button's vocabulary. Omitted, the trigger
takes the theme control radius, which equals rounded-xl in the default
theme, so the settings selects follow a theme's control corner instead of
its dialog step. The marketplace sort and the prompts filter ask for
shape="default" rather than overriding the radius locally.
The settings nav tabs, the agent builder footer buttons and the builder
action controls move from rounded-xl to the control radius.
The Click UI drift guard now measures primitive color and shape parity
against Click UI's component tokens and holds a floor per mode.
* test: Choose the Theme per Navigation Instead of Stacking Init Scripts
* fix: Paint Panel Surfaces From Theme Roles
The tool dialog's logo tile takes the fixed surface that logo tiles use elsewhere,
and the skill image preview icon takes a series slot like the other categorical
marks. The tools category filter drops a raw emerald border and shows its active
state on the filter icon in the accent role instead; the label still names the
active category.
The avatar cropper's media hint and the SharePoint picker background are recorded
in the theme allowlist.
* test: Cover the Skills Category Filter's Active State
* test: Match the Category Filter Trigger Exactly
* fix: Paint Chat Surfaces From Theme Roles
The code window header and skill output panel stop repainting their
surface-secondary container with a per-mode fill that matched it in every bundled
theme, the drag-and-drop backdrop becomes the theme scrim, the OpenAI file-source
chip takes the primary surface, and the banner drops a stopless dark gradient.
ActionIcon, RetrievalIcon, WrenchIcon and the message ProgressCircle had no
importers and carried fixed Lottie colours, so they are removed. The media,
artwork and exported-document colours that stay fixed are recorded in the theme
allowlist.
* fix: Keep the Drag-and-Drop Backdrop Black Until Scrim Opacity Is a Role
The theme scrim is gray in default light, so at the same alpha it dimmed the page
less than the black backdrop it replaced. The backdrop keeps its original colour
and the allowlist records the scrim-opacity role it waits on.
* test: Cover the Code Window Header Showing Its Panel Surface
* test: Exercise the Code Window Header Through a File-Authoring Card
The mock code interpreter exposes bash_tool, whose card has its own header; a skill
file created by the model opens the file-authoring card that renders CodeWindowHeader.
* fix: Keep the OpenAI File-Source Chip's Dark Opacity
The chip composites over the file's own preview, where surface-primary at 75% darkened
it against the shipped black at 65%. It keeps its colours and joins the code-execution
badge in the allowlist as a media overlay.
* fix: ship the Inter, Roboto Mono and Inconsolata faces with the package
Move the woff2 files from client/public/fonts into packages/client/src/theme/fonts
and every @font-face rule (the app's Inter and Roboto Mono faces and
client/src/fonts.css) into theme/fonts.css, which tokens.css imports and the build
copies into dist with the files. Each url() names the file by a new
@librechat/client/fonts/* export: a relative path breaks under Vite with
@tailwindcss/postcss, because Vite inlines the import before Tailwind runs and
Tailwind reports every emitted declaration as coming from the root stylesheet.
The app loads the same files through the same import and still emits them to
assets/fonts with their licences, now read from the package; the $fonts alias is
gone. Inter and Roboto Mono gain their OFL texts beside Inconsolata's. The README's
consumer stylesheet moves the package stylesheet import above @config, where
Vite's inliner no longer skips it. The static checks' bundle freshness test reads a
subpath pattern export as the directory it maps into.
* docs: describe how the theme fonts resolve through the package export
* fix: leave subpath pattern exports out of the bundle freshness check
A pattern such as ./dist/fonts/* names no entry the design rules load, and
reading it as its directory made the freshness check depend on a directory the
synthetic build in suppression-baseline.spec.ts never dates. The scenario's
mirror of the manifest reading skips patterns the same way.
* fix: publish the theme defaults and scale mapping with the package
Move the radius, font and shadow utility mapping from the app stylesheet into
tokens.css, and the stock values of every theme-owned property (the 94 color
roles, the appearance roles and the brand colors, with the primitive scales they
are written against) into a new defaults.css that tokens.css imports and the
build ships beside it. The app reads both through its existing tokens.css
import, so @librechat/client/theme.css now gives a consumer the same utilities
and values the app renders. controls.css folds into the defaults, and the preset
no longer pins rounded-sm, rounded-md and rounded-lg to --radius.
A new mock scenario compiles the app and a consumer from what each loads,
checks the build output and exports, and compares computed styles in light and
dark; defaults.spec.ts holds defaults.css equal to the runtime themes.
* docs: note that hosts load the stock Inter and Roboto Mono faces
* perf: keep the mermaid chunk off the boot path
The chunk groups capture each module's dependencies with it, and mermaid
depends on Vite's dynamic-import preload helper, the Buffer and process shims,
DOMPurify, uuid and dayjs. Those landed in the 3 MB mermaid chunk, and because
the boot chunks import them, every page preloaded and evaluated mermaid before
its first API request, on the path to the largest contentful paint.
Two higher-priority groups claim them first (runtime-shims and shared-libs), so
mermaid loads only through useMermaid's dynamic import. A new mock scenario
checks that a chat without diagrams never fetches the chunk and that a diagram
reply still loads and renders it.
* test: wait on the reply and the composer instead of network idle
The app keeps connections open, so network idle never arrives.
* fix: Apply controlled themes before paint and give prose links, inline code and QR backdrops theme roles
ThemeProvider folds a controlled theme prop change into state during the
render that delivers it and applies the theme in a layout effect, so a
deployment theme, a shared link's theme and the viewer's own theme reach
the root before the commit paints. Storage writes stay after the commit.
Adds rgb-link-prose (link in light, primary text in dark, following a
theme's own link) and registers rgb-surface-qr, and paints inline prose
code with the surface-active-alt and surface-hover-alt roles.
* test: Record the frames a controlled theme paints and the prose and QR roles
* feat: Give the Dialog Scrims Theme-Owned Opacity Roles
The three dialog families carried their scrim alpha in each utility
(bg-surface-overlay/80, /90, /65), so a theme could recolor a scrim but
not set how much it dims. Add scrimOpacity, alertScrimOpacity and
modalScrimOpacity, exposed as bg-scrim, bg-scrim-alert and
bg-scrim-modal in the published tokens, defaulting to the 80%, 90% and
65% each family drew.
ClickHouse takes Click UI's 0.75 from dialog.color.opaqueBackground,
cited in the drift guard. A bundled scrim never lifts the page and, on
a light canvas, separates the dialog by 3:1, which is why ClickHouse
dark keeps a black overlay rather than Click UI's lighter gray.
* test: Read the Dialog Scrim Through Its Role Class in the Morph and Settings Scenarios
Both scenarios pinned the old bg-surface-overlay/80 literal: the morph
counted elements carrying it and Settings compared its scrim with a
probe of it, whose color-mix value serializes differently from the
role's rgb(). They now use bg-scrim, the class DIALOG_SCRIM_CLASS
exports.
* feat: Give Display Type and the Type Scale Their Own Theme Roles
Add a displayFontFamily appearance role, which the dialog titles compose
through font-display, and textXs..text2xl and leadingXs..leading2xl,
which the app's text-* utilities now read like the radius and shadow
scales. The defaults are Tailwind's own sizes and line heights and the
UI family, so the default theme is unchanged, and a theme that names
only its fontFamily keeps its headings in it.
ClickHouse takes Click UI's display family, font sizes 1 to 6 and its
1.5 product line height, cited in the drift guard. Basier Square is
named but not self-hosted: it is a commercial face, so browsers without
it fall through to the Inter stack.
* fix: Publish the Display Family Utility With the Theme Tokens
* fix: Keep the ClickHouse 2xl Step Below the Unthemed 3xl
Click UI has no size between 1.25rem and 2rem, and 2rem would overtake
Tailwind's unthemed text-3xl (1.875rem), so a responsive text-2xl to
sm:text-3xl heading shrank as the viewport grew. ClickHouse keeps
LibreChat's 1.5rem for 2xl at Click UI's line height, recorded as a
decision in the drift guard, and a registry test holds every bundled
2xl below 3xl.
* feat: Give Pressed and Disabled States Their Own Theme Roles
Add surface-pressed and surface-inverted-pressed, which IconButton takes
while held under a hover. They default to the hover fills a pointer
press has always shown, and a theme that names only its hovers presses
in them, so the default theme is unchanged.
Add surface-disabled, text-disabled and border-disabled, and a
disabledStyle appearance role: dim keeps the half-opacity treatment,
fill paints a disabled Checkbox, Switch, Field or IconButton in the
disabled roles through the theme-disabled variant. The root mirrors
the choice as data-theme-disabled. ClickHouse fills from Click UI's
disabled and active tokens, cited in the drift guard.
* fix: Ink Disabled Placeholders and Press the Submit Button in the State Roles
* fix: Ignore Color Roles a Client Predates and Fill Every Disabled Composer Slot
A cached client reading a newer server's inline theme rejected the
whole definition over one color role it did not know. The client
reader now ignores a well-formed rgb- role holding a triplet with a
warning, as it already did for appearance keys; the server keeps
rejecting unknown roles, where one is a typo in librechat.yaml.
The disabled fill moves into composerSubmitClasses, so the plain send
and stop buttons in Composer and DuringRunSendButton take it too.
* fix: Share One Disabled Fill Recipe Across Every Shared Control
Select, InputNumber, SecretInput, FilterInput and the Slider thumb only
dimmed under a fill theme. The theme-disabled classes now live in one
disabledFillClasses recipe that every shared control composes, and the
variant also matches Radix's data-disabled so a disabled slider thumb
takes the fill.
The deployment-theme scenario for an invalid theme now serves a
malformed color value, since a well-formed role the client predates is
ignored rather than rejected.
* fix: Fill a Disabled InputNumber Through Its Wrapper
rc-input-number puts the component's classes on a wrapper around the
disabled input, so the theme-disabled variant never matched it. Add a
theme-disabled-within variant for a wrapper that contains a disabled
control, and a disabledWithinFillClasses recipe InputNumber composes;
the input inherits the ink and shows the fill through its transparent
background. The default theme's disabled InputNumber is unchanged.
* fix: Carry the Disabled Roles to Every Shared Disabled Surface
Sweep every shared component that dims when disabled. Button-like
controls (AlertDialog and Dialog footer buttons, DialogTemplate's
select, the MultiSelect trigger) compose the fill recipe. Menu items,
tabs and the send-actions rows keep their surface and take the ink
through disabledInkClasses, as Click UI's genericMenu does, and Label
and FilterInput's floating label follow their control through
peerDisabledInkClasses. The variant also matches aria-disabled rows.
OGDialogTemplate and Button stay with the open PR that owns them.
* fix: Gray the Icons of Disabled Menu Rows Under a Fill Theme
A send-action or menu row's icon owns its color (text-status-warning,
text-text-secondary), so the inherited disabled ink left it bright.
The item ink recipe now recolors descendant icons too.
* fix: Fill the Dropdown Trigger and the OTP Group Under a Fill Theme
The default Dropdown trigger only dimmed, and its disabled:hover resets
tied with the fill; the shared recipe now pins the disabled hover fill
and ink too, and the trigger composes it. InputOTP dims its group with
has-[:disabled], so it takes the wrapper recipe.
* fix: Reach Child-Owned Ink and Slot Edges in the Disabled Recipes
A disabled control's own children, the SelectDropDown labels or a
checkbox glyph, kept their explicit ink, and a disabled OTP group's
slots kept their border. The fill and item recipes now carry the
disabled ink to descendants, the wrapper recipe carries the disabled
edge and ink to the group's parts, and SelectDropDown composes the fill
recipe.
* test: Probe Disabled Row Icons With the Recipe's Descendant Selector
* fix: Give Radio Segments, SecretInput Actions and DropdownPopup Items the Disabled Ink
These three dim their disabled state through a JS branch or Ariakit's
aria-disabled rather than a disabled: utility, so the earlier sweep
missed them. Each composes the item ink recipe, whose opacity reset
outranks the branch's opacity-50.
* feat: Give Keyboard Focus Its Own Theme Roles
Add rgb-focus-outline for the app-wide keyboard focus outline and
rgb-focus-control for the shared primitives' focus ring. The defaults
reproduce the default theme exactly: a black and white outline, and
primitive rings in the primary text ink. A theme that names only
rgb-ring-primary keeps drawing its outline in that ring, and one that
names only rgb-text-primary keeps ringing its controls in that ink, so
the data-theme-ring marker is no longer needed.
ClickHouse sets both roles to Click UI outline.default, cited in the
drift guard. Every bundled palette holds both roles at 3:1 on its
canvases. Button keeps its ring until the open PR that owns it lands.
* test: Read the Opaque Ring Layer in the Focus Ring Probe
* fix: Draw the Popover Focus Outline in the Control Focus Role
* refactor: Supply App-Global Chat Settings From the Host
The chat hooks read the during-run default, steer interrupts, save drafts,
temporary chat and the artifacts reset from a ChatSettings context that Root
fills from the existing Recoil atoms. liveAppliedSteerIds moves to Jotai with
its reader and writer, and manual compaction submits through the useChat
facade.
* fix: Keep Compaction Off the Message Frame and Read One Interrupt Default
Compaction reads status through useChatActions, which re-renders only when
the status changes instead of mapping every stream frame. The during-run send
button reads the interrupt default from the steering controls instead of a
second store read.
* fix: Supply Only Preferences Every Chat Reader Can Take From the Host
saveDrafts and isTemporary are also read by useNewConvo, useAutoSave and the
upload path, and by ChatView, ChatForm and the resume hook, so a host value
for them would split one preference across two sources. They stay on the
store until all of their readers can move together.
* fix: Fall Back From an Invalid interface.theme Instead of Exiting
A mistake in interface.theme failed the whole librechat.yaml and stopped
the server. The loader now checks the theme with the rules the client
applies before painting it, drops a theme the client would reject with a
warning naming each problem by path, and loads the rest of the config.
Unknown appearance tokens only warn, matching the reader. Startup and
reload behave the same.
The token lists and the validator move to librechat-data-provider so the
client registry and the server check against one list.
* fix: Reject Unknown Bundled Theme Names at Config Load
A misspelled bundled name such as clickhous passed the server and was
rejected only in the browser. The bundled names now live in
librechat-data-provider; the client's theme map is typed against that
list and the loader drops a name outside it with a warning.
* test: Pin the ClickHouse Theme to a Click UI v0.12.0 Snapshot
Add clickui.json, the Click UI v0.12.0 tokens the ClickHouse theme cites,
and a drift guard that matches every color, radius, font, shadow and motion
value to its token, naming both on a mismatch. Correct the citations that did
not resolve, set shadow2xs, motionNormal and the light code text from Click
UI, and document the dark scrim as a deliberate deviation.
Draw the global keyboard focus outline in ring-primary under any theme
definition, keeping the default theme's black and white outline and the
high-contrast rule unchanged.
* test: Anchor the Focus Probe Tab on a Focused Sentinel
Tab moves on from the previously focused element after a blur, so it never
reached the prepended probe. Focus a sentinel outside the tab order first.
* test: Require Three RGB Channels in the Click UI Drift Guard
A truncated theme triplet such as '0 138' matched a source whose leading
channels agreed. Compare only complete, finite triplets.
* fix: Follow the Theme Ring Only When the Theme Names One
A theme that leaves rgb-ring-primary to the default resolves it to #595959,
which a dark surface carries at about 2.3:1. The provider now marks the root
with data-theme-ring when the active mode names its ring, and the global
focus outline follows the ring only under that mark; every other theme keeps
the default black and white outline. Each e2e mode also gets its own page
instead of relying on init-script order.
* 💬 refactor: ChatForm UI redesign with unified palette
Replace the scattered composer controls (badge row, tools dropdown, attach
menu, floating steer chips) with one system: a searchable palette with
favourites, a composer bar holding active tools plus effort and speech
controls, a tray staging files and manual context, and a queue rail that
buffers messages while a reply is running.
Collapses the 94 development and review-fix commits of PR #14473 into one
commit so the branch could be rebased onto dev once; the upstream merge of
this pull request is a squash merge either way. The pre-rebase history is
kept locally at tag prerebase-backup (afde3e4b9a).
* fix: finish the dev rebase of the composer redesign
The rebase onto dev left four defects the merge itself produced, plus one
review finding the branch had never answered:
- ChatForm's `min-w-0` JSX comment lost its closing brace, so the file did
not parse at all.
- `useSteering` carried both dev's local draft consumer and the host-injected
one under the same name; the host's `consumeDraft` owns the storage key
autosave actually parked on, so the local copy is gone.
- `useUploadOptions` kept dev's `awaitingAgentProvider` gate after the merge
replaced the block that derived `isSavedAgent`; the predicate is back, from
the same `isEphemeralAgentId` the shared allowance helper uses.
- `MessagesView` still reserved a band for the in-flight steer overlay and
read `steerOverlayHeightFamily`, which nothing writes now that pending
steers render inside the streaming reply. The reservation is gone and
`ScrollButton`'s lift is optional; composer panels keep standing the button
down through `composerOverlayCountFamily`.
- The aggregate MCP chip keyed on `mcp:pinned`, which a server literally
named `pinned` also produces (`mcp:${serverName}`), so both entries shared
one React key and one packing slot. It uses a key outside that namespace.
Also drops the locale keys whose only consumers the redesign deleted, ports
dev's extracted-text delivery path into `useAttachExisting` (every
attach-existing surface routes through it now), repoints the mock e2e
built-in helpers at the palette, and repairs test fixtures the merge left
stale (a query mock that blanked agent metadata, a Jotai store mismatch, an
ambiguous tray query, an ephemeral-agent fixture standing in for a saved one).
* test: cover the composer redesign's acceptance scenarios end to end
Twelve mock-harness scenarios, one `@scenario:` tagged test each, over the
behaviours this change is accountable for: the single searchable palette
(open, search, favourite across a reload, keyboard-only operation), the
composer bar and tray (active tool chip, a staged file riding with the
message, a per-turn thinking level that applies to the next turn only), the
queue rail (Enter buffers during a run and drains after it, keyboard reorder
changes the send order, the draft survives run completion), and the composer's
portaled surfaces (Stop reaches the run from inside a portal, a pending tool
approval in the bar resumes the run).
Two contract scenarios are deliberately absent, with reasons: split-pane Stop
isolation is not runtime-reachable because `ChatRoute` mounts one pane, and
code-approval mode mid-run needs an attached stateful code environment the
mock lab does not configure, so dev's `ChatForm.approvalMode` unit spec stays
its coverage.
* fix: make the composer's acceptance run and the whole client suite green
The first verification run of the twelve scenarios passed six and failed six,
and the full client jest suite (which CI runs sharded, unlike the diff-scoped
local precheck) failed fourteen tests. One of each group was a real defect; the
rest were queries and fixtures still aimed at surfaces the redesign replaced.
Defects:
- `useQueueDrain` held its send lock after a revealed queued turn settled, so
the next server-owned row could never be selected. The reveal path now owns
its own lock and releases it when the revealed turn clears.
- `ChatForm` recomputed the during-run slot without `consumeComposer` in its
dependency list, and still carried a landing flag the host prop replaced.
Tests and lab:
- The five `ChatForm.attachments` focus cases now drive the palette dialog and
the tray's quote chips instead of the deleted attach menu and quote popup,
with the same strict focus assertions; the harness gains the `main` landmark
the palette portals into, and the touch fixture a complete `matchMedia`.
- `Palette.spec` no longer closes a hoisted `jest.mock` factory over a later
`const`; `useSteering.spec` follows dev's toast wording.
- The six failing scenarios were repointed at what this lab actually renders:
uploads go through the palette's local-computer row, the reasoning control
needs `interface.parameters`, the queued-row contract is the row and its
drain rather than a control the submit consumes, the palette's keyboard path
uses a built-in the lab grants, the slow-reply wait uses the fixture's chunk
marker, and the approval scenario selects its agent through the builder flow
the passing approval specs use.
* fix: close the eight findings Codex raised on the rebased head
Codex reviewed the rebased heads and reported eight issues; all eight were
real. Two of them concern code this session added.
Behaviour:
- A custom or OpenRouter endpoint that declares neither `paramDefinitions`
nor a `reasoningFormat` no longer gets a reasoning control inferred from the
generic settings list. Capability now has to be declared (a reasoning-key
parameter definition, a non-disabled `reasoningFormat`, or a known reasoning
provider as the default params endpoint); a `disabled` format always
resolves nothing, and every consumer — the composer control, the pre-send
validation and the server — asks the same resolver, so a gateway backed by a
non-reasoning model can no longer offer a level the turn would fail on.
- Restoring a queued attachment keeps its `llmDeliveryPath`, so an extracted
document is routed as its text rather than as raw provider content.
- The placeholder-to-durable reasoning migration now requires this pane's own
live submission, so opening another conversation from a new chat no longer
carries the pending level into it. The created/sync case it was written for
still carries.
- The queue rail treats an acknowledged server-owned row as an ordering
boundary: local rows still reorder among themselves, a server-owned row
cannot move, and a move that would cross one is refused out loud — the
queued-turn API has enqueue, list and cancel only, so a client-side reorder
across that boundary could not be made authoritative.
- The palette resolves an agent's skill scope with `resolveAgentSkillsScope`
instead of treating an empty list as an open catalog, and the attach target
resolves `useResponsesApi` with `resolveUseResponsesApi`, matching what the
run will actually do.
Policy:
- The reasoning-override behaviour moved out of the CJS middleware into
`packages/api/src/modelSpecs/reasoningOverride.ts`, which returns a
discriminated result; the middleware is wiring again, and the behavioural
cases moved with it.
- `showComposerTips` and `showFilesDialog` are Jotai atoms now, the tips
preference keeping its storage key and reaching the composer as a prop.
* fix: make the redesigned composer reachable from the mock e2e helpers
CI's Playwright shard failed four `file-provisioning` tests on this branch
because the shared upload helpers still clicked `#attach-file-menu-button` and
`#attach-file-button`, entry points the redesign removed — attaching now goes
through the palette, which also owns the file input. Both helpers drive the
palette now, `uploadViaLegacyOption` mapping its two destination names onto the
palette rows that route to the code environment and the vector store, and the
specs that drove the old surfaces directly (`code-upload`, `thread-renderers`,
`unified-upload`) follow.
Also closes the last client/server disagreement the acceptance run exposed: a
custom endpoint that declares a reasoning key in `paramDefinitions` had that
declaration merged into the provider defaults and then discarded by model-name
inference, so the composer offered a level for `mock-model-a` that the server
rejected with "Invalid reasoning override". An explicitly declared key now
selects the merged setting before inference, while an undeclared generic custom
or OpenRouter target still resolves nothing.
The twelfth acceptance scenario is rescoped: a paused approval keeps the
composer usable and resumes to completion, but the composer-hosted approval
panel and chip are not asserted, because the pending action is keyed by the
stream's conversation id while the composer stays bound to the new-chat
sentinel — in code byte-identical to dev, and mirroring the write onto the
sentinel did not surface the panel. Tracked at berry-13/LibreChat#39.
* test: finish the acceptance run and keep the client typecheck green
- The thinking scenario read `reasoningOverride` from a `userMessage` wrapper
the chat request does not have; the trace shows the request carrying
`{ key: 'effort', value: 'max' }` at the root, so the assertion reads it
there. Source behaviour was already correct.
- The approval scenario timed out on mobile because the Agent Builder drawer
stayed open after selecting the agent and swallowed the pointer events meant
for the composer; it now closes the sidebar and waits for the drawer to go
inert before sending.
- Two specs typed a named custom endpoint's label as an `EModelEndpoint`, which
CI's client typecheck rejects (the label is exactly what the case is about).
All twelve acceptance scenarios now pass in desktop light, desktop dark and
mobile emulation.
* fix: close the nine findings Codex raised on the verified head
Two were defects in code this session added while closing earlier findings:
- A reasoning key declared in `paramDefinitions` but absent from the generic
settings list was accepted as capability and then never inserted, so the
resolver exposed and validated the generic `reasoning_effort` instead. A
declared key is now appended to the effective settings and resolved before
any fallback, for `effort`, `thinkingLevel` and `thinkingBudget` alike.
- The queue's reorder guard validated from the destination row's own index, so
`target === index` rejected every pointer drag and only keyboard reordering
worked. It validates the dragged row's identity now; the server-owned
boundary, its announcement and the disabled grips are unchanged.
The rest:
- Stop no longer falls through to the document-wide control once a focused
pane has been resolved: focus in an idle pane while another generates used
to abort the other pane's run.
- `PendingSteers` is mounted once in `MessagesView`, above the renderer
switch, so a sent or failed steer keeps its in-flight, retry and
send-as-new surface with `FLAT_THREAD` off, where the recursive renderer
had none and the composer no longer carries a fallback.
- The file-list policy moved out of the CJS route into
`packages/api/src/files/list.ts`, and its cap is `interface`-level config
(`fileListLimit`, default 100) instead of a literal in the route.
- The steer-arm confirmation window is `interface.steerArmConfirmationTimeoutMs`
with its current 10s as the default, travelling the same path as the other
client-consumed timings.
- The pause predicates judge the active branch's tail rather than the last
assistant in array order, so an inactive sibling branch can no longer
enable or disable the inline escalation control.
* fix: move persisted reasoning validation into api package
* fix: restore composer entry points and defaults
* style: normalize reasoning helper formatting
* test: open mobile sidebar before account menu
* test: use settled sidebar layout for mobile account menu
* test: handle empty mobile sidebar in account scenario
* test: target account menu in mobile drawer
* fix: keep composer surfaces and mobile drawer reachable
* fix: let thread approvals receive pointer events
* fix: stabilize composer steering and shortcuts
* fix: commit queue drags and centralize file modal
* fix: return a canceled steer's text to the composer
Canceling a pending steer dropped the words the user had written: the
bubble disappeared and nothing carried the text back, so a reclaim that
the server confirmed still lost the message.
Hand the reclaimed text to `pendingComposerTextByConvoId`, which the
composer already drains on its first render, and append rather than
overwrite when the composer holds a draft the user typed while the cancel
request was in flight.
* fix: close the six findings Codex raised on the pre-rebase head
Pending steers keyed off the Recoil conversation id, which lags the route
during warm-cache navigation, so the surface could sit at the destination
thread's tail while its Cancel and Escalate acted on the source run. It
now keys off the rendered tree, as the mount window already does.
Arming a steer removes the control that was activated, dropping keyboard
focus on the document body. `useSteerEscalate` already exposes `onArmed`
for exactly that; the bubble now passes one that moves focus to the row's
Cancel action.
Opening the file manager from the account menu left its dialog with no
opener: the modal rejects an active element inside a menu, and nothing
supplied a fallback, so closing it returned focus to the body. The opener
is published alongside the open flag and cleared on close, which keeps the
shortcut path's composer-focus behaviour intact.
A steer the server rejected was swept into the queue with no marking, and
the run-end drain then sent those words as a fresh turn without the user
touching Retry or Send as new. Converted failures now carry
`needsExplicitSend`, which the drain skips the same way it skips a row the
rail is mid-edit on.
The queued-turn reconciliation window was a hard-coded 60s, unreachable by
an operator behind a slow proxy or a delayed read replica. It ships as
`interface.queuedTurnReconciliationTimeoutMs` with that value as default,
beside the steer arm timeout it belongs with.
Reasoning-override shape validation moved to
`parseReasoningOverrideRequest` in packages/api, leaving the middleware to
map a rejection onto its own error response.
* fix: keep the composer's reasoning level in step with the parameters panel
`stableConversation` memoized the conversation without any reasoning field
in its key, so editing the level in the Parameters panel produced a new
conversation object that every listed dependency ignored. The composer
went on showing the old level while the next ordinary send already used
the new one.
* test: stop the sidebar helper waiting for a drawer desktop never renders
`ensureSidebarOnScreen` measured `#mobile-drawer` whenever the sidebar had
no Pinned region. Desktop renders no drawer, and `boundingBox()` on a
locator that matches nothing waits for it, so every scenario that reached
that branch at a desktop viewport timed out instead of continuing. Check
for the drawer's absence rather than measuring it.
* test: open the mobile drawer from its inert state, not its geometry
A closed drawer stays mounted and keeps its layout box, so measuring the
Pinned region reported it on screen while every control inside was still
untappable. `inert` is the state the component actually sets when it
closes, so the helper reads that and clicks the header opener before
anything else measures.
* fix: return a canceled steer whole, through one recovery boundary
A steer that stops being a steer was handed back three different ways,
and each one lost something. The immediate cancel wrote `steer.text`
into the composer handoff atom, dropping the attachments, quoted
excerpts, manual skill picks and reasoning override that `settleReclaimed`
then deleted with the pending object. The cancel that won its race
against its own POST cleared the pending marker and restored nothing at
all, so those words disappeared. A definitively rejected steer could
only be resent unchanged, because the row offering to edit it was gone.
Give the recovery one boundary. `useSteerRehome` hands the whole steer
to the composer through the guarded restore the queue rail already uses,
and converts it into an ordinary queued follow-up when the composer
refuses: a draft in the box, another chat on screen, a paused question,
or a composer that unmounted while the reclaim was in flight. The chip
goes either way, since a pending row claiming the server still holds the
words is a lie once they live somewhere else.
The composer is a sibling of the message list rather than its ancestor,
so the thread had no way to reach that restore. The pane's host now
holds the reference and both sides read it there, which also means a
reclaim resolving after the composer is gone refuses instead of writing
into a dead form.
Because the restore refuses rather than overwrites, a draft typed while
a cancel is in flight is no longer glued to the returned steer; the
steer becomes its own queued message and the draft is left as typed.
* test: drive the surfaces these scenarios actually reach
Four scenarios were failing against surfaces the user never touches.
The approval scenario clicked Approve on the thread's copy of the
pending card. The composer's review panel floats over the tail of the
thread and owns the batch submit, so that copy sits behind it and the
click retried until the test timed out. Drive the panel instead; the
thread card stays asserted as the record of the request.
The palette scenario counted rows the moment the dialog painted, then
expected clearing the filter to restore that number. The catalog arrives
in waves: the MCP server list is released by the app-wide warmup timer
and the skills catalog only starts fetching once the dialog's open
effect has run, so the count kept climbing under the test. Wait for the
last wave and for the count to hold still, then snapshot.
The sidebar helper opened the mobile drawer twice through two signals
that disagree. It waited for `inert` to clear, which happens on a state
commit several frames ahead of the slide, then fell through to a
bounding box still reading off-screen and clicked the opener again, into
a machine already opening. One open, through `inert`, then wait for the
drawer to settle in the viewport.
Splitting the cancel scenario follows the recovery change: an empty
composer takes the steer back, and a staged draft keeps the box while
the steer becomes a queued message of its own.
* fix: hold a server-rejected steer for an explicit send
Editing a definitively failed steer falls back to the queue when the
composer will not take it. Queued like any other follow-up, the run-end
drain would send it the moment the reply finished, which turns the
server's refusal into an automatic new turn carrying the payload it
refused.
Mark that fallback the way the rest of the rejected-steer path is
marked, so the row waits for the user to send it.
* test: restore two assertions this branch broke while rewriting specs
Both specs were updated for the redesigned surfaces and both lost a step
they needed, and both fail deterministically in CI as a result.
thread-fold pressed Enter mid-run and waited for a queued row, under a
comment claiming Enter queues by default. It does not: the default
during-run action is steer, so Enter produced a pending steer and the
queued row the next line waits for could never appear. Ctrl/Cmd+Enter is
what takes the other action, which is the one this test needs.
mcp-oauth-readiness asserted on the server card's Cancel button straight
after clicking Connect. Connect opens the OAuth dialog over the panel,
and it is modal, so while it is open the card underneath is marked
aria-hidden and Cancel is not in the accessibility tree at all, which is
why the failure reads "element(s) not found" rather than "not visible".
The rewrite dropped the dismissal the previous flow performed; restore it
as a helper both call sites share.
* test: reach the account menu without the pinned section
The two account-menu tests borrowed `ensureSidebarOnScreen`, which also
settles the Pinned section's placement for the pinned specs. They never
touch that region, and on a run where an earlier test left a pinned
conversation the section exists but has not been laid out yet, so both
tests spent fifteen seconds polling it and failed on the wait rather than
on anything they assert.
Give them a helper that does only what they need: open the mobile drawer
from `inert`, the one signal that distinguishes a closed drawer from an
open one while it stays mounted, then wait for the slide to reach the
viewport. Desktop renders no drawer and needs nothing.
* test: open the drawer by what it shows, not one early read
The helper decided once, from a single read of `inert` taken right after
navigation. On a fresh mobile load the drawer can still be mounting, so
that read reports a state the app has not settled into, the helper
returns believing the drawer is open, and the account menu inside it is
never shown.
Retry against the thing the caller actually needs, the account menu being
visible, re-checking before each attempt. While the drawer is open its
control is labelled "Close sidebar", so the retry cannot toggle it shut.
* fix: keep the composer observing MCP authorization where its tools apply
The old badge row mounted its provider with observeToolAuthorization set
whenever the composer showed ephemeral tools, so on a plain endpoint the
chat kept MCP connection status and discovered tools in the query cache.
The redesigned composer mounts the same provider without the prop, so it
defaults to off and nothing on the chat side loads that state any more.
The Agent Builder's tool library reads that cache. With it empty, every
server reads as Unknown with no tools, and clicking one opens its detail
dialog offering to connect instead of adding it to the agent.
Pass the same gate the badge row used: showTools is the old
showEphemeralBadges expression.
* test: let the palette close before reopening it, and name what is asserted
enableBuiltinTool pressed Escape and returned while the palette was still
animating out. A caller that reopened it straight away resumed the same
popover, scroll offset included, so the Attach rows at the top sat outside
the virtualized window and upload helpers timed out looking for them. Wait
for the dialog to be gone before returning.
The staged-file assertions matched the bare file name, which the palette
now also lists under recent uploads while it closes; scope them to the
tray they are about.
The OAuth dialog dismissal waited for a title built from the server's
display name, but the dialog is titled with the server name.
* feat: make the queued-send lock expiry configurable
The per-pane lock that stops the run-end drain and an explicit Send now
from submitting the same queued message twice gives up a broken claim
after a fixed 60 seconds. Until then every drain and every Send now for
that pane is refused, and a deployment had no way to tune that window.
Expose it as interface.queuedSendLockTimeoutMs beside the queued-turn
reconciliation timeout, defaulting to the same 60 seconds so nothing
changes unless an operator sets it. The interface loader copies each
field explicitly, so the value is added there too; without that line a
configured value would never reach the client.
* fix: decline interrupt-and-steer while reasoning is staged
With a reasoning choice staged, the Interrupt & steer row is disabled,
but its keyboard chord still reached interruptSteer, which fell back to
interruptAndSend: it aborted the run and threw away the partial answer
the action promises to keep.
A staged reasoning choice belongs to a new generation that a steer into
the live one cannot carry, so decline the chord exactly as the row does,
the same way the function already declines while paused on an approval.
The words stay in the composer, where Enter queues them as a turn of
their own.
* fix: wake the queue drain when a late steer retry is queued
Retry marks a failed steer sending, and a sending steer is kept out of
the run-end sweep so its own POST settles it. When the run finishes
while that POST is in flight, the drain spends its one-shot signal on a
queue the steer is not in yet; the late ack or NO_ACTIVE_RUN then moves
it into the queue, where it sat until the user sent it by hand or
another run ended.
Every other late conversion re-posts the last completed run end through
the composer's rewakeDrain. The retry lives in the thread, so the
composer now publishes that function through the same host it already
uses to hand a reclaimed steer back to the draft, and the retry calls it
after converting. It stays a no-op unless the run completed cleanly and
no newer signal is armed, or when no composer is mounted.
* fix: honour thinking-budget ranges wider than 200,000
The request-scoped reasoning override capped thinkingBudget at 200,000
in the shared schema. An operator's paramDefinitions may advertise a
wider range, and the composer renders it, but every value above the cap
was dropped by the control's parse and rejected by the server's before
the range-aware check could run.
That check already exists: resolveReasoningOverride validates the value
against the endpoint's resolved range. Drop the fixed ceiling so the
resolved range is the bound, keeping the integer and lower-bound rules.
The new case covers both directions on a custom endpoint whose range
goes to 500,000.
* refactor: persist the preliminary reasoning override through the shared helper
The agents controller validated the request's reasoning override with
the schema itself before writing the preliminary user message. That is
behavior in the CJS layer, and it duplicated resolvePersistedReasoningOverride,
which @librechat/api already exports and BaseClient already uses for the
same field.
Call the helper instead. A preliminary message is always a fresh turn,
so it is never an edit or a compaction. The two controller specs that
replace @librechat/api with a hand-built mock now take the real helper.
* fix: clear the file manager's state at the session boundary
The file manager's open flag and its opener live in Jotai's default
store, which outlives the authenticated route. A file manager left open
at logout reopened for the next account, still holding a ref to the
previous session's unmounted opener. Reset both in useClearStates, next
to the other composer state this branch already clears there.
* fix: decline interrupt-and-steer on the first turn while reasoning is staged
* test: read the phone layout from the viewport and keep the drawer open
* fix: name both reasons a queued message cannot be edited yet
* test: reuse a published file through the composer palette
* style: move the composer onto the design-system rules
* style: move the file row's new layout styles into classes
* fix: keep the composer's reused thumbnails, focus ring and disabled dimming intact
* fix: bring a steer back when its deferred cancel does not remove it
* fix: treat a run as over only when every pane holding it is idle
* refactor: pass enter-to-send into the composer from its host
* feat: make the palette's recent-file count configurable
* fix: let a file already in the composer be reattached at the limit
* test: hide the chip before replaying a deferred cancel
* refactor: restore the resume request through one packages/api facade
* fix: keep a conversation's staged reasoning across a visit to another model
* fix: keep a connected MCP server's variables configurable from the palette
* fix: resolve every pane-scoped shortcut through the pane holding focus
* refactor: keep every reasoning-override decision out of the /api wiring
* perf: skip paging the skill catalog for an agent that can show no skill
* test: type the reasoning snapshot the override helper records
* test: pin the applied budget and the untouched parameters in the override specs
* style: give the stop button the send button's shared submit recipe
* fix: name only the newline key when no chord sends
* fix: name the stop key only once the stop control can act
* fix: clear every staged reasoning selection at sign-out
* fix: stop advertising send-now while staged reasoning forces a queue
* fix: resolve the reasoning target of an ephemeral agent in the composer and the request
* test: type the staged reasoning the sign-out spec seeds
* fix: check the composer will take a queued row before discarding its parked copy
* style: carry the composer's classes onto Tailwind v4
* style: sort the composer's classes in Tailwind v4 order
* fix: return a refused queued row to the durable queue instead of memory
* i18n: localize the whole attachment-limit message
* refactor: validate the reasoning override inside the packages/api facade
* style: give the composer's submit slot its own IconButton variant
* test: pin the submit slot to IconButton's submit variant
* test: stub the seen-state hook the rebased MessagesView now calls
* fix: turn file search on for an embedded file attached through the shared hook
* chore: record the composer's design-rule counts against the rebased base
* fix: keep the composer's container query inert under Tailwind v4
* test: point the v4 focus scenarios at the redesigned composer's controls
* fix: keep a re-queued row on its own lineage after the run ends
* test: assert a re-queued row carries its lineage
* refactor: pass the dictation preferences into the composer from its host
* test: read the usage atom from the store the ask hook renders under
* fix: read a pending steer's pause from its own pane's branch
* a11y: keep focus on the quote chips until the last one is removed
* fix: keep staged reasoning for the next message across a manual compaction
* test: expect focus on the remaining quote chip after removing one of two
* style: sort the imports the canary rebase merged
* refactor: remove the side-panel files table the account menu replaced
* fix: settle a retried steer's late ack against a racing cancel
A cancel clicked while the retry POST was in flight left the optimistic
marker unread: the ack re-minted a live chip and injected words the user
had asked back, and a failed retry after that cancel lost them entirely,
because the failure path maps chips the optimistic cancel had removed.
The ack path now defers to the marker the way the composer's does, the
failure path re-adds the chip before settling it, and the queued row a
late ack produces binds its recovery identity only under protocol v2,
the same default the composer's degradation path trusts.
* fix: strip a replayed reasoning override that no longer validates on resume
A paused turn's reasoning override is replayed onto its resume request
and then validated as if it were fresh client input, so an operator
changing the endpoint's reasoning config between pause and approve made
the checkpoint permanently unresumable with a 400. The replay now marks
the request, and a replayed override that fails validation is stripped
so the resume proceeds on defaults; a fresh override that fails is still
rejected.
* refactor: derive the enforced spec fields inside the modelSpecs facade
The set of keys an enforced model spec locks is the input contract of
the reasoning-override refusal rule, and deriving it in the route
middleware left the policy untestable from packages/api and free to
diverge from the facade that owns the rest of the preset application.
applyModelSpecPreset now returns the enforced fields beside the private
ones it applies, and the middleware consumes them.
* fix: restore the pre-override parameters when a resume degrades
Stripping a replayed reasoning override that no longer validates removed
only the metadata field, while the resume had already replayed the paused
turn's model parameters with the override applied into the endpoint
option, so the provider still received what validation had refused. The
strip now inverts the application from the replayed base: the recorded
pre-override value returns and the thinking flag the override forced on
comes off.
* fix: resolve the effective responses route for upload targets
The attach target returned the conversation or agent's saved
useResponsesApi flag, so an Azure chat auto-routed to the Responses API
by responsesApiRouting preflighted its uploads against the chat
completions route and lost provider-document uploads. The resolver now
applies resolveEffectiveUseResponsesApi with the endpoint, model,
web-search state and routing policy, the same resolution the during-run
upload target makes.
* fix: name the interrupt chord in the hint only while it still interrupts
The during-run hint advertised Alt+Enter Interrupt & send even when the
chord had been claimed: a submitMessage rebound to Alt+Enter, a chord
yielded to a global shortcut, or shortcuts disabled altogether each make
the keydown resolver return something else. The hint now reads the same
resolver verdict the during-run send button does and omits the chord it
no longer owns.
* test: assert the unseen count the scenario adds rather than a zero baseline
The tab title counts every unseen conversation, and a shared
verification run leaves legitimately-unseen conversations from earlier
scenarios behind, so the three title-count scenarios failed on other
scenarios' leftovers instead of the behavior they seed. Each now reads
its baseline and asserts the count it adds and clears.
* test: hold the unseen count to its direction on the mobile project
The away poll that lights a dot also discovers conversations the client
had not counted yet, so on the mobile project the title count can jump
past any baseline taken mid-test. Opening still has to clear this
conversation's contribution and an away reply still has to raise the
count, which is what the assertions now hold.
* test: enable the MCP Apps server through the redesigned palette
The MCP Apps integration spec still drove the old composer's MCP
Servers menu, which the redesign replaced with toggleable rows in the
attach-and-tools palette; the lane failed on the missing menu on every
run. The helper now opens the palette, waits for the server wave, and
flips the row's pressed state.
* fix: preserve numeric reasoning defaults on blur
* test: reopen persisted code cards before checking highlights
* test: follow the current file manager navigation
* fix: offer only reasoning options a request override can carry
* fix: keep the favorited copy when skills share a name
* fix: restore composer focus by the interaction that closed the palette
* test: expect a keyboard close to refocus the composer on every project
* chore: re-record the design-rule backlog on the rebased composer
* chore: drop the files table and label the rebase brought back
* test: tolerate other projects' unseen conversations when marking one unread
* test: open the file table header scenario from the account menu
* fix: let the stop shortcut reach a hidden stop control from a dialog
* fix: keep staged reasoning out of tool-limit recovery and MCP App turns
* fix: degrade a queued turn's stale reasoning override instead of dead-lettering it
* fix: refresh the attach target when web search changes the route
* fix: re-arm the run-end drain after moving a steer to the queue
* test: hover the MCP row before reaching its revealed Connect action
* test: pick the parts scenario's MCP server from the tools palette
* fix: cancel a recovered row's parked steer only after the composer takes it
A recovered row's only durable copy is the steer parked on the server, and
once its run has ended nothing can re-create it, so a composer that refused
the row after the cancel left the words in memory until the next reload.
The composer now takes the words first; a cancel that then fails holds the
row out of the run-end drain instead of sending it a second time.
* fix: name only the idle composer affordances that act
* fix: keep steer escalation off until the pane owns the rendered conversation
The branch tail that decides whether a run is paused is keyed by the pane's
conversation, which can still name the previous chat while a cached one
renders. Until the pane catches up, a pause on any cached branch now keeps
Escalate disabled instead of reading a missing or foreign tail as unpaused.
* fix: route the host's Enter and auto-send preferences into the composer's nested hooks
ChatForm takes enterToSend and autoSendText from its host, but useTextarea,
DuringRunSendButton and both speech engines still read the store atoms, so a
host whose values differ advertised one behavior while the handlers ran
another. Pass the host values through instead.
* fix: settle a browser dictation take only after its final transcript lands
stopListening resolves on the recognizer's end event, after the final result
reached useSpeechRecognition's reducer but before React committed it, so the
take settled while the composer still held the interim text. Settle from an
effect declared after the transcript effect, off a state update queued behind
the result.
* fix: drop a restored reasoning override the current setting no longer offers
The staged override was only checked for a matching parameter key, so a value
restored after a capability change on the same target stayed staged, blocked
live steering, and was dropped at send time. Validate it against the setting's
options and range.
* test: scan the attach and tools palette for accessibility
The a11y suite still opened the removed tools dropdown. Open the palette
instead, and add a scenario that runs Axe over the open palette.
* fix: make the palette's virtualized list the grid that owns its rows
The rows sat under a wrapper grid behind react-virtualized's scroller, which
always renders aria-label and aria-readonly and so could not be made
presentational. Axe reported both the disallowed attribute and the grid's
invalid child. The scroller is now the labelled grid, with a rowgroup holding
the rows.
* fix: state the palette grid's row total and each row's position
Only the rows in view are mounted, so without aria-rowcount and aria-rowindex
assistive technology reads the rendered window as the whole grid.
* fix: key an empty destination's pending steers to the route
An empty tree has no message naming its conversation, so the pending-steer
surface fell back to the chat context, which still names the source chat
during warm navigation, and its controls acted on the source run. Fall back
to the chat's messages key, which follows the route.
* fix: keep assistant uploads available when the file config fails
Assistants upload directly and never needed the destination policy, which is
why the replaced AttachFileChat rendered their uploader outside the policy
gate. Restore that exception so a failed or paused file config request does
not remove every attachment entry.
* 📡 fix: Explain Model Streams That Die Mid-Response
* 📡 fix: Bound Model Stream Failures to Response Bodies
---------
Co-authored-by: Lia <lia@librechat.ai>
* feat(client): Rebuild the chat list filter as a properties menu
The old menu listed every option at once. It now opens on three rows that
state where the list stands, Show, Sort and Filter, each with its current
value and a submenu holding the choices. Reading the list's state no longer
means reading the whole menu.
Filter holds the five facets and a search field that looks inside them, so
typing openai finds the endpoint rather than the category that contains it,
and typing a category name offers everything under it. Reset sits in the
header next to the title, and the active facet count rides on the trigger.
The facets are one derived atom, so the query key and the request are built
from the same description of the filter. Date cutoffs snap to local
midnight: taken from the current instant they would differ on every render
and refetch the list each time.
The panels also fade their content at the foot when a list runs past the
fold, and the chat list's loading label shimmers the way every other
in-flight label in the app does.
* feat(client): Show a project's chats under that project only
A chat assigned to a project was listed twice in one sidebar: once under
the project and once in Chats. The chats list now asks for the chats that
belong to no project, which the list query and its cache helpers already
knew how to answer.
Search and the archived view keep asking for everything. Projects is not
rendered while a search is on, and an archived chat has no project row to
appear under, so excluding them from either would leave a chat with no way
back to it.
* fix(client): Keep facet-filtered caches honest and the facets session-scoped
Cache writers judged list membership only by project, archive, tags and
sort, so an SSE update or a fork could insert a chat the server's filter
had excluded. The endpoint and date facets are decided from the row
itself, and the attachment and sharing facets mark their variants for a
refetch, since only the server can place a row in them.
The date cutoffs now derive from a day anchor a mounted consumer
advances at local midnight, so a tab left open stops serving
yesterday's today. Facets reset at the authentication boundary with the
rest of the session state, the bookmark query stops firing for roles
without BOOKMARKS:USE, an empty Chats section under existing projects
says so instead of claiming the account is empty, and the design-rule
violations this branch had suppressed in Conversations and
ChatFilterMenu are fixed and their entries pruned.
* test(client): Mock the projects query the chats section now reads
* test(e2e): Acceptance scenarios for the chat list properties menu
* fix(client): Call the account empty only once its projects have loaded
An empty Chats list read as "No chats yet" whenever the projects query
had not succeeded, so a slow or failed projects request claimed an
account was empty when its chats may all live under projects. The
account-level wording now needs a loaded, empty project list; otherwise
the section says only that nothing sits outside a project.
* test(e2e): Scenario for the Filter submenu under a role without bookmarks
* fix(client): Drop a cached chat from an endpoint filter it no longer matches
The endpoint facet only guarded insertion, so a chat already in an
endpoint-filtered list stayed there after switching provider until the
next refetch. List membership now includes the endpoint for existing
rows as well; a row with no endpoint stays until the server decides.
* fix(client): List every chat when the projects fail to load
Scoping Chats to unassigned chats hid every project chat behind the
Projects section, which shows nothing when its request fails. A failed
project list without cached data now leaves Chats unscoped, so those
chats stay reachable, and an empty unscoped list reads as an empty
account.
* test(e2e): Scenario for project chats when the projects fail to load
* fix(client): Count the bookmark filter once in the menu trigger
The trigger added one per selected bookmark while the Filter row counts
the whole bookmark group as one, so a screen reader heard two different
totals for the same state. Bookmarks now count once, like the endpoint
facet.
* fix(client): Stop the endpoint filter at the configured limit
The Endpoint facet let a user pick more endpoints than
conversationList.maxEndpointFilters allows, and the list request past it
answered 400, leaving Chats in its error state. The post-login startup
config now publishes the limit, resolved by the same helper the list
parser uses, and the facet disables further endpoints once the
selection reaches it, with a note saying why.
* test(e2e): Scenario for the endpoint filter limit
* fix(client): Hold the endpoint filter until its limit is known
The sidebar renders before the post-login startup config resolves, and
until then the Endpoint facet had no limit to enforce, so a selection
made in that window could pass conversationList.maxEndpointFilters.
Endpoints are now offered disabled until that config has loaded.
* fix(client): Keep a chosen bookmark removable and a failed config unlocked
A selected bookmark whose last chat dropped it fell out of both bookmark
lists, so only Reset could clear it; a selected tag now stays listed.
The Endpoint facet waits only while the startup config is loading, so a
config that failed no longer locks every endpoint without saying why.
* test(e2e): Scenario for a chosen bookmark whose count drops to zero
* fix(client): Keep a chosen endpoint removable when the deployment stops listing it
An endpoint that left the endpoints config, or a config that failed to
load, dropped out of both endpoint lists while it stayed selected, so it
kept narrowing the chats and could not be turned off. Both lists now come
from one helper that keeps the selection listed.
* docs(client): Punctuate the list membership comments without dashes
* fix: Publish the chat list filter limits the list route enforces
The startup config published maxEndpointFilters from the caller's merged
config while the list route reads the deployment's base config, so a
principal override could let the sidebar allow a selection the route
refused, or disable one it accepted; and the endpoint name limit was not
published at all, so a long custom endpoint name could be selected and
turn the list into a 400. Both limits now come from one loader the list
route also uses and are published together as conversationListLimits;
the Endpoint facet leaves out names past the length limit.
* fix(client): Keep a deleted bookmark removable from the filter
A selected bookmark that was deleted no longer came back from the tags
query, so neither bookmark list could show it and only Reset cleared it.
Both lists now come from one helper that keeps every selected bookmark.
* fix(client): Offer a retry when a project's chats fail to load
Chats lists only chats outside any project, so a project whose own chat
request failed hid its chats behind the empty state. The project now says
its chats could not load and offers a retry.
* test(e2e): Scenario for a project whose chats fail to load
* 📟 fix: Present Background Task Checks as a Distinct Activity
* 🧹 fix: Use a Flat Category Icon Branch
* 📟 fix: Label Live Background Polls as Checks
---------
Co-authored-by: Lia <lia@librechat.ai>
* 🏷️ fix: Title an Open Live Card Only by a Generated Label
An open card still repeated its newest row: a running call's intent was
counted as a label and put on the header above the very row that shows it,
and with no label at all the header said "Thinking…", which the grouped
thought row says too.
Open, the header now takes only a generated label from the span, the
newest batch label or a thought's label, and otherwise a generic running
line that no row uses. A call's intent, a reasoning sentence, commentary
and the generic thinking line are all row text and are walked past.
* 🧪 test: Pin the Open Card's Title to the Running Line While a Thought Streams
* feat: expose the messages key and the submitted response on the chat contract
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: keep the submitted response in the chat host instead of reading the submission atom
The contract's initialResponse is now the response ask submitted, held beside
the host's own state and exposed only while the turn is in flight, so a settled
turn and a run restored after a reload report none, and the submission atom
gains no reader.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: take the submission ask builds as a value and document the empty messages key
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: scope the submitted response to its chat and promise seeded positions, not identity
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* docs: note that a new chat's submitted response ends at its saved id
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: end the submitted response with its turn and note that the stream can continue the last seeded part
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* refactor: drop the submitted response from the chat contract
A run can be restored, replaced or torn down outside ask, so a copy of the
submitted response kept in the host describes the wrong run on those paths. The
contract keeps messagesKey only.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* feat: add an AI SDK-shaped useChat facade over the chat contract
useChat reads ChatContext and presents it as @ai-sdk/react@4.0.117 does:
UI-shaped messages, a submitted/streaming/ready/error status derived
from the in-flight flag and the latest message, and sendMessage,
regenerate, stop and setMessages forwarding to the contract. It holds
no state and changes no existing consumer.
* feat: apply the client tool outcome rules in the useChat facade
The parts mapping reads only the markers stored on a tool call; the facade supplies getToolMeta as its resolveToolFailure, so memory failure prose and background task status attachments reach the UI parts view the same way they reach the tool cards.
* fix: make the useChat facade reactive, placeholder-aware and cached per message
messages re-reads getMessages whenever the message query cache is written, through useSyncExternalStore, so setMessages and other cache writes show without an unrelated context update. Views are cached per message reference, so a streamed chunk remaps only the messages it replaced. status ignores placeholder parts through isEmptyContentPart, the error reads an Assistants error part's text, and a message inserted through setMessages joins the active conversation under the message before it.
* fix: remap streamed responses and scope stepless tool calls in the facade
The content handler reuses a response object across frames and replaces its content, so views are cached per message and per snapshot of the fields a view reads rather than by the object alone. The tool outcome resolver passes getToolMeta the step ids a repeated provider id already owns, so a call without a step is not failed by another step's attachment.
* fix: follow in-place stream frames and tolerate empty tool call slots in the facade
The content handler replaces a response's content on the same object, so
structural sharing keeps the cached message array and the external store saw
no change. The snapshot is now keyed by the message cache write count.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: catch unobserved cache writes, keep inserts in the active chat and report the error part
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: key the facade snapshot on the active query's own update count
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: use the filters form of findAll for the facade's messages scan
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* perf: look up the facade's holder queries once per cached array
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: read the chat id, first delta and active branch from the chat contract
The facade subscribes to the one message query the contract names instead of
matching cached arrays, reports that query's conversation as the chat id, counts
only parts the submitted response was not seeded with as streamed output, and
parents an appended message to the active branch. A memory call's failure prose
is its error text now that the resolver's reason becomes the part's errorText.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: keep inserted messages in the active chat, unassigned in a new one
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: keep stored messages in their conversation and count generated files as streamed
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: read seeded parts by position and report failed user tails as errors
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* fix: count a continued last seeded part as streamed and remap views on metadata changes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
* refactor: read streaming status from content alone now the contract has no submitted response
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01GQdLgSaTheiBFHYnUDgobw
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* 📏 fix: Keep the Context Gauge's Used Count Above Its Own Breakdown
* 📏 fix: Include summary tokens in context gauge breakdown floor
---------
Co-authored-by: Lia <lia@librechat.ai>
* 🪁 fix: Prioritize Running Chats in Sidebar History
* 🪁 perf: Reuse Date Groups Across Active-Job Polls
* 🪁 perf: Avoid Allocating a Set on Every Sidebar Render
---------
Co-authored-by: Lia <lia@librechat.ai>
* perf: Collapse Settled Reasoning Text in AnimatedText
AnimatedText re-split and re-classified the entire accumulated reasoning
string on every streamed token and rendered one element per word, making
a long reasoning stream quadratic. Classification is now append-only: the
leading run of words that finished fading is committed as a single text
node and never re-split, and only the unsettled tail is classified.
* perf: Probe the Settled Reasoning Prefix Instead of Rescanning It
The append check compared the whole settled prefix on every token, which
kept the stream quadratic in character comparisons. It now probes the
prefix head and the region at its boundary, and the settled region is
rendered from the current text so a rewrite can never show stale
characters. Adds a long reasoning stream scenario to the mock harness.
* perf: Settle Spaceless Reasoning Runs Before Their Final Segment
A CJK or Thai reasoning stream never reaches whitespace, so the settled
prefix could not advance and the whole run was re-segmented per token.
Inside a spaceless last word everything before its final segment is now
treated as stable; re-segmenting there only changes fade granularity.
* perf: Bound the Unsettled Reasoning Tail by Length
Settling at part boundaries left any run that never reaches whitespace
(unspaced Hangul, a spaceless script without Intl.Segmenter, long URLs)
in the tail, re-classified on every token. The unsettled region is now
capped at 64 characters regardless of segmentation, replacing the
spaceless-specific rule; only non-animated text is ever cut, so a cut
inside a part is invisible. Tests measure the classified length per
render instead of the DOM shape.
* fix: Keep the Settled Reasoning Cut on a Glyph Boundary
The length cap could cut inside a surrogate pair or a combining or ZWJ
sequence, splitting one glyph across two text nodes. The cut now steps
back past continuation code units.
* fix: Cut the Settled Reasoning Prefix at Grapheme Boundaries
Code-unit rules missed emoji modifiers and regional-indicator flags. The
cut now takes the last grapheme boundary from Intl.Segmenter over the
unsettled tail, which starts at a committed cluster boundary and is
bounded by the cap; the code-unit rule remains as the fallback.
* fix: Skip the Reasoning Tail Cap Without a Grapheme Segmenter
The code-unit fallback could still split emoji modifiers and flags. With
no Intl.Segmenter there is no safe cut, so the cap is not applied and the
tail stops at the last word, as it did before.
---------
Co-authored-by: Lia <lia@librechat.ai>
* 💭 feat: Bring the Streaming Thought Peek Back Under a Collapsed Live Card
#14546 previewed streaming reasoning as its trailing sentences in a short
fading window under the thought's header. The live fold (#16118) put that
thought inside a collapsed card, which unmounts its rows, so the peek went
with them and one throttled sentence on the header stood for a paragraph
of live reasoning.
A collapsed live card whose tail is a streaming thought now renders the
same peek under its header, outside the fold, in the cursor's place. It
gives way to the rows when the card opens and to the cursor when a call,
not a thought, is at the tail.
* 🧪 test: Watch Live Reasoning Stream Through the Real App
A mock-lane scenario streams three sentences of reasoning at a readable
pace; the spec samples the collapsed card's header and asserts every line
was the generic one or a whole sentence, sees the thought peek under the
card, opens the card mid-stream and checks the header became a title, then
checks the settled thought's header sits at the shared row scale.
The peek strips the thought's tags itself: handed the stream straight from
a live card, it had shown a literal `<think>` before the first words.
* 🎨 style: Sort Activity Phase Imports After Branch Rewrite
---------
Co-authored-by: Lia <lia@librechat.ai>
An open card whose label has not landed yet kept the live line, so its
sticky header quoted the sentence the reader could already see streaming in
the rows beneath it. Open, the header now names the span by its newest
label: a thought's generated label or the generic thinking line, a batch
label, a call's own line. Reasoning sentences and prose commentary are for
the collapsed header, which stands in for rows the reader cannot see.
The live header previewed the sentence the model was still writing, word
by word, so it read as fragments ("This is a simple greeting, not") that
changed before they made sense and were replaced by the next fragment. It
now shows the last sentence that ENDED, holds it for at least a second, and
keeps its generic line until the first sentence finishes. A Latin mark
closes a sentence only before whitespace or the end of the text, so
decimals and version numbers stay whole; CJK marks need no space.
The measurement that let short sentences stream freely until they filled
the row (a ResizeObserver and a per-paint scroll-width read) goes with it:
the line no longer changes per word, so there is nothing to meter.