* fix: Follow the Interface Direction in Radix, the Composer and First Paint
Wrap the app in a Radix DirectionProvider driven by i18n so Tabs, Select and
menus stop rendering LTR subtrees in right-to-left locales, and swipe toasts
toward the inline end.
Chat direction becomes auto, ltr or rtl (auto by default; stored LTR/RTL
values are kept). The composer applies a forced direction as a dir attribute
instead of reversing rows, which double-flipped under an RTL document, and the
text fields default to dir=auto. The unused Jotai chat direction atom that
pinned every TextareaAutosize to ltr is removed.
index.html sets dir before first paint, and the language setting no longer
races i18n over document.lang.
* fix: Keep the Empty Composer Placeholder on the Interface's Reading Edge
A tool call that is the only one in its group, inside a phase of several, renders its panel bare: open, with no row of its own. The full-content request only fired when the reader expanded the card, so a bare panel stayed on the server preview, showing truncated output and keeping its copy buttons disabled. Bash, generic and the other useToolCallState cards now request the stored part whenever they render bare.
* refactor: Draw Provider Icons as Theme-Following Brand Marks
Replace the raster and oversized provider art with single colour SVG marks
that inherit currentColor. Marks come from simple-icons (CC0) or are cut
from the logo each brand publishes, as noted per entry in marks.ts.
Lemonade, MLX, ShuttleAI and Unify keep their raster art for now.
The Cohere landing padding only compensated for the old PNG, so it goes.
* fix: Use the Groq Wordmark G Instead of the Site Favicon Bolt
* ⌨️ feat: Keyboard submit for ask_user_question answer box
The interactive ask_user_question card's answer textarea had no key
handling, so the answer could only be submitted by clicking. Route its
keydown through resolveComposerKeyDown so it follows the composer's
Enter-to-send preference, a rebound submit chord, and IME composition.
The run is paused rather than generating, so it resolves as idle and
Ctrl/Cmd+Enter submits instead of reaching the during-run actions.
* ⌨️ feat: Keyboard focus and confirm for batched ask_user_question
- Focus the first question's controls when a batch arrives, only when
focus is idle (body, or the empty composer that sent the turn), and
focus into the next step on advance instead of the fieldset: the picked
or first option, else the answer field.
- Enter on the already-picked single-select option confirms the step;
Enter on an unpicked option stays its click, so a pick never submits
by itself (WCAG 3.2.2). Multi-select rows only toggle.
- Ctrl/Cmd+Enter confirm from the answer field like Enter; Shift+Enter
never does.
* 🐛 fix: Keep Held Enter and Remounts From Picking Batched Answers
Review findings on the batched ask_user_question card, and what this
commit does about each.
1. Held or double-tapped Enter could pick answers and submit the batch.
After Enter confirmed a step, focus moved to the next step's first
option. A keyboard Enter on a focused button is its click, so the
key repeat of a held Enter (or a quick double-tap) picked that option
and advanced again. On the last step the next repeat landed on the
now-pressed option, confirmed it, and submitted the whole batch with
answers the user never chose.
Fix: after a step change, focus goes to the step's picked option,
else its answer field. An unpicked option is only a focus target when
a batch first arrives. Repeats now reach a text field, which can page
on but never picks an option or submits on the user's behalf.
2. Focus-on-arrival ran on every mount, not only on arrival. Its effect
has empty deps, so moving the batch between the composer popover and
the chat card, or remounting a message that still holds an
unanswered batch (reopening a conversation, a Share view), pulled
focus onto an option of that batch. The next Enter or Space would
then answer it.
Fix: `AskUserQuestions` takes a `focusOnArrival` prop, set only by
the surface showing the live pause (the popover, or the chat card
while it is live and visible), so other mounts never take focus.
Focus is placed at most once per batch, tracked by action id, so a
move between surfaces is not treated as a new arrival.
Tests: a held Enter from an answer field picks nothing and never
submits; a surface that isn't the live pause leaves focus alone; a
remount of the same batch doesn't take focus again. All three fail
without this change. The Chat/Messages/Content, Chat/Input, Chat/ask
and hooks/Input suites pass, apart from TokenUsage/index.spec.tsx,
which fails the same way without this change. tsc, ESLint and Prettier
are clean on the changed files.
* feat: Escape to Skip Ask Cards and Move the Popover to Chat
Escape in the chat card skips the question or batch, but only before
anything is entered, so a stray Escape cannot discard an answer and
resume the run. Escape with focus inside the composer popover moves the
question to the chat, as it already did from the composer, and returns
focus to the composer once the popover closes. Also drops the em dashes
from the batched card's comments.
* fix: Read Enter-to-Send From the Chat Surface Host in the Ask Card
* test: Cover Ask Batch Keyboard Control in E2E Scenarios
* fix: Keep Held and Modified Enter From Submitting Ask Answers
Ignore auto-repeated Enter before confirming a picked option, so a held
Enter cannot both pick and submit the last question. In the answer
field, Shift, Alt and repeated Enter now prevent the enclosing chat
form's implicit submit instead of returning early. Escape that cancels
an IME composition no longer moves the popover to the chat, and the
arrival-focus flag lives in the batch's Jotai form state.
* fix: Honor a Submit Shortcut Rebound Off Enter in Ask Answer Fields
The window shortcut handler yields submitMessage in every editing field
except the composer, so a chord such as Ctrl+J did nothing in the ask
answer fields. Both the single-question answer box and the batched
answer field now act on it, through a shared isReboundSubmitChord.
---------
Co-authored-by: elie605 <ebtien@gmail.com>
* fix: Hold the Locked Composer to One Line Under a Batch Question
A batch ask_user_question card sits above the composer with its own 70vh cap, while a multi-line draft kept the disabled textarea at 45 to 55vh. Together they could exceed the viewport and push the prompt and the move-to-chat control off screen. The composer now takes its collapsed one-line height while the batch owns it, and the collapse toggle stands down since it cannot expand a locked composer.
* test: Cover the Batch Card Above a Long Composer Draft
* fix: Wait for Chat History Before URL Auto-Submit
Existing chats refuse a send until their message history has loaded and
settled, so a ?q=...&submit=true link on a slow-loading chat consumed its
one auto-submit attempt on a refusal and left the prompt unsent. Hold the
submission in its preparing state and retry when the history query
resolves, giving up only if the history request fails.
* test: Cover URL Auto-Submit on a Slow-Loading Chat
* fix: Retry URL Auto-Submit After the History Render
A query cache listener fired while the history was being written, before
the composer re-rendered, so the send derived no latest message and forked
from the root. Observe the messages query instead and retry from an effect,
which runs after the render that carries the loaded history.
* fix: Pair English Plural Keys as _one/_other for Locize
Locize groups keys by plural suffix and generates every form a language
needs. Families that paired a bare base key with _one surfaced empty-source
_other rows, and literal keys ending in _other or _one spawned unrelated
plural siblings. Rename the bases to _other, rename the colliding literal
keys, and guard the English file against a regression.
* test: Cover Plural Version Count and Other Issue Feedback Tag
* fix: Announce a Single Running or Finished Chat in the Singular
* fix: Resolve Plural Forms From Base Keys So Locale Plural Rules Apply
Selecting an explicit _other key bypassed i18next's plural resolution, so a
single running chat was announced in the plural and locale forms such as
_zero and _many were never chosen. Call sites pass the base key with a
numeric count again, and TranslationKeys accepts plural base keys.
* fix: Let i18next Pick Duration Plural Forms From a Numeric Count
Duration and elapsed announcements selected an explicit _one or _other key
and passed the formatted number as count, so locale forms such as _few and
_many were unreachable. They now pass the base key with a numeric count and
interpolate the locale-formatted number as value.
* fix: Route Subagent Files Delivered After a Mid-Run Reload
* test: Stub Stored Part Queries in the ContentParts Integration Suite
* fix: Recover Stored Part Lookups That Fail Without a Reader
* fix: Resolve Late Subagent Files With One Previewed Message Read
* fix: Read Late Subagent Hints Again Until the File Is Saved
* fix: Count Failed Reads Toward the Late Hint Poll Bound
* fix: Let Unresolved Late Hint Reads Recover on Focus and Reconnect
* fix: Resolve Late Subagent File Owners From the Stored Transcripts
* fix: Ask for Late Subagent File Owners in Bounded Batches
* fix: Render Files Saved by Tools Inside a Subagent Run
* test: Cover Subagent Nested Tool Images After Reload
* fix: Hoist Subagent Child-Run Files With Grouped Tool Calls
* fix: Route Live Subagent Child Files Through the Part Attachment Map
* fix: Keep Subagent Child File Routing Through Previews and Ownership Rules
* test: Match Subagent Image Names Exactly
* fix: Key Subagent Child Files by Their Owning Part
* fix: Apply Attachment Scope Rules to Subagent Child Calls
* fix: Carry Child Call Scope Through Previews and Live Progress
* fix: Retain Live Subagent Progress Readers and Seen Child Calls
* fix: Keep Live Progress Keys Retained Across Streamed Deltas
* fix: Assign Subagent Child Files Once Per Message
* fix: Route Late and Id-less Subagent Child Files
* fix: Keep Seen Subagent Child Calls With Their Invocation
* fix: Use Streamed Index and Call-Level Owner for Subagent Child Files
* fix: Reserve Scoped Child Files From Unscoped Top-Level Calls
* fix: Record Subagent Child Calls When Progress Is Written
* fix: Keep Agent Scope on Remembered Subagent Child Calls
* refactor: Share One Subagent Child Call Collector
* fix: Derive Live Signature From Remembered Child Calls
* fix: Keep Child Calls a Bounded Batch Evicts
* fix: Bound Remembered Subagent Child Calls and Include Replay Segments
* fix: Refresh Owners of Shown Files in the Bounded Call History
* fix: Keep Remembered Call Order Stable Across Eviction Replays
* fix: Scope Preview Hints to Attachments and Walk Nested Calls Iteratively
* fix: Bound Wildcard Preview Hints and Own Live Calls by Their Run
* fix: Own Stored Subagent Child Calls by Their Persisted Identity
happy-dom 20 holds each MutationObserver callback only through a WeakRef, so a garbage collection between two mutations silently unregisters the hook's observer and the removal is never delivered. Under load that GC lands mid-test and the spec reads data-active="true" after the attribute is gone. jsdom keeps registered observers alive the way browsers do.
* feat: open durable subagents from background tasks
* fix: isolate subagent focus and discovery from navigation
* fix: retain subagent task identity through navigation
* test: verify task opening without changing the rendered popup
* fix: preserve retained task activity across navigation
* fix: import the core progress key after dev integration
* fix: refresh subagent panel return focus on explicit open
* fix: prefer the subagent panel opener when restoring focus
* test: retain the actual reply stamp across permission grant
* test: scope permission-grant notifications to the target reply
* test: observe the pending reply before granting notification permission
---------
Co-authored-by: Lia <lia@librechat.ai>
* refactor: Drop Queue Reset Helper From The Client Queue Module
* refactor: Drop Unused Shortcut Getters And Subagent Batch Wrappers
* refactor: Move HITL Approval Payload Builder To Test Fixtures And Drop Dead Helpers
* 🔐 fix: Skip Reasoning Labels for Encrypted Reasoning
OpenAI models on Bedrock (and Claude's redacted thinking on Bedrock) return
reasoning as encrypted `redactedContent` bytes. The current agents SDK
base64-encodes those bytes into the reasoning stream, so they arrive as THINK
text: the reasoning-label wiring summarized the blob ("Inspecting an encoded
data snapshot"), activity and phase labels quoted it as reasoning evidence, and
the UI rendered it as thinking.
Add a shared `stripOpaqueReasoning` / `isOpaqueReasoning` predicate in
data-provider that recognizes long unbroken base64/base64url runs (Bedrock
redactedContent, Anthropic redacted_thinking data, OpenAI encrypted_content,
Gemini thought signatures) and keep them out of every label path:
- reasoning labels drop encoded deltas, ignore persisted payloads on resume,
and re-check the accumulated text before a billed label call
- activity-label and phase-summary evidence carries readable excerpts only;
phase anchoring still uses the raw excerpt
- the THINK renderer, grouped reasoning rows and the live header show a
neutral "Reasoning (encrypted)" marker instead of the payload, and ignore
labels previously generated from it
Readable reasoning on every provider, including readable text mixed with an
encoded payload in the same step, is still labeled.
* fix: Classify streamed and persisted payloads by readable content
- isOpaqueReasoning / stripOpaqueReasoning ignore the persisted `<think>`
wrapper, so stored encrypted reasoning renders as the encrypted marker and
never reaches label evidence as tags-only "text"
- add advanceOpaqueReasoning, a streaming cursor that holds a trailing
payload-alphabet run until it settles as a word or grows into a payload,
so a payload split across sub-threshold deltas is dropped as it streams
- reasoning labels count only settled readable characters plus a word-sized
pending tail, so ciphertext fragments never advance the minChars or
updateChars gates or trigger extra billed label calls
* fix: Show readable reasoning around payloads and settle words before counting
- the reasoning disclosures, grouped rows, live header and conversation
export render and copy the readable text of a THINK part with encoded
payloads removed; the encrypted marker stays for parts with nothing readable
- reasoning labels count and send only settled readable text: an undecided
trailing run never advances minChars/updateChars or reaches the prompt,
whatever the configured thresholds, and a short final word settles when the
step closes
* fix: Classify reasoning before every bound and across resume
- once a run is classified as a payload, it continues through the next
non-payload character even when that arrives in the same delta as readable
text, so no tail fragment of the payload is released as readable
- add sliceReasoningTail: the activity-phase rolling reasoning window (and its
snapshot/restore bounds) never starts inside an unbroken run, so a payload cut
by the window cannot leave a fragment too short to classify; the window stays
a substring of the raw THINK text, so phase anchoring is unchanged
- a resumed reasoning step replays its persisted text through the stream
filter and keeps the resulting cursor, so a payload prefix that was still
undecided at the snapshot is dropped once the continuation classifies it
* fix: Classify phase evidence while it streams and mask payloads while pending
- activity phases classify each reasoning delta with the stream cursor and keep
a readable rolling window beside the raw anchor window; tracked activities
carry readableExcerpts as prompt evidence, so a window cut into a payload can
never forward a fragment (older snapshots fall back to stripping); the
cursor-free sliceReasoningTail is removed
- add readableReasoning / flushOpaqueReasoning: while reasoning streams, the
disclosures and live header hold back a trailing run that could still grow
into a payload instead of showing its first characters
- the message editor offers no field for payload-only reasoning and shows only
the readable text of mixed reasoning; an unchanged part is never written, so
its stored payload stays intact
* fix: Replay legacy phase snapshots through the reasoning stream filter
A pending-reasoning lane restored from a snapshot without a readable window
or cursor is replayed through advanceOpaqueReasoning instead of stripped, so a
payload prefix that was still undecided at the snapshot stays in the cursor
and is dropped once the continuation classifies it.
* fix: Drop the cut leading run of a full legacy phase window
A legacy pending-reasoning window that filled the 600-char bound may have been
cut inside a payload, so its leading payload-alphabet run is dropped before the
replay; a shorter window is the whole lane and is replayed as is.
* fix: Keep payload-bearing reasoning read-only and drop payload-derived titles on resume
- reasoning that carries an encoded payload is not offered in the message
editor: the edit endpoint replaces the whole value and the payload cannot be
positioned in an edited draft, so editing would silently delete it
- a resumed reasoning step whose persisted text has no readable reasoning
drops its existing title (which could only describe the payload) instead of
carrying it forward where a short readable continuation would revive it
* fix: Reset a resumed payload-derived title on the client, only for payload-only steps
- a resumed step drops its title only when its persisted reasoning is entirely
an encoded payload, so a short readable token (still undecided for the
stream cursor) keeps a valid title
- dropping it emits a same-step on_reasoning_label reset, which the client
already applies, so a resumed client clears the snapshot title too
* fix: release OpenID delivery lease when the client disconnected before delivery
A refresh whose client disconnects before the response is sent claims the
delivery lease and then arms finish/close listeners on a response that has
already closed. Neither event fires again, so the lease is never released and
every later refresh for the same generation polls until the 30-second lease
expires.
Arm the listeners only while the response is open; otherwise release the lease
directly after the send.
* fix: defer the silent refresh until a prerendered page is shown
A prerendered page runs the startup silent refresh before the user sees it.
When the browser discards the prerender, the refresh is aborted mid-flight
after the server has already rotated the refresh token, and the page the user
actually opens has to recover from a session it never received.
Wait for prerenderingchange before refreshing so a prerender has no effect on
the session.
* refactor: move OpenID response delivery into packages/api