* fix: Bound Generation Status Reads So a Silent Request Cannot Strand the Pane
* test: Cover Status Reads That Never Answer Across Terminal, Error and Foreground Paths
* test: Match Scenario Replies Exactly
* fix: Keep the Provider Message on Classified Rate Limit Errors
* fix: Honor maxProviderErrorChars for Classified Generation Errors
* test: Cover Classified Gateway Errors in the Mock Harness
* test: Cover a Zero Provider Detail Limit on Generation Errors
* refactor: Resolve Classified Provider Errors in the TypeScript Backend
* fix: Redact Provider Details Before Truncating Them
* fix: Wrap Label and Prose Text Between Words Instead of Mid-Word
* test: Cover Word Wrapping in the MCP Trust Disclosure and Bookmark Delete Dialog
* test: Reopen the MCP dialog per width and treat hyphens as break points
* test: Open the bookmark delete confirmation from the keyboard
* fix: Let the MCP Dialog Segmented Controls Wrap on Narrow Screens
* perf: Prefetch Conversation Messages From the Chat Route
The messages request started only when ChatView mounted, after startup config, roles, endpoints, models and the conversation record resolved, although it needs only the route id. ChatRoute now starts it as soon as the id is known for a conversation the cache has never held, sharing the view's query function, and the view's first mount uses that payload instead of fetching it again.
* fix: Keep the Messages Prefetch Mark for the Observer That Uses It
Child observers of the same conversation mount first and cleared the mark before the conversation view subscribed, so the view loaded the messages a second time.
* fix: Show a Failed Messages Prefetch Instead of Requesting It Again
A prefetch that failed left no data, so the view loaded the messages again on mount, and a link to a missing conversation sent two requests. The view now shows a fresh prefetch failure, and the prefetch retries like the view's own load except on a 4xx.
* test: Cover Route-Level Messages Loading on Cold Load, Sidebar Switch and Missing Conversation
* fix: Retry Rate-Limited and Timed-Out Messages Prefetches
* fix: Prove Prefetched History by Reply Identity and Honor Invalidation
A URL load's prefetch starts before the conversation record is cached, so it captured no reply stamp and the seen check loaded an unread conversation's history a second time. Once the record lands, the prefetched history is proven when it holds the exact reply the record names. The mount guards also yield to an invalidation, and the failed-prefetch guard is the default for every observer of the conversation's messages.
* fix: Match Vendor Chunk Rules Against the Module Path, Not the Checkout
The vendor chunk rules tested the absolute module path, so a repository under a directory whose name contains a rule's text (fetch, ky, zod) sent every dependency to that chunk and the PWA precache limit failed the build. The rules now see the path from the first node_modules segment on.
* test: Cover Vendor Chunk Rules Under a Hostile Checkout Path
* fix: Require a Click to Submit Email Verification Links
Mail security scanners (Microsoft Defender Safe Links, Sophos, Hornetsecurity)
open links in incoming mail and run the page's scripts, so the verify page
submitted the single-use token on load and the user's own click then failed.
The page now waits for an explicit confirm button for both sign-up
verification and email-change confirmation.
* test: Cover verification links opened before the user clicks
* test: Isolate the verify click scenario from the submission limiter
* test: Confirm the sign-up link in the email e2e setup
* test: Exercise the shared Button in the verify page spec
* fix: Skip Attachments Deleted by Storage Retention Instead of Failing the Turn
A file that storage retention removed from an earlier message made every
later turn of the conversation fail, because the encoders rethrew the
storage miss and the user has no way to remove an attachment from history.
History replay now passes a handler to the encoders, which leave missing
files out of the request and report them; the message's file context tells
the model which files were deleted and were not read. Current-turn
attachments keep the strict error, since the user can remove and re-upload
them. Code Interpreter priming names files missing from storage in the
tool context and only requires resource recovery for other failures.
* refactor: Keep Code Reupload Failure Policy in packages/api
Tally priming reupload failures in createCodeReuploadFailures so the
decision that a file deleted from storage is named instead of requiring
resource recovery lives in TypeScript, and add mock e2e scenarios for a
follow-up after an attachment is deleted from storage.
* fix: Classify Code Files as Deleted Only at the Storage Read
A 404 from the Code API upload was counted as a source file deleted from
storage, which skipped resource recovery. openCodeFileSource marks a miss
where the stored object is read, and only that error counts as deleted.
Move the history replay of attachments with tolerated misses into
replayMessageAttachments so BaseClient only calls it.
* fix: Name Local Code Files Whose Stream Reports a Missing Object
Local storage returns a read stream that raises ENOENT only once read, so
openCodeFileSource now holds a stream until its first chunk or error and
names a miss there. The missing-files context assignment moves into
setMissingCodeFilesContext so both tool loaders only call it.
* fix: Close a Held Code File Stream When Priming Is Cancelled
openCodeFileSource waits for a source stream's first chunk; when that
wait fails, as on cancellation, the caller never receives the stream, so
destroy it before rethrowing instead of leaking its file descriptor.
* fix: Treat Empty Sandpack Bundler URLs as Unset
An empty SANDPACK_BUNDLER_URL or SANDPACK_STATIC_BUNDLER_URL (common with
${VAR:-} passthrough in Docker Compose) was handed to Sandpack as the
bundler URL, leaving artifact panels blank. buildSandpackOptions now falls
back to the default bundler for an empty value, and the shared-link startup
payload omits empty bundler URLs.
Related to #17790
* test: Cover Empty Sandpack Bundler URLs in Artifact Previews
* fix: Poll the Flux Cluster polling_url Returned by the Global Endpoint
* fix: Only Forward the Flux Key to Same-Origin or BFL Polling URLs
* test: Cover Flux polling_url Routing in the Mock Harness
* fix: Match mcp:<server>:<tool> Tool Approval Patterns Against MCP Tool Keys
* fix: Normalize Globbed MCP Server Names in Tool Approval Patterns
* fix: Keep Colons Inside MCP Server Names in Tool Approval Patterns
* test: Submit the Scoped Approval and Give the Control Agent remember_fact
* fix: Match mcp:<server>:<tool> Rules Per Segment Instead of Rewriting Them
* fix: Match Glob-Free mcp: Approval Rules Against Their Exact Key
* fix: Never Grant Through a Server Glob That Normalization Widened
* fix: Never Grant a Globbed mcp: Rule on an Ambiguous Tool Key
* fix: Mirror Both Server Name Normalization Branches in mcp: Server Globs
* fix: Treat Boundary Underscores in mcp: Server Globs as Optional
* fix: Refresh Copied Client Bundle Files After Build
fs.cp keeps the source mtime on Windows, so dist/tokens.css and the other
copied theme files never looked newer than packages/client inputs written
later by a checkout, and the pre-commit freshness check failed after every
rebuild. Stamp the copied outputs in tsdown's build:done hook.
Related to #17731
* test: Cover copied client bundle freshness where cp keeps mtimes
* fix: Accept Content-Array Outputs in Responses function_call_output
The Open Responses spec allows function_call_output.output to be a string or
an array of input_text, input_image and input_file parts, but request
validation refused anything that was not a string. Validate the array parts
and convert them with the same part mapping the message path uses, so the
replayed tool result keeps its text and images in order.
* fix: Inspect Tool-Output Media Against File Policies
Image URLs and file names in a function_call_output content array reached the
model without the file-policy fragments the same parts get in a message.
Collect those fragments for both item types in packages/api and call that
from the Responses controller.
* test: Cover Content-Array Tool Outputs in the Responses API
* fix: Remember Tool Approvals for MCP Servers Using Env References
* fix: Resolve Approval Env Only for Tiers That Expand It, Add E2E Scenarios
* fix: Reject User Placeholders in Resolved Env Values and Take the Env Source as a Parameter
* test: Expect the Resume Reply After Approving the Env MCP Tool
* fix: Check Fully Expanded Config Values for User Placeholders
* fix: Resolve Approval Env Into a Prototype-Free Map From Own Properties
* fix: Count Memory Tokens After the o200k Encoding Loads
Memory create and update routes, the by-id handler, and the memory agent's
set_memory tool persisted Tokenizer.getTokenCount, which returns the UTF-8
byte length for long text while o200k_base is still loading. Await
countTokens instead so the stored tokenCount is the real count.
Related to #17842
* test: Cover Memory Token Counts Saved Before the Tokenizer Loads
* test: Mock countTokens in Memory Tool and Route Specs
* fix: Key Sidebar Chat Rows by Conversation Id
Virtualized chat rows were keyed by row index, so a row's local state (an
open delete dialog, a rename field) stayed in its slot while another chat
moved into it, and confirming deleted the wrong chat.
* test: Cover Sidebar Row State When Rows Move
* test: Scope Sidebar Row Scenario Clicks to the Open Menu
* test: Release the Held Archive Reply Before Removing Its Route
* fix: Stop Sidebar Toggle Buttons Forwarding the Click Event as afterSlide
The desktop toggle and the mobile close button passed their onClick handlers
straight through to handleCollapse(afterSlide), so React's click event reached
setSidebarOpen as the afterSlide callback and threw when invoked (directly, or
via requestAnimationFrame on the animated mobile drawer).
Related to #17861
* test: Cover Closing the Sidebar Without a Runtime Error
* fix: Report Zeroed Summary in SharedLink Permissions Dry Run With No Links
The dry run returned early without a summary when the sharedlinks collection was empty, so the CLI crashed reading result.summary.totalLinks. Dry run now falls through to the summary counts; live runs keep the early no-op return.
Related to #17887
* test: Cover SharedLink Migration CLI on Empty and Legacy Collections
* fix: Keep the Projects Section Expanded by Default
* fix: Label the Streaming Dot With a Tooltip and an Accessible Name
* fix: Move the Welcome Screen Footer Directly Under the Composer
* fix: Drop the Agent Name From Responses When the Selector Already Shows It
* fix: Align the Sidebar Section Headings With the Date Labels
* feat: Make the Welcome Footer Placement and Response Author Configurable
* fix: Hide the Memories Panel on the Agents Route
* test: Cover the Launch QA Chat Chrome in E2E Scenarios
* test: Assert the Section Heading Padding Merge Without Mocking App Internals
* fix: Read the Response Author Setting Through the Messages Host
* test: Give the Consumer Spec's Messages Host the Author Setting
* refactor: Move the Queued Follow-Up Row Into @librechat/chat
The package owns which queued turn is drawn, when it follows the reader to the end of the thread, and where the composer's queued actions attach. The app keeps who sent it, their icon and how the text reads, passed in as props.
* test: Tag the Queued Follow-Up Row Scenario
The spec already drives the queued turn through its reveal and handoff in the browser; the tag lets the gate record it as evidence for the row the package now draws.
* chore: Retest Against Current Dev
* test: Keep Jotai's Real Exports in the Search Route Spec
PendingTurn now ships from the package components entry, which loads the steer rows and the pane submission atom at import. The spec replaced the whole jotai module with useAtomValue alone, so atom() was undefined and the suite failed to load.
* refactor: Move Conversation Starters Into @librechat/chat With a Starter Button Variant
The welcome screen's suggested first messages render through a package component built on a new starter Button variant instead of a raw button. The app keeps resolving which starters apply (agent, spec, assistant) and sends the chosen text.
* test: Pin the Starter Variant Under the ClickHouse Reference Theme
A starter takes its fill and ink from the theme roles in light and dark, so a theme that repaints the surface repaints the chip.
* fix: Cap the Starter Width at the Call Site, Not in the Button Variant
The Button primitive spec rejects fixed sizes in the variant recipe, so the 16rem cap moves to the package component that lays the chips out.
* fix: Keep the Starter Stagger Ahead of the Unlayered Fade Rule
The unlayered fade-in rule sets the animation shorthand, which resets the delay and fill mode a layered utility sets, so every chip animated at once. The variant now declares the whole animation in one utility, and the cap scenario asserts the computed delay and fill mode.
* fix: Ship the Starter Fade From the Shared Tailwind Preset
The starter variant referenced a keyframe that only the app stylesheet defines, so a host using the published preset would get no entrance. The keyframe and its staggered animation now live in the preset next to the loading dots, and the variant uses the animate-starter-in utility.
* chore: Retest Against Current Dev
* refactor: Move the Scroll Button Into @librechat/chat
* refactor: Make the Package Scroll Button the Message List Default
* test: Pin the Scroll Control Fade and Its Reduced-Motion Path
* fix: Drop the Scroll Control's IntersectionObserver Guard
The observer starts in an effect, which never runs in a server render, and every browser has the API. The guard existed only for jsdom, which now gets an inert observer from the package's jest setup.
* chore: Retest Against Current Dev
* fix: Show MCP Server Titles in the Tool Library and Search by Them
Tool Library cards and agent tool rows showed the MCP server's config key because buildCatalog named each item after it, and search could only match that key. Name items from the resolved title in server.metadata, falling back to the key. Search also matches the key through the item id, as the MCP Servers panel does, so existing key searches keep working.
* test: Cover MCP Server Titles in the Tool Library and Agent Tool Rows
* test: Find the E2E Memory Card and Row by Title in MCP Specs
* fix: Resolve User and Group Balance Overrides on Sign-In
Login routes pass the lean user document from the passport strategy, which has _id but no id, so the balance sync resolved config from the role alone and wrote role-level balance settings over user and group overrides on every login.
* test: Cover the Sign-In Balance Sync With a User Override Scenario
* fix: Report Union, Object, Array and Later Fields in Tool Validation Feedback
Type checks now cover integer, array, object and null, accept any member
of a declared type union, and scan every schema property. The cap of 8
applies to reported fields instead of inspected properties, so a mismatch
past the eighth property is still named.
Related to #17875
* fix: Read anyOf, oneOf and nullable Types in Tool Validation Feedback
Pydantic and zod describe an optional object as an anyOf of object and null,
and the Gemini sanitizer collapses unions into a type plus nullable: true.
Both shapes now report the field, and a valid null is no longer flagged.
Adds a mock MCP tool with a nullable object after its eighth property and a
scenario that checks the model receives the field name but not the value.
* refactor: Read Subagent Conversation Data Through the Subagents Host
* refactor: Move Subagent Conversation and Shared Activity Dialog Into @librechat/chat
* test: Spec the Subagent Conversation and Shared Dialog Against Fake Hosts
* fix: Keep the Panel Selection Atom Out of the Components Entry
* test: Open a Shared Child Run in Its Activity Dialog
* fix: Style the Moved Subagent Dialog With Variants and Static Classes
* test: Pin That a Shared Child Run Card Stays Closed
* chore: Retest Against Current Dev
* refactor: Derive a Send's Code Decision Inside @librechat/chat
TurnServices.resolveCode computed the code workspace and approval a send
carries in the app. It becomes readCodeInputs, a storage read of the raw values,
and the package derives the decision through resolveTurnCode. The read goes
through a ref updated each render, so a captured adapter still reads the newest
render's values.
* test: Close Mocked Readers Over Each Render in the Freshness Spec
* test: Close the Mobile Drawer Through the Race-Tolerant Helper in Two Agent Specs
Two specs closed the drawer after selecting an agent with a bare click on
Close sidebar. The drawer closes itself as the selection lands, so the click
waited on a button that stopped being visible and ran into the 120 s test
timeout on phones. closeMobileDrawer, which #17764 added for this, handles that
race and keeps the inert check.
* refactor: Ship Icon Sizes, Status Type and the Label Sweep From @librechat/client
* test: Prove a Consumer Gets the Shared Part Classes From the Published Stylesheet
* test: Stop Asserting the Blockified Display of the Elapsed Reading
* fix: End the Status Row Type Chain in a Length for Standalone Consumers
* test: Cover a Deployment Skill File Removed at Runtime
* fix: Name the Skipped Skill File in the Console Log
The default console format prints the message and request context but drops metadata, so a relativePath passed as metadata only reached the JSON file log.
* test: Keep the Removed-File Fixture Skill Out of the Palette
As a user-invocable skill it added a row to the virtualized composer palette and pushed the Your files section out of the rendered window, which broke palette-show-all.spec.ts.
Multer assumes latin1 for multipart filenames, and the skills routes have no decode step, so a non-ASCII name reached the handlers garbled. A file name content filter could not match it on import, and the file upload route stored the garbled name.
* fix: Encode Unicode Filenames in Project File Uploads
Project uploads sent the raw filename, which multer reads as latin1 and the upload route then decodes with decodeURIComponent. Non-ASCII names were garbled, a stray percent sign failed with a 400, and a valid escape such as %20 was silently rewritten. Percent-encode the name the same way chat uploads do.
* test: Cover Unicode and Percent Project Upload Filenames
The withheld-reply scenario restored the unread baseline in Mongo right after the reply tab finished, while that tab was still posting its own seen acknowledgement. When the acknowledgement landed second it overwrote the baseline, the first tab polled a conversation the server considered read, and its dot never appeared. The sibling scenario already waits for the acknowledgement; this one now does too.
* fix: Let Native Azure Deployments Declare Their Reasoning Efforts
The composer effort control needs explicit capability evidence on Azure,
because a deployment name does not identify its model, but the native
Azure config had no way to supply any, so the control could never appear.
Add an optional per-model reasoningEfforts list to Azure groups. The
endpoints config serves only model names and efforts, and the composer,
the submit check and the server override validation all resolve them
through getReasoningParamDefinitions. Undeclared deployments stay without
a control and model spec locks still apply.
* test: Assert Auto as the Reset Button in the Azure Effort Scenario
* test: Ignore Looping Animations When the Memory Alert Settles
The spec waited for every animation in the document to finish. A looping one
never does, so it hung whenever earlier specs in its shard left a chat list
long enough to show pulsing skeleton rows. Adding any spec can move it there:
the shard planner places it in shard 9 instead of 6 with this branch.