- Map the four NVIDIA Nemotron models on Bedrock to their AWS-documented context windows (256K for Nemotron 3 Super and Nano, 128K for the Nano v2 pair), so agents stop falling back to the 32K default that MCP tool schemas exhaust once a tool result arrives.
- Price the same models at their Bedrock on-demand rates so the context and pricing maps stay in parity.
Bound model-facing Flux images to the 5 MiB base64 budget for standard and finetuned generation. Preserve original image data for attachment storage under the existing resolution policy.
Fixes#6954.
* fix: Read the Account and Session Together and Test Cancel on a Real Database
- Start the setup user and refresh session lookups together in the cancel controller instead of one after the other, as the opt-in confirmation already does
- Run the cancel controller spec on mongodb-memory-server with the real data-schemas methods, so the session and compare-and-swap predicates under test are the production ones, and add a negative case for each
- Keep only a module wiring test in the CJS controller spec, since the database behavior is now covered where it can run for real
* test: Pin the Cancel Response When the Session Lookup Fails
Both lookups now run before any cookie or write, so a failed session lookup answers 500 with the browser and the stored state untouched. Cover that on the real database.
* fix: Keep the Saved Files Draft Until the File List Loads
After a reload the live file map is empty until the server's file list arrives, and the restore that rebuilds it needs that list. The save effect rewrote the stored draft from the empty map in between, so a pasted-text chip or any other attachment was gone before it could be restored. It now waits for the list. This is what turned the composer-attachments restore scenario red on dev.
* 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
* 📏 fix: Measure Usage Percent Against the Allowance Each Period Granted
Store the credits granted at the start of each balance period on the record and measure the usage percent against it, so changing refillAmount or startBalance no longer moves a user's percent until their next grant. Legacy records fall back to the configured amount.
* 🕰️ fix: Ignore a Stored Allotment a Later Refill Has Replaced
Stamp each stored allotment with the time of its grant and measure against it only while no refill is more recent, so a refill by a replica that predates the field falls back to the configured amount instead of reporting usage against the prior period's allowance.
* 🔒 fix: Write the Allotment Grant Atomically and Return Its Period Start
Insert lastRefill with the credit, allotment and grant time when a balance record is created, so a racing creator never moves another's period start past its grant. Return lastRefill even without auto-refill so the client can place a stored allotment in its period, and fall back when it cannot.
* 🔐 fix: Authorize Restored MCP Tools Before Agent Revert Writes Them
Reverting wrote the version snapshot, including its mcpServerNames, before
filterAuthorizedTools ran. Agent-scoped MCP access is granted from that index,
so a restored server vouched for itself and an editor without access to it
could re-attach it. A snapshot without mcpServerNames also left the current
index in place, keeping agent-scoped access to servers no tool used.
The handler now authorizes the restored tools against the agent as it stands,
treating only its current tools as existing, and passes the authorized tools
and rebuilt index to revertAgentVersion so the restore is one write pinned to
the revision it was checked against. revertAgentVersion always rewrites the
index for the restored tools.
* 🔐 fix: Exclude the Target Agent's Own Grant When Authorizing a Revert
Moves the restore authorization into authorizeAgentRestore in packages/api,
leaving the revert controller as wiring.
A newly restored MCP tool whose server the user reaches only through agents
must now be reachable through an agent other than the one being reverted,
checked fresh against the database. The target's current or cached index
(including one left stale by an earlier revert) no longer vouches for it.
Prior server names are matched through their normalized aliases, and only for
tools the agent already had. When nothing resolves, the pre-restore index is
kept instead of deriving names from tool keys.
A revision-filtered restore that misses is reported as a 409 conflict.
The route spec restores the real ~/config so the MCP registry is consulted
rather than reported unavailable.
* 🔐 fix: Check Newly Restored MCP Servers Against Live Grants
The restore gate now runs for every approved MCP tool, not only consumeOnly
ones. A tool the agent doesn't already have, on a DB-backed server, is
admitted when a current tool still uses that server, or on a live direct
server grant, or through another viewable agent. A grant the registry still
caches after revocation no longer becomes durable through the restored index.
Admission is memoized per server. A snapshot tools array that sanitizes to
nothing restores an explicit empty set. Active-server matching no longer
reuses the index fallback, so a stale entry is never read as in use.
* 🔐 fix: Treat Legacy Cached DB Configs as User-Sourced in the Restore Gate
A config cached before an upgrade may carry a dbId without source. The
restore gate now uses isUserSourced, so such configs get the live access
check instead of passing as operator-managed.
* 🔐 fix: Recheck Unindexed Existing MCP Tools in the Restore Gate
Holding a tool no longer exempts it from the live check. Only a current tool
whose server the agent's index already lists skips it. Otherwise an unindexed
tool on a legacy agent could pass on a revoked grant the registry still
caches, and the restore would add its server to the index.
* ⚡ perf: Run Restore Admission Checks for Distinct Servers Concurrently
filterAuthorizedTools now starts every admission check during its single pass
and awaits them together, so each distinct DB-backed server no longer adds a
serial lookup wave to the revert. Tool order is preserved and server names
are recorded only for admitted tools.
* refactor: Move Part Helpers and activityLabels Into the Chat Package
activityLabels moves to packages/chat/src/core/parts and exports from
@librechat/chat. useLazyCollapseBody, useProgress and useThrottledValue move
to packages/chat/src/react/parts (@librechat/chat/react). useExpandCollapse,
CopyButton, ActionButton, LangIcon and its icon paths move to
packages/chat/src/components/parts (@librechat/chat/components), since they
read @librechat/client. Specs move with them, client importers and mocks point
at the package entries, and design-lint suppressions follow the moved files.
The fold-type set in activityLabels is built on first use so the module reads
no data-provider enum at load, with a spec pinning its members.
* test: Point the About Spec's Localize Mock at the Package's Copy Button
* chore: Retest Against Current Dev
* fix: Only Hand Out OpenID Access Tokens the Bearer Verifier Accepts
The OpenID reuse path in refreshController and the post-refresh fallback in
getOpenIDAppAuthToken returned the cached access token as the app auth token
whenever its exp was in the future. Entra issues Microsoft Graph access tokens
to sign-in clients by default; their aud names Graph, so the bearer verifier
rejects them and every protected request answers 401 until the client loops.
Both paths now require the access token's aud to name the client id or an
OPENID_AUDIENCE entry, the same audiences the verifier accepts. The reuse path
refreshes with the IdP instead, and the post-refresh fallback keeps opaque
tokens as before. Audience parsing is shared with openIdJwtStrategy.
Related to #17791
* test: Exercise Real OpenID Audience Helpers in AuthService Spec
The AuthService spec stubbed the audience helpers as pass-throughs, so it
could not catch a foreign-audience access token escaping the post-refresh
fallback. Use the real exports and cover Graph-audience, configured-audience
and opaque access tokens through setOpenIDAuthTokens.
* 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: 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 Attachment Row Into @librechat/chat Behind the Files Host
* test: Cover the Attachment Row Against a Fake Files Host
* test: Run the Files Host Deletion Against the Real Hooks and Mock Only the Service Call
* test: Drive the Fallback Deletion Through the Real File Card
* test: Run the Row's Host Cases Against the Real Thumbnail and File Card
* test: Read the Host Preview From the Thumbnail Image Source
* 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
* refactor: Retire Host Tool Outcome, Sibling and Approval Members
The package now owns the tool outcome rules and the attachment merge they
read, selects the visible sibling through its own atom, and derives the steer
path's code approval mode from raw inputs. ChatHost loses getMappingOptions and
useSetSiblingIdx; SteerHost.useCodeApprovalMode becomes useCodeApprovalInputs.
* refactor: Select the Branch Tail Inside @librechat/chat
The tail of the branch on screen becomes package logic, read through
useBranchTail. ChatHost loses useLatestMessage and useLatestMessageId;
SteerHost.useLatestMessage becomes the useRouteConversationId adapter.
* test: Exercise Steering Against a Real Message Cache
The steering spec mocked React Query to return one message list for every key, so
it could not notice a read of another conversation's cache or a tail that stops
following the cache. It now provides a real QueryClient, seeds each conversation
key, and covers the route and another chat holding different tails and a cache
that hydrates after mount.
* refactor: Read the Thread Panel's Jotai State Through the Chat Surface
* refactor: Move the Subagent Thread Panel Into @librechat/chat
* test: Spec the Subagent Thread Panel Against Fake Hosts
* test: Cover the App Subagents Host Adapters the Panel Move Relies On
* fix: Style the Moved Subagent Panel With Shared Variants
* test: Pin the Shared Variants the Panel Draws Its Picker and Inline Actions With
* perf: Keep the Subagent Thread Panel Out of the Boot Bundle