* refactor: Move the Fork and Feedback Actions into @librechat/chat
Fork reads the viewer's fork defaults and starts the fork through two new
MessagesHost members, keeping its toasts and rate-limit copy in the package; the
app supplies them from its settings atoms and the fork mutation, whose own
callbacks open the copy so a row that unmounts mid-request still lands there.
Feedback moves as is. @ariakit/react becomes a peer dependency of the package.
* test: Cover Fork and Feedback in the Mock Harness
* chore: Sort MessageRow Spec Imports
* fix: Resolve Fork Labels Per Call and Pin the Fork Toast Locator
* test: Give the Consumer Spec Messages Host the Fork Members
* fix: End an Answer's Claim Only When Its Generation Ends or a Newer Pause Replaces It
* fix: Match a Stop to the Answered Generation by Its Record When No Timestamp Names It
* fix: Date an Undated Answer by the Pause a Later Source Reports
- Let agents use Langfuse prompts by production label or pinned version,
with the existing permission and usage checks.
- Add a configuration switch, disabled by default. Preserve saved links
while disabled, without resolving them or falling back to inline text.
- Share source-specific production and exact selections across agent
links and the prompt service. Reject source mismatches.
- Add revision lists for the Agent Builder version picker. Resolve
Langfuse instructions on each initialization without caching.
- Improve disabled and unavailable states and test source changes,
revision loading, and permission failures.
* 🥐 chore: Standardize Bun-First Workspace Typechecks
* fix: Preserve Bun 1.4.3 Checker Compatibility
* chore: Pin Bun 1.4.4 for Local and CI Typechecks
---------
Co-authored-by: Lia <lia@librechat.ai>
The login tests reach getAppConfig through the balance middleware, which
loads librechat.yaml from the repository root by default. The file is
gitignored, so a developer who keeps one locally runs the spec against
their own config; an unresolved `${VAR}` MCP URL in it makes
configSchema.safeParse throw and every login returns 500. Point
CONFIG_PATH at a missing file so the spec runs with no config, as CI does.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* 📵 chore: Add an Offline OSS Scanner Environment
* 🛡️ docs: Add the OSS Scanner Threat Model
Tells the scanner LibreChat's trust boundaries, the components that matter, how to run the tests offline, how we rate severity, and what is out of scope.
---------
Co-authored-by: Lia <lia@librechat.ai>
Co-authored-by: Danny Avila <danny@librechat.ai>
* refactor: Move Composer Input Controls and Shortcut Helpers Into @librechat/chat
* test: Cover Interrupt Now on a Queued Composer Row
* test: Read the Moved Escalate Button in the Composer Color Guardrail
* refactor: Reuse the Core Draft Check for the Send Button
* fix: Point the Acceptance Harness at the Moved Composer Shortcuts
* perf: Serve JWT Bearer User Lookups From the Auth User Cache
* perf: Batch Agent Permission Reads on the File Access Path
* perf: Remember a Loaded Google Service Key Across Endpoint Config Reads
* perf: Run Independent Startup, Tool, Skill and Erasure Reads Concurrently
* fix: Drop JWT User Cache Writes That Race a User Mutation
Confirm each cache write with a second read and evict when the user changed,
so an in-flight lookup cannot repopulate an entry a credential change just
evicted. The cache mode now arrives through the loader dependencies.
* refactor: Move Agent Skill Access Reads Into packages/api
The chat-completions and Responses controllers call loadSkillAccess instead of
orchestrating the three reads in CJS. The preview route spec mocks the batched
file access helper the middleware now calls.
* test: Cover Agent File Access, JWT Bursts and Remote Skill Reads End to End
* fix: Serve a Cached JWT User Only While Its updatedAt Is Current
A cache hit now reads the stored updatedAt and deletion fence instead of the
fence alone, so a document cached from a read that raced a user mutation is
never served, even when the write landed after that mutation's eviction.
* refactor: Move Startup Migration Checks Into packages/api
* refactor: Inject the Startup Migration Checks Instead of Mongoose Models
* fix: Serve Cached OpenID Users Only While Their updatedAt Is Current
Moves the freshness check into getCurrentCachedAuthUserDoc and uses it on the
OpenID bearer path as well as the local JWT one, so a document cached from a
read that raced a credential or role change is never served there either.
Fixesberry-13/LibreChat#432
* fix: Compare Authorization Fields on Every Auth Cache Hit
updatedAt has millisecond resolution, so the freshness read also compares the
token retirement stamps, role, tenant and two-factor state.
* fix: Annotate the Auth Freshness Projection for Isolated Declarations
* perf: Reuse the OpenID Cache Freshness Read for Deletion Admission
A cache hit already proved the deletion fence is clear, so only users read from
the database take the separate fence query.
* refactor: Keep the OpenID Cache Freshness and Fence Decisions in packages/api
* test: Exercise JWT Cache Freshness Against Real User Methods in MongoDB
* fix: Stamp the JWT User Lookup So the Cache Write Fence Applies
* fix: Reload a Google Service Key File When It Changes or Disappears
The availability check remembers a loaded key only while the file's mtime and
size are unchanged, and fetches URL sources every time as before.
* fix: Recheck a Service Key File After a Permission Change
* refactor: Move resume-on-load and the stream steer state into the chat package
* refactor: Drop the useSessionEffects bridge now that resume on load lives in the chat package
* 🚤 fix: Keep Host Tool Loading Independent of Worker Status
* fix: Classify Workspace Tools at the Agent Boundary
---------
Co-authored-by: Lia <lia@librechat.ai>
Pushes to main that touch packages/data-provider/package.json without bumping the version failed with "You cannot publish over the previously published versions". Check whether the exact version already exists on the registry and skip packing and publishing when it does, as the client workflow does.
splitMarkdown and editableParts move to packages/chat/src/core/parts and are
exported from the package root, so hasEditablePart is available to package
components. The micromark and mdast packages the splitter parses with become
peer and dev dependencies of @librechat/chat at the client's ranges, kept
external by the build, and the package jest config transforms them like the
client's does.
* feat: Retire Expired and Older Pauses and Reconcile the Held Pause With Sync Snapshots
* fix: Let an Expired Pause Retire the Held One It Ties
* fix: Hold a Pause the Sync Frame Buffers and Follow an Answer Whose Pause a Snapshot Released
* fix: Keep a Stale Buffered Pause Out of the Host Reducer
* fix: Retire a Released Pause When Its Generation Ends
* fix: Let Only the Pause a Sync Frame Stands for Reach the Host Reducer
* feat: Answer, Restore and Stop a Paused Generation From the Chat Core
* fix: Retire a Pause the Server Refused Once Its Probe Owns the Chat
* fix: Keep a Pause the Server Refused Over for Later Reads and Replays
* feat: Reattach to a Running Generation and Keep a Dropped One Stoppable
* fix: Keep a Continued Generation's Endpoint When the Host Selection Changes
* fix: End What a Dropped Generation Leaves When a Retry, a Stop or a Status Read Settles It
* test: Cover a Reattachment That Yields to a Turn the Host Starts While It Selects the Endpoint
* fix: Settle What a Reconnect Answers for and Let It Go Before a Store Is Told
* fix: Return a Failed Chat to Ready as One Write
* refactor: Move Subagent Progress State and Content Folding Into @librechat/chat
* refactor: Attach the Subagent Activity Feed Through the Chat Transport
* test: Tag the Subagent Activity Panel E2E Specs With Scenarios
* test: Pin Replay, Ready and Live Steer Ordering Across an Activity Reconnect
* fix: Stop the Subagent Activity Feed on Refusals and Bound Idle Reconnects
* fix: Keep Attaching the Subagent Feed While the Child Thread Is Still Reserved
* test: Read Steer Order the Same While the Child Turn Streams and Once It Settles
* refactor: Move Composer Attachment Components Into @librechat/chat
* test: Cover the Pasted Text Editor From Paste to Saved Edit
* test: Keep the Real Data Provider Enums in the ChatView Subagent Spec
* refactor: Build the Source Badge Styles at Render Time
* test: Reopen the Pasted Text Correction From Its Replacement Chip
* chore: Re-record the Attachment Design Suppressions on Dev
* test: Merge the SteerPart Package Mocks Into One Factory
* fix: Fence Auth User Cache Writes After Eviction and Evict Favorite Holders
An eviction now leaves a per-user fence in the auth user cache that outlives
every document read before it. The reader skips a cached document while its
user is fenced, and a lookup slower than one cache TTL is not cached, so a
request that read the user before a change can no longer repopulate the cache
afterwards, and a document the reverse index lost is hidden as well. Role and
user methods share one bulk eviction helper, and the terms reset only waits
out the TTL when an eviction failed.
Deleting an agent, deleting a user's agents and removing an agent from
favorites now evict the cached auth documents of the users whose favorites
changed.
* fix: Anchor Cached Auth Document Expiry to the Lookup Start
A stalled reverse-index write could land a stale document after the eviction
fence expired, with a fresh TTL. The document now carries an expiry two cache
TTLs after its lookup began; its Redis TTL is computed after the index write
and the reader refuses it past that expiry.
* refactor: Time the Auth User Lookup in the API Package
The lookup start that anchors a cached document's lifetime is now stamped by
timeAuthUserLookup in packages/api, so the OpenID JWT strategy only wires the
lookup through it.
* fix: Wait Out the Full Cached Document Lifetime After Failed Eviction
A document read just before a change can live two cache TTLs from its lookup,
so a credential change or terms reset whose eviction failed now waits that
long before reporting success.
* fix: Recheck Cached Auth Document Expiry After the Fence Read
A fence read that stalls can return after both the fence and the document
expired, so the reader checks the document's expiry again before serving it.
* feat: Stop the Turn in Flight From the Chat Core
* fix: Adopt the Conversation a Late Stream's Stop Names and Record a Replacement Before Its Rollback Notifies
* refactor: Move Tool Disclosure State and Leaf Part Components Into @librechat/chat
Disclosure contexts and atoms, tool intent labels, useRowHandoff and
useFollowScroll move to packages/chat/src/react/parts (exported from
@librechat/chat/react); CancelledIcon, InProgressCall, AuthorHeader, Stdout,
BareStatus and the row class constants move to
packages/chat/src/components/parts (exported from @librechat/chat/components).
Disclosure reads the package host, and ToolContentQuery is declared from
data-provider and react-query types. The AuthorHeader and useFollowScroll specs
move with their modules.
* test: Cover the Author Header That Resumes After a Steer
* test: Keep the Package Entry Intact When Mocking AuthorHeader
* refactor: Move Pure Part Helpers Into @librechat/chat
Tool output parsers (JSON fields, background handles, wakeups, background
tasks, command results) and the small routing, params and rerun helpers move
from client Content/ into packages/chat/src/core/parts, exported from the
package root. Client importers read them from @librechat/chat; parseJsonField
becomes a named export so the root barrel can carry it.
* test: Cover Wake-Up Turns Rendering as Task Cards
* fix: Point the API Background Spec at the Moved Parsers
* fix: Bound outbound HTTP calls that had no timeout
Add timeouts.toolsMs, speechMs and imageGenerationMs to the timeouts block and apply the
existing budgets to the RAG delete paths, built-in tool APIs, TTS/STT providers (TTS also
aborts when the client closes), Stable Diffusion, Firebase reads and Mailgun. Calls made
before app config exists (remote config, RAG health, service key URL, Google refresh) get a
fixed 10 second budget.
* fix: Honor configured budgets for RAG deletes and per-chunk TTS deadlines
* refactor: Share one budget helper between speech and image CJS callers
* test: Cover a stalled RAG delete with a mock scenario
* test: Wait for the RAG embed before stalling the delete
* test: Embed at upload so the stalled delete has a vector to remove
* fix: Close a started TTS response on a late chunk timeout and share the service budget helper
* fix: Fail startup when the remote config fetch times out
* feat: Add a Minimal Consumer Example for @librechat/chat
A Vite React page under packages/chat/example renders a themed chat from the package's public entries and the @librechat/client theme, with an in-memory transport standing in for the server. Its jsdom smoke test runs in the package's jest suite, the package typecheck covers it, and ESLint keeps it to public imports and semantic theme roles.
* fix: Let Enter Confirm an IME Composition in the Example Composer
* fix: Report a Failed Turn in the Example and Restore Its Text
* fix: Localize the Example Copy and Guard the Safari IME Enter
* fix: Drop a Rejected Turn, Load the Component CSS, and Keep the i18n Instance in the Example Build
* refactor: Replace the Example App With a Public-Entry Consumer Spec
The owner does not want an example app in the repository. packages/chat/test/consumer.spec.tsx keeps what it proved: a chat built only from @librechat/chat, /react, /components and @librechat/client, over an in-memory transport, streams a reply into the message list under the ClickHouse theme. It sits outside src because chat/boundary rejects self-imports there; jest maps the entry names to the sources and the package typecheck covers it. The example app, its Vite, Tailwind and PostCSS config, its lint block, script and dependency lines are removed.
* chore: Declare the Consumer Spec's i18next Dev Dependencies
* chore: Retest Against Current Dev
* feat: Reject Module-Load Use of Other Packages in the Chat Package
App specs mock whole packages (@librechat/client, librechat-data-provider), so a package module that calls into one or reads a member of one while it loads throws before any test runs. chat/boundary now reports a call, new, tagged template or member read through an external import at module scope in packages/chat/src outside specs; React and Jotai are exempt through a loadSafe option.
* fix: Count Immediately Invoked Functions and Computed Field Keys as Module Load
* fix: See Through Erased TypeScript Wrappers in the Module-Load Check
* fix: Count Destructuring, Spreads and Wrapped IIFEs as Module-Load Reads
* refactor: Judge Every Load-Time Reference to an External Import
The check now walks each external import's references once and reports any that runs on load outside an inert position, instead of listing the node types that read an import. Class heritage, in, instanceof and iteration are reported with no new visitor.
* fix: Follow Branches to Their Consumer and Skip Erased Heritage Types
* fix: Count Constructed Classes and Destructuring Assignments as Module Load
* fix: Count Constructors of Constructed Classes and Branch-Fed IIFEs as Module Load
* fix: Follow Const Aliases of External Imports in the Module-Load Check
* fix: Find Aliases Behind Wrappers and Branches
* fix: Track require Bindings, Skip Writes and Defer Generator Bodies in the Module-Load Check
* fix: Follow Assignment Aliases of External Imports
* fix: Collect require Bindings From Exported Declarations
* refactor: Cover Module-Load Failures With a Mocked-Dependency Load Spec
A packages/chat spec now loads the index, react and components entries with every package the sources import (React and Jotai aside) mocked to a module whose exports are missing, the way an app spec's factory mock leaves it, and asserts nothing throws. That covers aliases, require, invoked functions and the rest at runtime, so chat/boundary keeps only the plain case: a call, new, tagged template or member read through a direct import binding at module scope.
* chore: Retest Against Current Dev
* fix: Exempt Spec Support Files and Node Built-ins From the Module-Load Check
* refactor: Move Message Row and Timestamp into @librechat/chat
Add a MessagesHost the message list reads the viewer's clock format and row
slots from, registered by the app at load, and move MessageRow,
MessageTimestamp, the row width helper, the timestamp formatters and the
minute ticker into the package with their specs. The user attachments a row
shows reach it through the host's files slot until they move.
* test: Cover Message Rows on Chat and Shared Routes
* chore: Prepare the Chat Package for Message Components
Give packages/chat a lint-only Tailwind entry so the design rules resolve the
library's theme utilities, allow the app status selectors its components carry,
expose the active language from @librechat/client, and let the locale check
accept keys that resolve through plural forms.
* refactor: Move Message Row Leaves into @librechat/chat
Move the hover style helpers, SubRow, SiblingSwitch, the elapsed timer with its
duration labels, HeaderLabel and the system event header into
packages/chat/src/components/messages, export them from
@librechat/chat/components, and point the client at the package.
* test: Cover Message Row Leaves in the Mock Harness
* chore: Drop the Chat Lint Entry in Favor of components.json
* test: Run the Elapsed Spec Against the Package
* fix: Label Composer Thinking Levels and State the One-Shot Scope
* test: Cover Composer Thinking Labels and One-Shot Scope in E2E
* test: Cover Thinking Focus for Option Lists Without Auto
* test: Wait for the Thinking Popover's Open Focus Before Arrow Keys
* test: Match the High Effort Radio Exactly
* fix: Keep the Composer Thinking Level for the Rest of the Conversation
* fix: Keep a Queued Turn's Picked Level on the Turns Queued After It
* test: Name the Numeric Reasoning Control by Its Thinking Label
* fix: Name the Queued Turn's Level on the Thinking Button
* test: Dismiss the Thinking Tooltip Before Typing on a Phone
* fix: Respect Enforced Specs and Parameters Edits Over the Inherited Thinking Level
* test: Expect the Model Spec Enforcement Flag in Chat Settings
* fix: Keep a Queued Turn Off a Level Its Target No Longer Runs
* fix: Record the Target a Queued Turn's Level Belongs To
* refactor: Inherit the Thinking Level on the Server
* perf: Answer the Branch Level Lookup Without Tree Links When Nothing Was Picked
* fix: Respect the Thinking Switch and Held Queue Rows Over the Inherited Level
* fix: Read the Branch Level Once and Inherit Only on Fresh Turns
* fix: Keep a Parameters Thinking Edit From Turning Thinking On
* fix: Withdraw a Staged Thinking Level When Thinking Is Switched Off
* fix: Scope the Thinking Switch Transition to the Conversation
* test: Drop the One-Shot Thinking Scenario the Conversation Level Replaces
* perf: Parallelize Serial Lookups on Agent Load, Discovery, Skills, and Schedules
* fix: Address Review Findings on Parallel Lookups
Stage same-level discovery results before collecting edges so a shared sibling is initialized once, bound resume snapshot fetches to small chunks so over-budget graphs stop early, and surface the first ordered image failure without waiting on later reads. Update the skill name test for bounded concurrency.
* test: Tag Existing Mock Specs as Acceptance Scenarios for Parallel Lookups
* feat: Make Parallel Lookup Bounds Configurable
Add a concurrency section to librechat.yaml for MCP catalog, skill name, skill file and resume agent lookups. Defaults equal the previous hard-coded values.
* style: Sort Imports for Configurable Lookup Bounds
* test: Match Model Spec Skill Resolution Args With App Config
* test: Close the Builder Drawer Before Chatting on Narrow Viewports
* fix: Surface Fatal Discovery Errors Without Awaiting Later Siblings
* fix: Reject Discovery on the First Fatal Load Regardless of Queue Order
* fix: Keep Image Encoding Fail-Fast With Parallel Reads
* fix: Reject Resume Snapshot Lookups Without Waiting on the Whole Chunk
* feat: Make Connected Agent Discovery Concurrency Configurable