* 🙈 feat: Let Config Profiles Hide librechat.yaml MCP Servers and Remove Their Fields
Config profiles are the `__base__`, role, group and user documents written through `/api/admin/config`. A profile can tombstone `mcpServers.<name>`, and the merge then removes that server from the principal's resolved `mcpConfig`.
* 🛡️ fix: Apply MCP Hiding to Server Instructions, Failed Removals and the Startup Audit
- Server instructions bound for the model now leave out librechat.yaml servers hidden from the acting user, including one a model spec pins. `applyContextToAgent` passes the request user through `formatInstructionsForContext`, which drops the user's hidden set from the user-less catalog it reads.
- A config-tier candidate that removes a librechat.yaml value (a field, or a key of a record such as `headers`) and fails inspection now stands as a failed config-source entry instead of falling back to the librechat.yaml base that still carries the removed value. A failed change that removes nothing still falls back, and an entry with no transport (fields written over a hidden server) keeps the documented librechat.yaml fallback.
- The startup tombstone report reads every tenant's profiles in the system tenant scope, so it also runs from entry points without an outer system scope under strict tenant isolation, and names each profile's tenant.
* 🗂️ fix: Project the Startup MCP Tombstone Report Query and Run It Alongside Connections
* 🔑 fix: Keep Removed OAuth Headers Out of the MCP OAuth Callback
* 📝 docs: Note That Removed OAuth Headers Stay Out of MCP Sign-In
* 🔐 fix: Recheck the MCP OAuth Flow's Server Before Initiation Redirects
OAuth initiation redirected to a flow's stored authorization URL, or started
a new authorization from the flow's stored server URL, without resolving the
server again, so a principal whose profile hid the server after the flow was
created could still be sent to its authorization endpoint. The callback's
active-server check and its OAuth header fallback lived in the CJS route.
resolveMCPOAuthTarget (packages/api/src/mcp/oauth/target.ts) now owns both:
it resolves the server in the principal's current config with a strict read,
refuses a flow admitted for another server, a server the principal can no
longer use, or a redefined one, and selects the headers the flow sends (the
initiation snapshot, else the principal's config). Initiation calls it before
either redirect (404 or 400); the callback calls it before the exchange and
from the token-store fence.
* 📝 docs: Note That a Hidden MCP Server Ends OAuth Sign-Ins Already Started
* 🚩 feat: Publish the yamlMcpHide Admin Config Meta Flag
Profiles can hide librechat.yaml MCP servers and a hidden server is an
access boundary, so GET /api/admin/config/meta now reports
features.yamlMcpHide for the admin panel to gate its hide UI on. The
meta.spec probe checks the behaviour itself: the admin API stores an
mcpServers.<name> tombstone, the merged config drops the server, the
registry withholds it from principal lookups and checkout but not from
app-level lookups, process-backed servers stay, and an unreadable
principal config withholds every hideable server.
* 🧪 test: Align MCP Tombstone Specs With the Portable Config Store
The admin config handlers now write whole documents through
writeConfigDocument, so the end-to-end tombstone spec wires it in place
of the removed upsertConfig/patchConfigFields deps. A reset of a whole
MCP record now also prunes the parents it leaves empty.
* 🗄️ refactor: Portable Admin Config Store with Revisions, Key Encoding and Coherent Writes
Admin config writes went straight to MongoDB with dotted-path `$set`, `$unset`, `$pull` and `$addToSet`, and nothing checked `configVersion`. That caused six problems.
* 🔁 fix: Reload librechat.yaml on Admin Config Writes
Base-only sections such as `filters` and `messageFilter` live only in librechat.yaml, and an admin
config write was how a running server picked up an edit to them. The coherent-cache change limited
admin-write invalidation to the override and MCP caches, so the YAML base was never re-read after
startup: the content-filter e2e suite, which activates filters by editing the YAML and writing a
reload sentinel, saw `filters: undefined` from GET /base and every activation timed out.
Admin config writes clear the YAML base config again, alongside the override and MCP caches, and
the clear is broadcast with the others, so every instance reloads its YAML. The reload keeps the
last good config on failure and republishes the tools derived from it, so the global tool cache is
left for it to overwrite rather than emptied first.
* 🧭 fix: Cap Dotted Field Path Depth and Hold Revision Preconditions on No-Op Writes
A dotted fieldPath was split without the 32-segment cap that segments already had, so a deep path reached the recursive path writer and failed with a 500. Both spellings now share MAX_CONFIG_PATH_DEPTH and answer 400 CONFIG_PATH_INVALID / too_deep on PATCH, tombstone and DELETE /fields.
Writes reduced to nothing (protected interface permissions, base-only or dedicated sections) answered 200 without checking expectedRevision. They now compare it with the stored document and answer 409 CONFIG_VERSION_CONFLICT with currentRevision on a mismatch, as the write itself would.
* 📡 fix: Retry the Config Invalidation Subscription When Redis Comes Back
A SUBSCRIBE that failed at startup (Redis unreachable, the client's retries spent) was only logged, and ioredis resubscribes only to channels it has joined, so the instance missed every other instance's config invalidation until it restarted. The subscription is now sent again each time the subscriber becomes ready until one succeeds; after that the client's own resubscribe takes over, and closing stops the retries.
* 🔁 fix: Drop YAML Reloads and Merges Overtaken by a Config Write
Admin config writes clear the YAML base config on every instance, so a write is how a running
server picks up an edit to librechat.yaml. Two races could undo that:
- A base reload that read the old YAML before a write's clear cached it after the clear, and the
base has no TTL, so the old base was served until the next write.
- A request that read the old base before the clear, and was still resolving principals, cached
its merge under the new override generation for the override TTL.
The config service now keeps a base-config generation that `clearAppConfigCache` bumps. A reload
publishes the base, the tools and the last good config only if its generation is still current,
a read after a clear never joins a reload started before it, and the newer reload starts once the
stale one settles so the loader's process-wide side effects end with the newest YAML. Merges
capture the generation, base included, before they read the base config.
Tests:
- service.spec: a clear during a reload (stale success, stale failure, a read waiting on the newer
reload) and a merge built on the pre-clear base.
- invalidation.redis.spec: a write on one instance makes another reload librechat.yaml.
- admin-config-reload.spec (api): every admin config write route, through the real router,
handlers, database and config service, reloads librechat.yaml; a request with nothing to write
does not.
- invalidateConfigCaches.spec no longer asserts an order the parallel clears do not have.
* 🧬 fix: Encode Config Keys in Objects Built in Another Realm
The key encoder treated an object as plain only when its prototype was this realm's
Object.prototype, and returned any other object unchanged, keys and all. Under Jest,
structuredClone (which the secret pipeline uses) returns objects from the host realm, so the
items of an array the pipeline had cloned, such as Azure groups or custom endpoints, reached
the engine with their dotted keys raw. Node in production has one realm, but the encoding must
not depend on prototype identity: a skipped object stores its dotted keys raw without an error.
A plain object is now one whose prototype is null or a root Object.prototype from any realm.
The encoding spec covers an object built in a vm context.
* 🧪 test: Check Dotted Record Keys Reach the Store's Whole-Document Write
dev's admin config route test asserted the keys on patchConfigFields, which the handlers no
longer call; it now asserts them on the draft writeConfigDocument receives.
* 🚩 feat: Publish the expectedRevision, prunesEmptyParents and awaitedInvalidation Config Flags
GET /api/admin/config/meta now reports the three write-coherence behaviours this branch
implements, so an admin client can drop the workarounds each one replaces:
- expectedRevision: responses carry `revision`; PUT, PATCH /fields, POST /fields/tombstone and
PATCH /active take `expectedRevision` in the body and both DELETE routes in the query. A
mismatch answers 409 CONFIG_VERSION_CONFLICT with currentRevision, a malformed value 400
INVALID_EXPECTED_REVISION; unpinned writes re-read and re-apply up to 3 times.
- prunesEmptyParents: a field delete or tombstone removes the objects it empties, stopping at
arrays; shells-only documents count as empty for assign-only lifecycle calls.
- awaitedInvalidation: writes answer after the config caches are cleared (broadcast with Redis),
GET /base bypasses the merged-config cache, and a merge read before a clear is never cached.
meta.spec probes each one against the behaviour: a stale and a malformed precondition on a
PATCH, the store operations' unset and tombstone on nested objects and an array element, and a
held invalidation, the GET /base read options and a merge raced by a clear.
* 🔀 fix: Keep Config Keys Raw Where the Engine Stores Them, So Older Servers Read Them During Upgrades
Servers before the key encoding (v0.8.9-rc1 among them) read every override key literally.
The Mongoose binding encoded every write, so during a rolling upgrade the first write from an
upgraded server turned a stored `gpt-4.1` or Bedrock inference profile into `gpt-4%2E1` for
every server not yet upgraded, and an older server's restore of a deleted group or role
config dropped the `keyEncoding` marker its schema lacks, after which upgraded servers read
the encoded keys literally.
The binding now sends keys raw first and encodes them only when the engine refuses a raw
field name (DocumentDB, Cosmos, FerretDB or MongoDB 4.4 for `$` names): the refused write
changed nothing, so it runs again encoded under the same revision precondition, and that
process encodes later writes holding such keys directly. MongoDB 4.4+ stores dotted names
(5.0+ also `$` names inside a value), so there an older server reads exactly what it wrote. A
document whose raw keys spell escape sequences while none holds a raw `.`, `$` or NUL is
still written encoded, since a raw read would decode it. A raw write clears the marker.
Reads decode a marked document (raw keys an older server added beside encoded ones, or wrote
under a marker it left, carry no escape sequence and read as stored) and an unmarked one
unless some key holds a raw `.`, `$` or NUL, which no encoded document does; that reads an
encoded document whose marker an older server dropped. Encoding stays in the binding for
every engine that needs it (decision #15); the residual cases are in documentdb-compat.md.
* ⏱️ fix: Remove Config Cache Writes a Clear Overtook in Flight
With an asynchronous APP_CONFIG store (Redis or a custom Keyv store), a clear can bump the
generation and finish its delete after a reload or merge checked the generation but before its
`set` landed. The older value then repopulated `_BASE_` until the next write, or a merged
config for the full override TTL.
The base reload, the last-good restoration and the merge cache now check the generation again
once their write lands and delete what they wrote if a clear overtook it. A deleted entry only
costs the next read a reload or a merge; a newer base load cannot be removed this way, since it
starts only after the stale one settles.
The spec mocked the installed `@librechat/api` and `librechat-data-provider`
packages as virtual modules and gave the data-provider mock only a few enums.
On CI the real `@librechat/api` build is sometimes loaded instead of the mock;
it then imports `@librechat/data-schemas`, which reads `ResourceType.AGENT` from
the partial data-provider mock and throws, failing `Tests: api (shard 1/3)`
on dev and on unrelated PRs.
Mock both packages as the real modules they are, and spread the actual
data-provider into its mock so every enum exists whichever API build loads.
* 🔑 feat: Keep Dotted Record Keys in Admin Config Writes and Publish the Config Contract
Keep dotted record keys in admin config write bodies behind a path-scoped
sanitizer, reject $-prefixed and prototype keys after auth, report dotted_key
issues with segments, and publish GET /api/admin/config/meta (features, rules,
limits, versions) so clients can gate on backend capabilities.
* 🧪 test: Expect Issue Segments in the Config Override e2e Scenario
Override validation issues now carry `segments`; assert the full issue shape.
* fix: Bound Outbound Calls Made Inside a Request
RAG queries for file search and message context, OCR provider calls,
action HTTP calls, the moderation check, the stored-image fetch for
base64 providers, and Flux result polling had no timeout. A stalled
upstream hung the chat turn, the upload or the tool call indefinitely,
and Flux kept polling forever on statuses that never become Ready.
Each now runs under a budget from a new librechat.yaml timeouts block
with finite defaults (getOutboundTimeoutMs in @librechat/api). Action
calls also receive the run's abort signal, so stopping a run closes the
socket.
* test: Cover a Stalled RAG Query in the Mock Harness
* fix: Share One Flux Budget Across Submit and Polls, Bound the Vertex Token Exchange
* refactor: Poll Flux Results Through pollUntilSettled
* refactor: Leave Flux Polling for a Separate Migration
* fix: Cap Configured Timeouts at the Node Timer Limit
* fix: Bound the Action Token Refresh and the Whole Image Download
* fix: Cap Every Outbound Budget by Wall-Clock Time
Axios's timeout only fires on an idle socket, so a RAG, OCR, moderation,
action or token response that keeps trickling bytes never ended.
outboundRequestBudget pairs each timeout with an AbortSignal deadline, and
action calls combine it with the run's own signal.
* fix: Name the Budget That Ran Out and Give Actions Five Minutes
The actions default rises to five minutes, since action endpoints are
arbitrary third-party APIs and the old minute could cut off legitimately
slow ones after an upgrade.
A budget overrun now fails with an error naming the timeouts key and its
value, whether it was the idle timeout or the wall-clock deadline, and
logAxiosError logs it as a warning so operators know what to raise. A
deadline cancellation is no longer read as a user stop, so an action that
runs out of time reports a timeout instead of looking like a stopped run.
* fix: Hold Outbound Deadlines Weakly So Finished Requests Leave No Timers
* fix: Check the Budget Before Reading an Error Message
* fix: Report a Wrapped Deadline as a Failure, Not a Stop
* fix: Bound the OAuth Code Exchange and Keep the OCR Budget Hint When Protected
* chore: Sort Imports in axios.ts After Rebase
* feat: Defer MCP OAuth Sign-In Until a Deferred Tool Is Used
Adds an oauthPrompt server option (on_start by default, or on_use) and an
operator-declared oauthToolCatalog. With on_use, an operator-configured
OAuth server whose selected agent tools are all deferred no longer prompts
at turn start: tool search offers the declared catalog (or the user's own
cached catalog), and sign-in happens on the first tool call through the
existing in-run OAuth step. Without a declared catalog the turn-start
prompt is kept. Both fields are operator-only.
Related to #16765
* test: Cover MCP OAuth Sign-In on First Deferred Tool Use
* fix: Bound the Declared OAuth Tool Schema to Non-Recursive Types
* refactor: Move MCP OAuth Deferral Orchestration Into packages/api
* refactor: Move the Cached-Catalog OAuth Prompt Branch Into packages/api
* fix: Defer MCP OAuth Only When the Offered Catalog Covers the Selection
* fix: Cover Runtime OAuth, Background Tools, and Tokenless Discovery in Deferral
* fix: Match Legacy Keys, Honor Background Capability, and Reject Unsupported Schema Keywords
* fix: Resolve Wildcard Background Opt-In and Tighten Declared Catalog Matching
* fix: Validate Declared Tool Schema Types and Enums
* fix: Accept Canonical Catalog Keys the Definitions Loader Accepts
* test: Pin the Declared Catalog Requirement for Cached User Catalogs
* fix: Resolve Deferred Tool Options Across MCP Spelling Aliases
Normalize uploaded image bytes in shared TypeScript and recover mislabeled image history without inferring codecs from generic HEIF brands.
Co-authored-by: Lia <lia@librechat.ai>
Co-authored-by: JeongCheolLee <jclee7503@gmail.com>
* fix: Contain Download and Speech Stream Failures and Cap Speech Input
The code output download piped the Code API stream with .pipe, which adds no
error listener, so a body that failed mid-stream (a bad gzip body, a reset)
became an uncaught exception and the process exited. pipeToResponse uses
stream.pipeline, ending the response and destroying the upstream when the
client leaves.
Text-to-speech piped provider streams the same way and waited only for
'end', so a provider error was uncaught and a client disconnect left the
chunk loop waiting. pipeSegment resolves on end and rejects on an error or
a disconnect, which ends the loop.
A speech request had no length cap, so one request could fan out into
thousands of serial paid provider calls. speech.tts.maxInputChars bounds it
(default 20,000 characters).
* docs: Document the Speech Input Cap in the Example Config
* fix: Reject a Speech Segment When the Client Already Left
* fix: Tear Down Both Ends of a Failed Speech Segment and Cap Streamed Speech
A failed provider segment now destroys the response as well as the source,
so a half-written answer cannot hang the client, and a response-side error
rejects the segment instead of going unhandled. Automatic playback reads a
message only up to speech.tts.maxInputChars, so long messages cannot fan
out past the cap either.
The download scenario now checks the behavior .pipe lacked: leaving a code
download closes the upstream connection right away.
* fix: Keep a Late Response Write Error From Going Unhandled
* test: Key Slow Code Downloads by File and Send a Browser User Agent
* fix: Read Speech Up to the Cap Instead of Refusing It
Manual speech now synthesizes the text up to speech.tts.maxInputChars, the
same as automatic playback, through capSpeechInput, so the CJS service only
passes the capped text along and the request keeps its existing error
contract. A streamed message that reaches the cap exactly now finishes, and
a segment that fails before any audio went out leaves the response open for
the caller's HTTP error.
* fix: Count Only Text Against the Speech Cap
* fix: Hash Off the Event Loop and Surface API Key Lookup Failures
Registration, email verification, password reset and verification resends
hashed and compared with bcryptjs's synchronous calls. bcryptjs is pure
JavaScript, so each call blocked the event loop for every request in the
process. They now use the promise APIs, and the password reset
dependencies are async.
validateAgentApiKey caught its own query failures and returned null, so a
database outage answered remote callers with 401 Invalid API key. It now
lets the failure reach requireApiKeyAuth, which answers with a 500.
* fix: Run bcrypt on a Worker Pool in packages/api
bcryptjs's promise API still computes on the main thread, only in chunks
of up to 100 ms, so concurrent registrations and resets kept stalling
unrelated requests (four cost-10 hashes delayed a 5 ms timer by about
190 ms). hashSecret and compareSecret now run bcryptjs on a small
worker_threads pool in packages/api, and AuthService.js calls them.
* feat: Configure the Password Hashing Worker Count
* fix: Verify Login and Step-Up Passwords on the Hash Pool
* fix: Skip Re-Embedding Legacy File Search Files an Agent Already Holds
Files embedded before per-namespace tracking carry no embeddedEntities, so
lazy provisioning treated them as unembedded and embedded them a second
time on the first search after upgrading, duplicating every passage.
A legacy record in the agent's own file_search resources was embedded for
it, so trust its embedded flag there instead of re-queueing it.
* fix: Adopt Legacy Search Embeddings Only for Their Sole Holder
Trusting the embedded flag for every legacy File Search file left an agent
that inherited one querying a namespace without its vectors. Legacy files
held by no other agent now record the active agent as their namespace, a
one-time write, and are not embedded again; a file another agent also
holds keeps the per-namespace re-embed.
Adds an e2e scenario that reshapes an uploaded agent file into the v0.8.7
record and checks that search queries it without a second embed.
* test: Cover Legacy Search Files Shared With Another Agent
* test: Read the Stored File Id in Legacy Search Scenarios
* fix: Pin Legacy Search Adoption to Loaded Records and Wire Every Agent Init
file_id is not unique, so adoption now matches each record by its owner and
tenant as well, and only rewrites the records priming loaded. The added
conversation and client agent initializations receive the dependency too,
so their sole-holder legacy files are not embedded a second time.
* fix: Adopt Each Legacy Search Record in One Conditional Write
* fix: Count a Concurrent Same-Agent Adoption as Adopted
* fix: Run Legacy Search Adoption Beside the Liveness Probe and Fall Back on Failure
* fix: Keep Committed Legacy Search Adoptions When a Sibling Write Fails
* fix: Confirm Legacy Search Embeddings With RAG Before Adopting Them
Inferring the embedding agent from current File Search holders could adopt a
file for an agent whose namespace held none of its vectors, for example after
the agent that embedded it was deleted. Priming now asks RAG for one chunk of
each legacy file under the agent's entity_id and records the namespace only
on a hit; a miss or an unreachable RAG keeps the file queued for embedding.
The data-schemas method no longer scans agents and matches untenanted records
as untenanted. The fake RAG server answers these presence checks from the
embeds it stores.
* fix: Require the Entity's Own Chunks and Keep Confirmed Vectors
RAG also returns chunks owned by the requesting user, so a presence check now
counts only chunks recorded under the agent itself. Once vectors are
confirmed, the turn skips the re-embed even if recording the namespace fails.
* fix: Bound the Legacy Search Check and Keep the Fake RAG Query Shape
The RAG presence check now times out (fileConfig.legacySearchProbeTimeoutMs,
5s by default) and falls back to embedding, so a stalled RAG cannot hang
priming. The fake RAG server records presence checks in their own list and
keeps the file_search query records it already returned.
* fix: Check Any Unrecorded Namespace Before Embedding a Search File
A file another agent embedded first lists only that agent, so the agent
whose older vectors went unrecorded skipped the check and embedded a second
copy. Priming now asks RAG for every agent File Search file that records no
embedding for the active agent, and recording adds the namespace to the ones
already there.
* fix: Ask RAG for Several Chunks So a User Copy Cannot Hide the Agent's
RAG ranks chunks across every owner of a file_id and, once the top one is
authorized, returns them all. A copy of an agent file embedded for the
requesting user could rank first with k: 1 and hide the agent's own chunks,
so the presence check now asks for ten and checks every owner.
* fix: Confirm Setup Embeddings in Shared-File Runs and Log Safely
The run-file host primes an agent's setup files on its own, so it now
receives the RAG presence check and the namespace recorder too; without
them a shared-file run could embed a legacy setup file a second time.
Failed checks are logged with getSafeErrorMetadata rather than the raw
error, which carries the vector store's response body.
* chore: Retire the DALL-E 3 Tool
OpenAI shut down dall-e-3 on May 12, 2026, so the built-in DALL-E 3 tool can no longer generate images on api.openai.com. Remove the tool, its manifest entry, registry definition, and env docs, along with the loadTools functions flag that only existed to register it. The OpenAI Image Tools cover the same use case. Stored agents that still list dalle load without it, and dalle stays in imageGenTools so past images keep rendering.
* test: Cover the Retired DALL-E 3 Tool in Mock Scenarios
* docs: Drop DALL-E from the README Image Generation List
* perf: Debounce People Picker Search
* perf: Index-Bound Word-Prefix Search for Users and Groups
Principal search and admin user search matched name, email and username
with case-insensitive regexes, which scan the whole users collection.
Users and groups now store lowercase, accent-folded token arrays derived
from each searchable field, indexed with tenantId, and searches match
anchored token prefixes. Schema middleware keeps the tokens in sync on
save, insertMany and update queries. Documents written before the tokens
existed stay findable through a fallback branch until
npm run migrate:search-tokens backfills them; startup warns while any
remain.
* perf: Build Search Token Indexes in the Backfill Migration
* fix: Reconcile Stale Search Tokens and Close Write-Path Gaps
The backfill now compares every document's tokens with its source fields and
repairs stale ones, so writes from an older server during a rolling deploy are
recoverable by rerunning it. Undefined assignments no longer clear tokens,
upsert inserts initialize every token field, every query word is required,
relevance scoring folds accents like the filter does, and the startup check
counts both collections in parallel and owns its error logging.
* perf: Drop the Second People Picker Debounce
SearchPicker already debounces query changes by 500ms; the extra 300ms delay
in UnifiedPeopleSearch stacked on top of it.
* fix: Match Nothing for Queries Over the Word Cap
Dropping the query word cap let one request build an unbounded filter. Restore
a 16 word cap, but a longer query now matches nothing instead of ignoring its
extra words, so a cap never produces a false positive.
* test: Cover Word-Prefix People Search and the Token Backfill End to End
* fix: Tokenize Fields an Upsert Copies From Its Filter
MongoDB copies equality conditions from an upsert's filter into the inserted
document, so their token fields now derive from that value instead of being
initialized empty.
* fix: Rank Accent-Folded User Matches Like the Search Filter
* test: Run the Token Backfill Without Blocking the Playwright Worker
* perf: Index Search Tokens Tenant First for Scoped Searches and Ignore Undefined Unsets
* test: Skip the Module-Scope Index Rebuild When the Scenario Runs the Backfill
* fix: Exit the Search Token Migration When Redis Clients Stay Open
* perf: Probe for Missing Search Tokens at Startup Instead of Counting Them
* test: Explain the Startup Probe the Warning Actually Issues
* fix: Keep Search Token Fields Out of Serialized Users and Groups
* fix: Keep the Admin Search Legacy Fallback Anchored at the Field Start
* fix: Score Groups Only on the Fields Search Matches
* perf: Start the Search Token Probe Alongside the Permission Checks
* fix: Warn at Startup When the Search Token Indexes Were Never Built
* fix: Score People Search Matches With the Filter's Word Separators
* perf: Check the Token Indexes Before Probing for Unmigrated Documents
* test: Drive the Failed Startup Probe Past the Index Check
* fix: Score People Search Matches Independently of Word Order
* refactor: Remove Unused Agent Helpers and Superseded Wrappers
* refactor: Remove Dead Auth, Middleware, Protection and Prompt Helpers
* refactor: Remove Dead MCP, File, Skill and Endpoint Code
* refactor: Remove Dead Stream, Telemetry, Utility and Type Exports
* chore: Drop Unused packages/api Dependencies From bun.lock
* refactor: Keep Published Compatibility and Deprecated packages/api Exports
* fix: Mask Email Addresses and Drop Personal Identifiers from Logs
* fix: Log User IDs in OpenID Admin Role and LDAP Domain Block Messages
* fix: Mask Unicode Email Addresses in Log Redaction
* test: Add Log Email Masking Scenarios
* fix: Bound the Email Redaction Matcher and Handle Quoted Local Parts
* fix: Mask Quoted and Special-Character Email Local Parts
* fix: Redact Whole Email Tokens, Top-Level Log Fields and Metadata Keys
* fix: Redact Top-Level Metadata Keys and Escaped Quotes in Email Local Parts
* fix: Redact Truncated Emails, Address Literals and Request Paths in Logs
* fix: Redact the Full Log String Before Truncating
* fix: Preserve Colliding and Object Keys When Redacting Log Metadata
* fix: Bound Log Redaction to a Window Past the Truncation Limit
* fix: Redact the Truncation Window Without a Whitespace Cut
* fix: Redact Serialized Quoted Addresses and Keep Key Collisions Linear
* fix: Bound Every Log Redaction Path and the URL Credential Pattern
* fix: Keep Long-Scheme Credential Redaction and Serialized Escaped Quotes
* test: Cover Every Secret Shape Across the Log Truncation Cap
* fix: Drop Unfinished Window Tokens, Mask Short TLDs and Redact Direct Error Stacks
* fix: Treat Input End as a Token Boundary and Mask Parenthesized Identifiers
* fix: Remove the Quoted Local Part Length Cap Without Quadratic Scans
* fix: Drop an Unfinished Quoted Address at the Window Edge
* fix: Check Quoted Locals Before Accepting a Window Delimiter
* fix: Count Only Opening Quotes When Trimming the Window
* fix: Track Unmatched Quoted Locals Instead of Counting Closing Quotes
* fix: Lift Residual Email Caps, Keep Punctuated Credential URIs Masked and Move OBO Log Id to the API Package
* fix: Mask Empty Quoted Locals and Tolerate Read-Only Log Fields
* fix: Return a Sanitized Copy When a Log Field Cannot Be Replaced
* fix: Treat a Failed Key Removal as a Failed Replacement and Guard Stack Writes
* fix: Skip Heavy Error Graphs and Accept Escaped Backslashes in Serialized Locals
* fix: Accept Lone Escaped Backslashes in Serialized Locals Without Rescans
* fix: Compact Axios Request Info Before Redaction and Skip Heavy Splat Error Graphs
* fix: Redact Address Keys on Nested Error Objects
* fix: Materialize and Redact Enumerable Getters on Nested Errors
* fix: Guard Throwing Top-Level Getters and Skip Heavy Fields Only on Error Records
* fix: Skip Serialized Quotes When Trimming an Unfinished Address
* fix: Bound Redaction With a Per-Record Node Budget and Accept JSON Control Escapes
* fix: Match Credential URIs From the Separator So Scheme Length Is Unbounded
* fix: Count Strings Against the Redaction Budget and Guard Nested Getters
* fix: Rebuild the Log Record When a Top-Level Accessor Ignores Writes
* fix: Rebuild the Log Record When a Stack Accessor Ignores Writes
* fix: Guard the Initial Stack Read on Direct Error Logs
* fix: Accept Escaped Solidus and Preserve the Outer Budget on Reentrant Logging
* fix: Guard the Axios Marker and Error-Like Probes Against Throwing Getters
* fix: Guard the Axios Marker on Nested Errors
* fix: Replace Skipped Heavy Fields Up Front, Budget Top-Level Fields and Mask Encoded Emails
* fix: Mask Encoded Address Keys and Guard Nested Axios Config Compaction
* fix: Redact URL Objects in Compact Request Info
* fix: Remove Encoded Email Caps and Strip Serializer Hooks From Clones
* fix: Strip Serializer Hooks From Error Clones and Top-Level Records
* fix: Neutralize Inherited Serializer Hooks on Log Records
* fix: Guard the Serializer Hook Probe
* fix: Replace Function Values in Log Records
* fix: Mask Percent-Encoded Domain Separators in Logged URLs
* fix: Bound the Encoded Email Domain at URL Delimiters
* fix: Shadow Throwing Serializer Getters on Nested Errors
* fix: Rebuild Accessor Messages and Mask Doubly Encoded Addresses
* fix: Redact Multiply Encoded Address Keys
* feat: Make the Log Redaction Node Budget Configurable
* fix: Guard Message and Serializer Property Reads
* fix: Contain Proxy Failures and Redact the Compact Request Method
* feat: Audit Resource ACL Changes and Fail Closed by Default
* fix: Keep Non-Reversible Admin Audit Handlers Fail-Open
* docs: Narrow AUDIT_LOG_FAIL_CLOSED Description
* fix: Audit and Restore ACL and Insights Changes in One Pass
* fix: Break Audit and Insights Permission Type Cycle
* test: Cover Resource ACL Audit Scenarios
* refactor: Audit Resource Permission Changes Before Writing
* test: Align Concurrent Revoke Audit Expectation
* fix: Skip Deleting ACL Entries That Were Not Audited
* fix: Read ACL Snapshots Through the Write Session
* fix: Never Block Revocations on Audit Outages and Normalize Principal Keys
* fix: Let an Explicit Public Disable Win Over a Public Update
* fix: Audit Duplicate ACL Rows as Combined Permissions
* fix: Audit Permission Changes Under One Target and Dedupe Revokes
* fix: Audit Capability Grants Before Writing Them
* feat: Serialize Resource Permission Updates With a Per-Resource Lease
* fix: Fence the Resource Lease Before Writes and Always Audit Grant Assignments
* fix: Renew the Resource Lease Before Each ACL Write Phase
* fix: Serialize System Grant Assignment and Revocation With a Per-Grant Lease
* fix: Keep Leases Alive During Audits, Bound Lock Index Wait, Canonicalize Audit Target IDs
* fix: Retry Deferred Revoke Audits Under the Lease and Audit Canonical Principal and Role IDs
* test: Expect Role IDs in Permission Audit Metadata
* fix: Read Mutation Snapshots From the Primary Outside Transactions
* fix: Flush Deferred Audits Before Cache Invalidation, Keep Renewals Monotonic, Keep Role Case in Grant Lease Keys
* refactor: Map the Busy Permission Error to Its Code in the API Package
* style: Sort Lease Imports and Type the Renewal Spec Read
* fix: Keep Success Audits After Commit, Report Downgrades as Revocations, Code Audit Outages
* refactor: Move the Audit Outage Code to Its Own Module
* fix: Translate Lock Failures, Lease Non-Transactional Sessions, and Code Grant Audit Outages
* fix: Sanitize Lease Renewal Failures at the Permission Boundary
* fix: Annotate the Unavailable Error Code for Isolated Declarations
* fix: Keep the Admin Route Fail-Closed Opt-In and Document Lease Clock Skew Safety
* fix: Translate Grant Lease Store Failures to a Retryable 503
* fix: Log Failed Grant Audit Writes With Safe Metadata Only
* test: Detect Raw Error Arguments in the Grant Audit Log Spec
* feat: Make Permission Lease Limits Configurable and Localize Retryable Sharing Failures
* fix: Bound Lock Index Readiness by the Lease Wait and Carry the Resource ID in Every Audit Entry
* test: Cover Lease Config Reaching the Leases and the Sharing Dialog Failure Messages
* fix: Load App Config on the Sharing and Grant Routes and Log Failed Lease Releases
* perf: Load App Config Only on the Grant Mutation Routes
* fix: Expire Dead-Lettered Trigger Deliveries After a Review Window
* fix: Delete Stored Image Uploads When Record Creation Fails
* fix: Clean Up Stored Objects on Every Storage-Before-Record Upload Path
* refactor: Resolve the Delete Strategy Lazily in Stored Upload Cleanup
* test: Cover Dead Letter Expiry After the Review Window
* fix: Carry Owner to Stored Upload Cleanup and Keep Batch Dead Letter Deadlines
* refactor: Select the Stored Upload Cleanup Target in the API Package
* fix: Check for a Committed Record Before Deleting an Upload and Keep Dead Letters Unless Configured
* fix: Check the Persisted Record Id for Converted Assistants Images
* fix: Arm a Dead Batch Root Deadline After Its Members Settle
* fix: Settle Expired Dead Batches and Skip the Record Guard Before the File Write
* fix: Skip the Record Guard for Pre-Write Failures and Build the Lookup in packages/api
* fix: Build the Record Lookup Lazily in the Upload Service
* fix: Pass the Dead-Letter Retention Window Through Experimental Startup
* fix: Carry the Vector Copy Into Failed-Upload Cleanup
* fix: Cover the Code Environment and Indeterminate Vector Copies in Upload Cleanup
* fix: Give URL-Upload Cleanup a Request When the Caller Omits One
* fix: Route Failed-Upload Cleanup Through Secondary-Storage Deletion
* fix: Attempt Every Copy Independently When Cleaning Up a Failed Upload
* fix: Track the Record Write in Every Upload Path Whose Retention Lookup Can Fail
* fix: Enforce Terms Acceptance on the Server When modalAcceptance Is On
* fix: Mark Backfilled Terms Acceptance Timestamps
* test: Cover Server-Enforced Terms Acceptance in the Mock Harness
* fix: Gate Management Routes and Skill Sync and Resolve Terms Policy Fail-Closed
* fix: Evict Cached Auth Documents When Terms Acceptance Is Reset
* fix: Guard Skill Sync Behind Terms Acceptance for Every Skill List Route
* fix: Move Terms Skill-Sync Guard Into the API Package and Evict All Users on Reset
* fix: Batch Terms Reset Cache Eviction and Sign the Spec User In Fresh
* fix: Let Stop Through the Terms Gate, Refresh Startup Config, and Wait Out Failed Eviction
* refactor: Narrow the Terms Exemptions and Share One Refusal Path Across Clients
* refactor: Compose the JWT Terms Gate Through an API Package Helper
* style: Sort Imports for the Shared Terms Refusal Helper
* test: Expose the Terms Gate Helper in the JWT Auth Spec Mock
* test: Cover the Terms Refusal Code on Pre-Stream Upload Errors
* fix: Boot-Paint a Cached Theme Only for the Tab's Signed-In Owner
The boot script and DeploymentTheme's first commit replayed the deployment-theme cache without checking its owner, so a browser that last held one identity's theme painted it for another until the config answered. The tab now records its signed-in owner in session storage and both paths replay the cache only when it matches.
* test: Assert Session Ends Forget the Tab Theme Owner
* test: Hold the First-Paint Config Answer Until the App Paints
* fix: Skip the Boot-Painted Theme Cache Once the Session Has Ended
* fix: Bind the Readable Session Hint to the Signed-In Owner
* test: End the Session for Real in the Expired-Hint Scenarios
* fix: Clear the Session Hint When a Local Refresh Finds No Session
* test: Give the Refresh Specs the Cookie Surface the Session Hint Uses
* test: Count the Signed-Out Config Answer in the No-Cached-Paint Scenarios
* fix: Retire the Session Hint on Every Rejected OpenID Refresh
* fix: Retire the Session Hint When a Refresh Carries No Token
* fix: Retire the Session Hint on Account Deletion and Password Reset
* refactor: Own the Session Hint Lifecycle in the Typed Auth Cookies
* fix: Write User Provider Keys With Random-IV v3 Format
* fix: Reject Undelimited Input in decryptV2 Instead of Returning It
* fix: Default 2FA Backup Codes to the Strong Format
* fix: Limit Admin Secret Preview to the Last Four Characters
* fix: Re-Derive Stored Secret Previews Under the Last-Four Policy
* test: Build Legacy Credential Fixtures With the Server Key
* fix: Load Request Config for User Key Writes and Mask Legacy Previews
* fix: Encrypt User Plugin Credentials With Random-IV v3 Format
* test: Cover Credential Encryption Formats in Mock E2E Scenarios
* test: Round-Trip Real Credential Helpers in PluginService Spec
* 🙈 fix: Remove SharePoint Picker Payload Logging (#16997)
* fix: Add credentials.writeFormat to Gate v3 Credential Writes During Rollouts
* test: Expect 201 When the Legacy Write Format Override Is Created
* 🔊 fix: Clear TTS Audio Cache On Session End and Hash Its Keys (#16998)
* fix: Clear TTS Audio Cache On Session End and Hash Its Keys
* fix: Read Manual TTS Replays From the Hashed Cache Key
* fix: Purge Legacy TTS Cache Keys and Fence Writes Across Logout
* fix: Keep TTS Cache Writes Off the Playback Path
* test: Expect Hashed Keys in the StreamAudio Cache Lookup
* test: Use a Single-Token Reply Label in the TTS Cache Scenario
* fix: Settle TTS Cache Clear Before IdP Logout and Across Tabs
* test: Cover the TTS Cache Fence When Another Tab Logs Out
* test: Wait for the Second Tab Before Logging Out in the TTS Scenario
* fix: Latch TTS Cache Writes Shut Until the Tab Authenticates Again
* fix: Reopen TTS Cache Writes When Authentication Completes In Place
* fix: Leave for the IdP Even When TTS Cache Cleanup Fails
* fix: Retry a Failed Legacy TTS Purge and Fence Other Tabs on Manual Clear
* fix: Rerun the Legacy TTS Purge After Each Clear and Sign-In
* fix: Rerun the Legacy TTS Purge When Another Tab Clears
* fix: Fence Pre-Auth TTS Writes, Retry Failed Clears, and Fall Back to Storage Events
* fix: Version the Hashed TTS Cache Key Path
* fix: Move TTS Audio to a New Cache and Keep It Closed Until a Failed Clear Succeeds
* fix: Persist a Pending TTS Cache Delete Across Navigation
* fix: Reopen TTS Caching Only on Session Start and Fence Cache Reads
* fix: Keep the TTS Delete Fence in Cache Storage and Close Reads After Logout
* refactor: Keep Read-Aloud Audio in Tab Memory Instead of Cache Storage
* test: Return to the Conversation After Re-Login in the TTS Logout Scenario
* test: Close the Phone Sidebar Before Replaying in the TTS Logout Scenario
* fix: Bound the In-Memory TTS Cache by Size as Well as Count
* fix: Scope Cached TTS Audio to the Signed-In User
* fix: Keep the Settings Cache Delete Available While Legacy Audio Remains
* fix: Recheck Legacy TTS Audio After a Settings Clear and on Unowned Handoff
* fix: Fence TTS Writes Started Before an Account Signs In
* test: Opt the Fresh TTS Account Into Enter-to-Send
* feat: Look Up Bitbucket Cloud Pull Requests
* fix: Keep the Pull Request Host Out of Overrides and Per-Host Keys
A role, group, user or tenant override can no longer change or remove pullRequests.host, so feature gating and link recording see the operator's host. Links are recorded only from the configured host, and the batch share key and the rate limit cooldown are scoped by host.
* fix: Compare Pull Request Link Hosts Through the Host Resolver
* test: Add Pull Request Scenarios Against a Fake GitHub Enterprise Host
* fix: Keep the Pull Request Provider Out of Overrides and the Batch Share Key
A role, group, user or tenant override can no longer change pullRequests.provider, so feature gating and link recording agree with the lookup, and a batch lookup for one provider is never joined by a request for another.
* fix: Read GitLab History Apart From Open Merge Requests
Merged and closed merge requests are listed on their own, so open fork merge requests on the branch name cannot crowd them out. Rate limit cooldowns are scoped by provider, and a workspace repository with a dot segment is rejected.
* fix: Revalidate Bitbucket Pull Requests and Skip Ones Without a Source Commit
Read a full closed page so a newer closed fork cannot hide an older pull request from the repository, recheck the source branch and fork status on the detail, and treat a pull request without a source commit as unmatchable instead of failing the lookup.
* fix: Follow Bitbucket's Next Link When Listing Pull Requests
* fix: Record the Lane Pull Request Without an Aggregation Pipeline Update
* fix: Match an Empty Lane Repository as Unknown
* fix: List Locked GitLab Merge Requests With the Open Ones
* fix: Replace an Expiring Fake Git Host Certificate
* feat: Show the Conversation Pull Request in the Sidebar List
Add a batch lookup route and a sidebar row mark: the state icon with the CI dot, opening the same card beside the row on hover. The row keeps the conversation's own title.
* fix: Address Review Findings on the Sidebar Pull Request Batch and Row Mark
Cap the whole batch with a configurable deadline, make the row mark a keyboard-reachable button that opens the card on focus, refresh rows when the window regains focus through the shared batch, and show a retryable failure state instead of treating a failed first lookup as no pull request.
* fix: Send Sidebar Pull Request Batches One at a Time
Queue every batch request behind the one before it, so an oversized list or an overlapping flush cannot multiply the server's configured lookup concurrency, and bound how long one request may hold the queue so a request that never answers cannot stop the rest.
* fix: Keep the Batch Request Timeout Above the Server Deadline
Default the client's request timeout to just past the longest batchTimeoutSeconds a deployment may configure, so a queued request never starts while the previous request's server-side lookups are still running.
* fix: Keep Batch Lookups Inside the Concurrency Limit and Per-Entry Failures Per Entry
Share one lookup limiter across every batch request so lookups that outlive a request's deadline keep their slot, answer a missing token only for the conversations that needed it, and point the row at its pull request description only while that text is in the page.
* fix: Bound Batch Reads, Scope Concurrency Per Credential, and Gate the Sidebar on the Batch Version
Start the batch deadline when the request arrives and bound the config and lane reads, count lookup slots per credential, advertise a batch route version and ask only a server that has it, take no row space for an absent mark, show a failed refresh over a cached empty answer, keep a malformed response from stalling the request queue, and repair the lane recording tests that still used a mock removed on dev.
* fix: Move Capability Resolution Into the Package and Harden the Batch Route and Queue
Resolve the pull request startup capabilities in packages/api so the route stays wiring only, bound the config and lane reads by the configured deadline, merge identical lanes before taking limiter slots, fall back to the single route when a replica has no batch route, and dispose the batch queue when the signed-in user changes.
* fix: Abort the Batch Fallback With Its Request So the Queue Never Overlaps It
Give the batch fetcher the request's abort signal and pass it to every single-route call, abort a request that times out or is disposed, and stop the fallback workers from starting more calls once it has.
* fix: Forward the Abort Signal to the Batch POST
Pass the request's abort signal through the batch fetcher, the data-provider call and Axios, so a batch request that times out or is disposed is cancelled instead of continuing beside the next one, and settle a request promptly when it is aborted.
* fix: Close the Pull Request Lookup Follow-Ups and Cancel Abandoned Sidebar Rows
* 🏝️ fix: Apply the No-Workspace Decision Wherever an Agent's Code Route Resolves
A conversation set to run without attached workspaces failed mid-turn with
"Choose an attached workspace before using this agent." when its agent's
default machine was attached and the turn loaded a skill or another tool at
execution time. Request-start initialization honoured the decision, but the
execution-time tool loader re-resolved the agent's attached default and then
demanded a workspace selection the conversation intentionally never made.
The per-agent code rule now lives in one typed function,
resolveAgentCodeExecution (packages/api/src/code/agent.ts): it applies the
conversation's no-workspace opt-out, the run-code grant, the stateful
capability and the stateful environment policy, then resolves the base route.
Request-start initialization, lazy subagent metadata and graph members, the
execution-time and definition tool loaders, and the legacy execute_code
loader all go through it, so an opted-out agent lands on the managed default
route with no code tools at every site. Per-call subagent machine choices
offer and accept no attached machine in such a conversation, and a source
invariant test keeps new resolver call sites on the shared rule.
* fix: Pass the Implicit Stateful Route Rollout Into the Shared Code Rule
The per-agent code rule read the implicit stateful route rollout from the environment itself. It now takes it as a parameter, and the request boundaries (initializeAgent and withRequestCodeInputs) resolve it from the deployment configuration, so the rule stays a pure function of its inputs.
* fix: Carry the Initializer's Code Opt-Out Into Tool Definition Loading
initializeAgent hands the tool loaders a partial agent record (id, tools, provider, model) together with the code route it already resolved. Without the agent's route fields the definition loaders could not see a no-workspace opt-out and recomputed code execution as enabled, so tool classification could advertise the programmatic bash tool that execution then refuses. initializeAgent now passes attachedEnvironmentOptOut through every tool loader, and the shared rule keeps an opt-out its caller already decided.
* feat: Record Lane Git State and Serve Conversation Pull Requests
A conversation now stores the branch, head and repository its code lane last reported, written only by the server after a finished bash command and excluded from ordinary reads, generic saves, unset fields and imports. GET /api/convos/:conversationId/pull-request finds the pull request for that branch through an injected GitHub source with a bounded cache, answers null for an absence, and maps failures to stable codes without upstream text. The feature is off unless endpoints.agents.pullRequests is enabled with a token reference.
* fix: Address Review Findings on Pull Request Lookup and Lane Writes
Scope the lookup cache, shared in-flight request and rate limit cooldown to the credential, so one tenant's token never serves another. Ask for open pull requests on their own before falling back to the newest closed one, read every page of check runs up to a bound and never report an incomplete rollup as passing, order lane writes per conversation so an older write cannot land last, and name both required token permissions in the example config.
* fix: Address Second Round Review Findings on Pull Request Lookup
Clear the stored lane when the conversation moves to another workspace or leaves attached execution, so the previous workspace's pull request is never served. Fence the lane write on the time it was reported, so an older report from another replica cannot replace a newer one. Honor GitHub's reset window up to an hour instead of ten minutes, and expose the request timeout, whole lookup timeout and check run page limit as endpoints.agents.pullRequests settings with the previous values as defaults.
* fix: Authorize Pull Request Repositories and Fence Lane Writes
Require an administrator allowlist of repositories before the server token is used, since a worker reports its own repository. Match a pull request to the commit the conversation last ran at, so a reused branch name does not return another pull request. Fence the lane write on the workspace the command ran in, so a write queued before a move cannot bring the old lane back, and record a subagent thread's lane on the visible root conversation. Key the lookup cache and shared request by the recorded commit and the lookup policy, and document the worker flag in the example config.
* fix: Order Lane Reports by a Database Sequence and Tighten Head Matching
Replace wall clock ordering of lane reports with a per conversation counter reserved in the database when the command settles, so replica clock skew cannot reject a newer report. Treat a pull request whose head is behind the recorded commit as no match, and expose the candidate and comparison bounds of the history search as settings. Read the app config and the owner scoped lane together, inject the HTTP transport from the route instead of falling back to a global, skip a report identical to the last one written, and name the Contents permission that commit comparison needs.
* fix: Queue Lane Reports by the Visible Conversation and Read Checks for the Current Head
Order lane report reservations by the conversation the write lands on, so two subagent threads of one conversation share a chain, and write every report instead of skipping a repeat, since another recorder may have changed the lane in between. Read a pull request's detail before its check runs and use the head it reports, so line counts, mergeability and checks describe one commit. Treat an expired temporary chat as gone when reserving, writing and reading the lane, and make the shared cache capacity a setting.
* fix: Fence Lane Reports by Attachment Epoch and Place Recorders Before Reporting
Strip dotted descendants of lane fields from saves and imports, partition the pull request cache per credential, resolve the lane write target when the tool is created, and fence reports with a codeAttachmentEpoch that advances only on an owner move or detach.
* fix: Fence Subagent Lane Reports by the Root Epoch and Cover Forks and Moved Heads
Read the visible root's attachment epoch for a subagent report, gate lane recording on pullRequests.enabled, make the credential partition bound a config field, revalidate the detail head against the recorded commit, and find fork pull requests through the recorded commit.
* fix: Keep the Credential Cache Bound Deployment Wide and Budget Head Revalidation
Apply the largest configured credential bound to the shared cache, count cache hits as use, spend the comparison budget on a moved head, and fix the data-provider config default test.
* fix: Cool Down Headerless Secondary Rate Limits
A 403 whose message names a secondary rate limit now starts the credential cooldown for at least a minute instead of being cached as a ten second failure.
* fix: Keep Reasoning Out of Every Read Aloud Path
Read aloud spoke the Thoughts section along with the answer. The hover
button flattened think parts into the string it handed to TTS, browser
autoplay and the hooks parsed content without skipping reasoning, and
server streaming spoke the stored text field, which an aborted turn
persists with its reasoning included.
Add getSpeechText to data-provider as the single source for spoken text:
text parts only for structured content, and the legacy :::thinking block
stripped from plain text, treating a still-open block as reasoning so
streamed speech only ever grows. Route the hover button, browser
autoplay and server streaming through it, and skip reasoning in the TTS
hooks' own content parsing.
* fix: Speak String Message Content Without Its Reasoning
Imported or client-built messages can carry their answer as string
content rather than parts; read it like plain text instead of falling
back to an empty text field.
* fix: Keep Answers Between Repeated Legacy Thinking Blocks
Strip every closed :::thinking block before treating an unterminated one
as reasoning, so a second closed block no longer swallows the answer
that follows it.
* fix: Skip Read Aloud and Autoplay When Only Reasoning Remains
A turn stopped mid-reasoning has no answer to speak once reasoning is
filtered. Gate autoplay on the same speech text it will play, so the
external engine never opens a stream that can only time out, and hide
the Read aloud action when there is nothing to read. Every speech path
reads a finalized message, so speak an unmatched :::thinking marker as
the answer text the UI shows instead of dropping the rest.
* refactor: Move Speech Chunk Polling Into packages/api
The decision of what text to stream as speech lived in the legacy CJS
service. Move the chunk processor to packages/api as
createSpeechChunkProcessor, taking the message cache and lookup from the
caller, and leave streamAudio.js wiring them in. Its tests move with it
onto an in-memory cache instead of module mocks.
* fix: Speak Exactly the Answer the Message View Shows
Share the legacy :::thinking parser between the message view and speech,
so Read Aloud strips the same block the view collapses instead of a
separate pattern drifting from it. Trim structured speech text so a
whitespace-only answer is not speakable, and key autoplay's audio cache
by the spoken text, as manual playback already is, so audio cached with
reasoning before this change is never replayed.
* fix: Fall Back to Text When String Content Is Empty
* 🧭 feat: Let a Parent Choose Its Subagent's Code Machine Per Call
A parent agent can now pass optional `machine` / `workspace` arguments on a
subagent call to pick where the child runs, using the host-argument
extension point added to @librechat/agents.
- Each lazy child declares enums built only from machines it may reach:
admitted in the conversation's sealed codeWorkspaces, on its default or
allowlist, in the requesting principal's tenant-scoped list, and served by
a ready worker. A child the conversation pinned to a machine offers only
that machine; a child with no reachable machine declares nothing.
- On use, a request-scoped routing session in packages/api re-validates the
choice against the current request and live worker status and refuses an
invalid one with a typed, model-visible error instead of falling back. A
routed child is pinned to the chosen machine at every routing site,
overriding parent inheritance.
- Omitted arguments keep inherit-by-default routing, extended so a child
follows the machine a per-call-routed parent runs on (passing through
agents without code). A subagent keeps one machine for the whole request.
- A child whose default workspace is unavailable stays guarded unless it has
per-call choices; then it stays callable and its description asks the
parent to pick a listed machine.
- Routed children keep their tool context per execution for tool loading,
provisioning and artifact handling.
- Paused-approval bindings, the attached-machine policy hook, skill-file
priming and tool-output references account for per-call choices.
The declarations are gated on the installed SDK exporting
SubagentHostArgumentError, so this change is inert until LibreChat depends on
an @librechat/agents release that includes it.
* 🧭 fix: Reserve Default Routes Early and Route a Routed Subagent's Graph Members
- Reserve a subagent's default machine when it is placed, as
initializeAgent will resolve it (including parent inheritance), so a
concurrent call cannot claim another machine during initialization.
- Graph members spawned by a per-call-routed subagent follow its machine
when they may, unless they already initialized on their own route this
request.
* 🧭 fix: Settle Concurrent Inherited Routes and Register Routed Graph Members
- Re-read a subagent's inherited route after resolving targets, so two
concurrent calls from differently routed parents settle on one machine.
- Attribute a workspace-only conflict to the workspace argument.
- Register graph members a routed subagent initialized, so each gets the
attached-machine policy of the machine it actually runs on.
* 🧭 fix: Re-read Inherited Routes After Resolving Subagent Targets
- Re-read a subagent's inherited route after the target resolution await,
so two concurrent omitted-machine calls from differently routed parents
settle on one machine.
- Attribute a workspace-only conflict to the workspace argument.
* 🧪 test: Prove a Routed Graph Member Takes Its Machine's Policy Through createRun
- Register a resolved child's graph members inline in the lazy resolver
instead of through an exported generic helper whose test did not
typecheck.
- Drive the check through createRun's real PreToolUse hook: a member
routed to a deny machine flips from its metadata machine's allow to deny
once its parent resolves.
* 🧭 fix: Hold a Lazily Loaded Graph Member's Default Machine While It Initializes
- Pass a graph member's code flags when it initializes on its default
route, so the routing session reserves that machine before the
asynchronous initialization instead of only after it, and a concurrent
call cannot place the same agent on another machine meanwhile.
* 🧭 fix: Validate the Approval Mode Against Every Machine a Subagent May Use
- Collect each agent's default attached route and each attached machine a
parent may route it to per call, and validate the requested approval
mode against that set in createRun and both client checks.
- A child whose default is unavailable but whose per-call machine permits
accept edits no longer fails run creation; full access still requires
every such machine to permit it.
* 🧭 fix: Admit, Prime, and Cancel Per-Call Subagent Routes Conservatively
- Build scheduled pause admission from every attached machine a parent
may route an agent to, so a permissive default does not hide a
per-call machine that asks.
- Walk lazy graph-member metadata when collecting execution routes, so
historical skill files are primed for every machine such a member may
run on.
- Stop placing a canceled call after its target resolution resolves, so
it never claims a machine for an execution that will not initialize.
* 🧭 fix: Release Unused Route Claims and Scope Per-Call Machines to Their Run
- Hold a subagent's route claim per placement: a failed or canceled
initialization gives the machine back unless another execution still
holds it, and a claim is permanent once a child on it attaches.
- Fold an agent's default route into its per-call choice digest, so each
agent contributes at most one paused-approval binding target.
- Use only an agent's route policies for pause admission when it has
them, so a missing default does not make a decided route look asking.
- Drop a routable child's provisional attached-machine membership when it
resolves off attached machines.
- Offer and accept no attached machine per call while run files are
shared, since shared children must run on managed code.
* 🧭 fix: Give Back a Route When Its Placement Verification Fails
- Release a placement's hold when attach finds the child initialized on a
different machine than the one it was routed to, so a retry that
follows the unavailable-machine guidance may name another machine.
* 🧭 fix: Commit a Lazy Child's Routes Only When Its Whole Resolution Succeeds
- Keep the routes a lazy child and its routed graph members hold pending
through the child's entire resolution (settleExecution), committing them
on success and giving them back on failure or cancellation unless
another execution still holds them.
- Placements outside any call (shared graph members) still commit when
they attach.
* 🧭 fix: Drop a Failed Lazy Resolution's Config from the Graph Cache
- When a lazy child's graph resolution fails or is canceled after its
config was cached for graph reuse, remove that entry with its route, so
a later graph never reuses a config for a machine the child gave back.
* 🧭 fix: Follow a Routed Agent Through Self-Spawned Runs and Scope Routed Member Configs
- A call whose parent run was never placed (an allowSelf spawn reuses its
routed inputs) inherits the routed claim of its parent agent, which keeps
one machine per request.
- Never share a graph member initialized for a routed parent's execution
through the graph-member cache, so a failed resolution leaves no config
behind for a route it gave back.
- Race lazy target discovery with the request signal.
* 🧭 fix: Pass a Machine On Through Self-Spawned Runs of Agents Without Code
- Record the machine each agent without stateful code passes on, and let a
call from that agent's unplaced run (an allowSelf spawn) follow it when
the agent passed on exactly one machine this request.
* 🧭 fix: Keep a Pass-Through Route Only When Its Resolution Succeeds
- Record the machine an agent without code passes on with the rest of its
call's routing state: kept when the resolution deciding it succeeds,
dropped when it fails, so a self-spawn never follows a failed attempt.
* 🧭 fix: Settle a Lazy Child's Routing State at the SDK Handoff
- Wrap the run's whole lazy resolution, through the inputs handed to the
SDK, in the descriptor's settle hook, so a cancel after the host
resolver returns still gives the call's routes back.
- Roll back the per-agent tool context a pending placement set (restoring
the entry it replaced) when its resolution fails, so a retry on another
machine seeds its own context.
- Drop a failed call's shared config from the graph cache through the
same settle hook.
* 🧭 fix: Keep Per-Agent Tool Contexts on Live Placements and Queue Target Probes
- Track every pending placement that uses an agent's per-agent tool
context: a committed one sets it, a failed one hands it to a surviving
placement, and only the last failure restores the earlier entry.
- Let an unplaced run of an agent (an allowSelf spawn) defer to the routed
run of the same agent that encloses it when looking up tool contexts.
- Queue live target probes at eight in flight, below the shared status
poller's ceiling, so concurrent descriptors never misreport a healthy
machine as unavailable.
* 🧭 refactor: Load Routed Graph Members Through a packages/api Service
- Move the routed graph-member load (shared reuse, skip list, view check,
execution identity, failure policy) into createRoutedGraphMemberLoader;
the initializer only injects its dependencies.
- Thread the request and call signals into target resolution so queued
probes are skipped once canceled.
- Probe nothing in shared-file runs, where every per-call target is
attached and unusable.
- Retire pending tool-context ownership when a committed placement
installs the per-agent entry.
* 🧭 fix: Share a Lazy Child's Config with Later Graphs Only After It Commits
- Publish a default-placed lazy child's config to the shared graph cache
from the settle hook's commit, so no concurrent graph can borrow a config
whose route may still be given back.
- Resolve a child's own appearance in its graph to the child config itself.
* 🧬 fix: Default Subagents to Their Parent's Code Machine and Workspace
A subagent resolved its code machine from its own saved fields only, so
when a chat routed the parent to an allowlisted machine the subagent fell
back to its own default and whatever workspace the conversation had
selected there, or failed the parent's whole turn when none existed.
Subagents now inherit the parent's resolved attached machine when it is
their own default or on their allowlist where per-chat choice applies,
the principal can use it, the sealed decision already selected it, and no
selection names the subagent as an explicit owner. Inheritance is derived
per request from the sealed decision and current agent configuration and
fed to every routing site; it is never persisted, so existing chats and
older clients get it without a decision change. The composer mirrors it.
A subagent whose machine has no usable workspace is advertised to the
parent as unavailable instead of failing the turn; a locked decision
stays fatal.
* fix: Inherit Across Every Spawning Path, Graph Members and Deferred Code Tools
A subagent now inherits only when every agent that can spawn it, at any
depth, runs on the same machine; spawn cycles keep their own routes.
Subagent graph members inherit like explicit subagents. The deferred
execute_code loader passes inheritance to its fallback resolver. Subagent
reads and VIEW checks move to a packages/api loader shared by inheritance,
lazy descriptors and graph members. The composer lets a draft's subagent
follow its parent's machine before a workspace is picked, and keeps an
inheriting subagent's former selection without asking for a move.
* fix: Prune Spawn Back-Edges, Gate the Composer on an Inheritance Capability
Spawning parents follow the run's path semantics, so a back-edge the
descriptor tree prunes no longer blocks inheritance, and agents that spawn
each other follow their single outside parent machine when it routes them
all there. Only admitted subagents count toward the walk's node limit.
Graph members re-read and re-check VIEW when they initialize. The composer
echoes an inheriting subagent's former selection without gating on that
machine, and mirrors inheritance only when the API advertises
codeWorkspaceInheritanceVersion; approval modes keep intersecting every
machine an agent could use.
* fix: Settle Each Mutually Spawning Group on Its Own
Independent spawn cycles no longer merge into one group, so a pair under
one parent follows that parent even when another pair follows a different
machine. Also types the data-provider spec for its ES5 target.
* fix: Bound the Inheritance Walk, Normalize Parallel Roots, Move Subagent Availability to TS
The walk reads in batches no larger than the remaining node budget, stops
once the limit is reached, and runs through the run's subagent load limiter.
Roots are keyed by saved agent ID, so a parallel root such as id____1 can
never receive an inheritance entry. Resolving a lazy subagent's code route,
reporting it unavailable, and guarding its descriptor now live in
packages/api (resolveSubagentCodeAvailability, guardUnavailableSubagent).
* fix: Keep Missing Subagent Loads Parallel Near the Node Limit
Batches never shrink below eight IDs, so missing or unviewable subagents
near the limit still load in parallel under the injected limiter, and at
most a handful of reads land past the limit.
* 🔗 feat(agents): use LibreChat prompt groups as agent instructions
Agents can link a prompt group revision or Production as their instructions. The link resolves at the start of each turn through a cached resolver. The cache clears on promote and delete. Editors without VIEW on the group see a restricted placeholder.
* fix: remove duplicate logger import from agent initialization tests
* fix: honor prompt management capabilities for linked agent instructions
* test: seed prompt manager capabilities in authorization fixtures
* fix: clear acknowledged linked prompt edits before version restore
* test: adapt merged chain coverage to prompt-aware payload helper
---------
Co-authored-by: Lia <lia@librechat.ai>
* perf: Async Local File Saves and Compile Hot-Path Regexes Once
* perf: Check Closing Code Fences Without a Per-Length Regex Cache
* test: Build the Delivery Route Config Before Routing Instead of Mutating It
mergeFileConfig now memoizes by config identity, so the CSV fallback tests
must set textFallbackWithoutTools when they build the request config rather
than mutate the object after the agent's routing already merged it.
* fix: Write Local Buffers Atomically So Concurrent Saves Cannot Interleave
Async writeFile calls to one deterministic path (OpenID and SAML avatars)
could truncate and interleave each other. Write through a temp file and a
rename instead, and add scenario specs for local saves and longer closing
fences.
* refactor: Move Local Save Logic Into packages/api
saveLocalBuffer, the temp-upload move and the atomic directory writes now live
in packages/api/src/storage/local.ts and receive their paths from the caller;
the Local strategy in /api keeps its logging and calls through. The traversal
check now runs before any directory is created.
* fix: Copy Moved Uploads Through a Temp File and Rename Into Place
Concurrent async copies to one destination could interleave just like direct
writes, so moveLocalFile now fills a sibling temp file and renames it.
* fix: Contain Every Local Write and Move Inside Its Target Directory
The traversal check that guarded saveLocalBuffer now runs in writeLocalFile
and moveLocalFile too, before any directory is created, and the buffer
save accepts only the images and uploads base paths.
* test: Define the Success Assertion Inside the Local File Saves Scenario
* fix: Give Atomic Write Temp Files a Fixed-Length Name So Long Targets Still Fit
* test: Send Image Dimensions and Read the Message Array in the Local Saves Scenarios
deleteCodeEnvFile rethrew every error except 404. A deployment that only serves
the stateful profile answers a default-profile delete with 409
execution_profile_mismatch on every retry, so the file landed in failedFileIds,
its record and agent references stayed, and agents holding it aborted every code
turn with "Failed to provision N file(s) to the code environment". The mismatch
is now handled like an absent object: logged, and the delete proceeds. Other
409s, 5xx, auth and network errors still fail the delete.
* 🔭 feat: Recover Stale Chunks Without the Error Page and Ship Client Logs to RUM
Stale chunks: a tab left open across a deploy that opens a code-split panel
requests a hashed chunk the new deploy removed. Vite's preload handler hands
the failure to the stale-asset recovery, which reloads, but the import then
resolves to `undefined` and `React.lazy` throws a TypeError into the route
error boundary, so the error page flashes before the reload. Lazy imports now
go through `lazyWithRecovery`/`importWithRecovery`, which turn that result into
a `ChunkLoadError`; the route boundary recognizes chunk failures (and any error
while a recovery reload is underway), asks the page recovery once, and renders
a localized "Updating to the latest version…" status. When the loop guard
declines, the normal error page with its reload action shows instead. The
page recovery is now idempotent within a page, so concurrent callers join the
in-flight reload rather than tripping the guard.
Client logs: with RUM in proxy mode, the client logger's warn/error calls,
error-boundary catches and stale-asset recovery events are exported as OTLP
log records through the authenticated `/api/rum/v1/logs` proxy. Records carry
allowlisted, scrubbed fields only, follow RUM session sampling, dedupe repeats
with a count, are budgeted per minute and per page, batch with backoff, flush
with keepalive on page hide, and stop after persistent failures. The proxy
now refuses non-OTLP content types and applies a per-user rate limit.
`RUM_CLIENT_LOGS=false` turns the export off.
* 🔒 fix: Keep Client Logs Opt-In and Drop Application Error Messages
- RUM_CLIENT_LOGS now defaults to off, like console capture, so enabling
RUM proxy mode never starts a new log stream on its own.
- Error messages are exported only for engine, browser and library
generated types (TypeError, SyntaxError, DOMException, AxiosError,
chunk failures); application errors keep type, status and stack only,
since their messages can echo prompt or response text. All quoted
fragments are now redacted.
- Batches are measured in encoded UTF-8 bytes of the exact request body,
and page hide drains every queued record that fits the keepalive quota.
- Queued stale-asset events are delivered as soon as the log exporter
starts, independent of the RUM SDK loading, and marked so they are
never logged twice.
* 🔒 fix: Export Only Known Generated Error Messages and Ack Queued Events
- Error messages are exported only when the whole text matches a known
engine, browser or library message shape (property access on
undefined, "x is not a function", fetch/abort/axios status, chunk and
CSS preload failures). The error class alone no longer qualifies, since
any code can throw a TypeError carrying arbitrary text.
- Queued stale-asset events are marked logged only after the collector
accepts the record carrying them, so an undelivered event replays on the
next page; in-flight events are skipped by the SDK path.
- A fractional RUM_PROXY_USER_MAX below one falls back to the default
instead of flooring to a limit of zero.
* 🔒 fix: Hold Unacknowledged Asset Events, Re-send In-Flight Logs on Hide
- The RUM SDK flush no longer drops queued stale-asset events whose client
log record is still unacknowledged: they stay queued and persisted,
flagged so their SDK action is not repeated, and leave the queue once
the collector accepts the record.
- Page hide re-sends the records of an unanswered regular request with
keepalive; each record belongs to the latest request carrying it, so a
late response to the original request neither marks nor requeues it.
- Client-log export stops when the authenticated layout unmounts.
- The proxy refuses /v1/logs while client logs, console capture and replay
are all off, so turning RUM_CLIENT_LOGS off also stops tabs that still
hold an older startup config.
* 🔒 fix: Export Only Complete Stack Frames, Never the Message Header
Stack reduction now strips the V8 `name: message` header (including every
line of a multi-line message) and keeps only lines with a complete V8,
Gecko or WebKit frame shape that ends in a URL or absolute path with a
line number. A message ending in a source-position-like suffix can no
longer pass as a frame.
* fix: centralize RUM delivery and asset recovery lifecycle
* fix: preserve RUM page lifetime and collector acknowledgements
* fix: retain page transport quotas and recover functional chunks
* fix: scope RUM client sessions to authenticated identity
---------
Co-authored-by: Lia <lia@librechat.ai>
* fix: Separate workspace admission and transport budgets
* fix: Validate workspace run reserves and retry test types
---------
Co-authored-by: Lia <lia@librechat.ai>