* feat: theme-adaptive SVG support for custom MCP and group icons
Custom icons (MCP server iconPath, model spec groupIcon) were rendered as
plain <img>, so monochrome SVGs kept fixed dark colors and were nearly
invisible in dark theme.
Introduce a shared CustomIcon component that detects monochrome SVG glyphs
and tints them with currentColor so they follow the active theme, while
multi-color SVG logos and raster images keep their original colors. The
monochrome decision parses the SVG's color tokens; content is fetched once,
cached, and any failure falls back to the original image. Monochrome SVGs
render via CSS mask, never inlined, so no SVG markup reaches the DOM.
Apply across all custom-icon surfaces: MCP settings cards, the chat MCP
dropdown, stacked MCP icons, tool-call headers, and model group icons.
Also support SVG in the MCP avatar uploader: add SVG to the accepted file
types and sanitize uploaded SVGs with DOMPurify before storing them, and
make the dialog preview theme-adaptive via the same component.
Add unit tests for SVG detection, monochrome analysis, sanitization, and
CustomIcon rendering.
* fix: reset adaptive icon tint state when the source changes
A reused CustomIcon instance kept the monochrome verdict from a previous
source until its effect re-ran, so switching to a raster image or a new
multi-color SVG could briefly render it as a currentColor silhouette.
Key the resolved verdict to the current source and reset it synchronously
during render (seeding from cache when available), so a stale verdict can
never tint a different icon. Add hook tests covering the source change.
* fix: do not tint custom SVGs that have an opaque background
A grayscale SVG with an opaque full-canvas background (for example an
exported logo with a white background rect and a black glyph) passed the
monochrome check and was drawn through a CSS mask. Masks key off the alpha
channel, so the opaque background filled the whole area with the tint color
and the icon collapsed into a solid block.
Detect a full-canvas opaque background rect and exclude such SVGs from
tinting, rendering them with their own colors instead. Transparent
single-color and multi-shade glyphs remain tintable. Add tests for the
background cases.
* fix: detect opaque SVG backgrounds without a viewBox
Opaque background detection only read canvas dimensions from the viewBox, so
an SVG that declares width and height on the root element but omits the
viewBox slipped through and was tinted into a solid block.
Fall back to the root svg width and height when no viewBox is present, and
match attribute names exactly so stroke-width is not mistaken for the canvas
width. Add tests for the no-viewBox cases.
* fix: parse comma-separated SVG viewBox values
The viewBox regex only accepted whitespace between values, so a valid
viewBox like "0,0,24,24" failed to parse and an opaque background went
undetected, tinting the icon into a solid block. Accept commas and
whitespace as separators. Add a test for the comma-separated case.
* refactor: detect SVG tintability with DOMParser instead of regexes
The monochrome/tintable decision scraped SVG markup with regexes, which kept
missing edge cases (opaque backgrounds, missing or comma-separated viewBox,
stroke-width vs canvas width, embedded raster images).
Parse the SVG once with DOMParser and inspect real elements and attributes:
reject embedded <image>/<foreignObject> content, detect a full-canvas opaque
background rect, read the canvas size from the viewBox or root width/height,
and gather paint colors from attributes, inline styles, and <style> blocks.
Unparseable input is treated as not tintable. Tests cover these cases.
* fix: only tint single-tone grayscale SVGs
A grayscale SVG can draw its background as a full-canvas path (not just a
rect), e.g. a white background path plus a black glyph. The rect-only
background check missed that, and the icon flattened to a solid currentColor
block under the CSS mask.
Tint only when the SVG resolves to a single grayscale tone. Any second tone
(a background shape drawn as a path or rect, an accent, or a second shade)
now preserves the icon's own colors, which covers full-canvas path
backgrounds without per-shape geometry parsing.
* fix: count default black fill when detecting tintable SVGs
* fix: treat currentColor as a tone and stroked closed shapes as filled
* fix: ignore non-rendering template paint and count filled open paths
* fix: scan defs and symbol paint rendered through use references
* fix: resolve currentColor against fixed color and match style fills for default paint
* fix: treat alpha-zero fills as transparent in SVG tint detection
* fix: resolve inherited and CSS fill for backgrounds and default paint
* fix: resolve CSS currentColor and default-black use instances
* fix: count default-filled polylines when detecting tintable SVGs
* fix: resolve inherited fill and style/CSS opacity in SVG tint detection
* fix: ignore display:none shapes and stop counting CSS color as paint
* fix: only count CSS paints that match rendered elements
* fix: match root for CSS currentColor and skip unreferenced defs tones
* fix: count paint only on rendered shapes and filter hidden CSS currentColor
* fix: resolve use-instance currentColor and skip hidden use default fills
* fix: resolve CSS paints onto rendered painters and skip overridden use fills
* fix: keep checking ancestor opacity after resolving paint opacity
* fix: resolve CSS paints by specificity and skip transparent default fills
* fix: apply CSS over presentation attributes and follow nested use targets
* fix: strip CSS comments, honor visibility, and inherit paint through nested use
* fix: skip transparent svg use references
* Account for SVG filter colors
* fix: propagate outer use color through nested currentColor references
* fix: preserve url() id casing and record every nested use chain
* fix: preserve SVGs whose filters recolor the source
* refactor: detect monochrome icons via canvas pixel sampling
Replace the SVG source parser with offscreen-canvas pixel sampling: draw the icon, read pixel data, and treat it as monochrome only when every non-transparent pixel is grayscale within tolerance and the image has some transparency (a fully opaque image would mask to a solid block). Verdicts are cached per src.
An explicit monochrome flag on CustomIcon skips detection; sampling is the fallback. Image load failures and canvas taint from non-CORS cross-origin icons fall back to untinted rendering instead of throwing.
* harden custom SVG icon handling across security, a11y, and perf
Sanitize user-provided MCP iconPath server-side; the client-side
DOMPurify pass was bypassable by posting iconPath straight to the API.
Adds sanitizeMcpIconPath in @librechat/api (allowlist SVG sanitizer that
preserves case-sensitive names) and runs it in the create/update
controllers, plus a length cap on iconPath in the shared schema.
- gate theme detection to same-origin/data sources so a remote icon is
no longer auto-fetched from every viewer's browser
- restore forced-colors (High Contrast) visibility and the broken-icon
onError fallback for tinted icons
- mark redundant icons decorative to avoid screen-reader double reads
- base64-encode inlined SVG data URIs and bound the verdict cache
- harden the client SVG sanitizer and guard FileReader failures
- cover the tinted render branch, sanitizer allowlist, and edge cases
* fix: preserve local SVG references while stripping external hrefs
The hardened sanitizers dropped <use> and every href outright, blanking
self-contained exporter output like <defs><path id="p"/></defs>
<use href="#p"/> and gradients inheriting stops via href="#g".
Re-allow <use> and href/xlink:href on both sanitizers, restricted to
same-document fragments: the client uses a dedicated DOMPurify instance
with an afterSanitizeAttributes hook (kept off the shared default
instance), and the server mirrors the rule with a sanitize-html tag
transform plus an empty allowedSchemes list as a second layer. External,
protocol-relative, relative-path, and javascript: hrefs are still
stripped, leaving the element inert.
* fix: keep xlink namespace and safe filters in the server icon sanitizer
The server allowlist stripped xmlns:xlink from the root while keeping
xlink:href, leaving an unbound prefix that is a parse error once the
icon is stored as an image/svg+xml data URI, so SVG 1.1 exporter output
went blank after save. It also dropped <filter>/fe* primitives that the
client sanitizer's svgFilters profile allows, so icons relying on
effects like feDropShadow lost them between preview and reload.
Allow xmlns:xlink and mirror the client's filter allowlist (elements
plus their presentation attributes); feImage passes through the existing
fragment-only href transform, so external references are still stripped.
* fix: require empty pixels for tinting and block external url() references
Two issues from review of the icon sanitizer and monochrome detector:
- scanMonochrome flagged any pixel below full opacity as transparency,
so an SVG that fills the whole canvas with a semi-transparent grayscale
wash was tinted and rendered as a solid currentColor block. Require a
genuinely empty pixel (alpha at or below the paint threshold) instead.
- The fragment-only reference rule only covered href/xlink:href, so a
url() in filter/fill/mask/clip-path/style kept external targets like
filter="url(https://evil/f.svg#f)". Both sanitizers now strip any
attribute carrying a non-fragment url(), preserving local url(#id).
* fix: preserve internal SVG stylesheets while scrubbing their references
Exporter SVGs that store multi-color paint in an internal <style> block
(e.g. .red{fill:#e00} with class="red") lost all styling because the
sanitizers dropped <style> outright, so multi-color logos rendered with
default paints after upload.
Re-allow <style> on both sanitizers and scrub its CSS: strip comments
and @import at-rules and rewrite any non-fragment url() to none, keeping
local url(#id) paints. These icons only render in inert <img>/CSS-mask
contexts isolated from the host page (no script, and no subresource
loads once external refs are scrubbed), so a script-free stylesheet is
safe to keep; a premature </style> close still has its trailing markup
re-sanitized by the parser.
* fix: normalize icon value before deciding it is not an SVG data URI
A posted iconPath with leading whitespace or C0 controls (e.g.
"\n data:image/svg+xml,...") slipped past the anchored data-URI check,
so sanitizeMcpIconPath returned it raw while a browser trims the prefix
and still renders the SVG, keeping its external <image>/url() and other
stripped content. Normalize the value the way URL parsing does, trimming
leading/trailing C0 controls and spaces and removing embedded
tab/newline, before the SVG check so these variants are sanitized too.
* fix: cap sanitized icon length and re-encode as compact base64
The schema length check ran on the input, but sanitizeMcpIconPath then
re-encoded the markup with encodeURIComponent, which expands SVG by well
over the base64 input size (a 235KB base64 icon became a 374KB percent
string). create stored that over-limit value, and editing the server
resubmitted the prefilled value and failed MCPServerUserInputSchema.
Re-encode the sanitized SVG as base64 (far more compact for angle-heavy
markup, which sanitizing can itself grow via explicit close tags) and
drop anything still over MAX_MCP_ICON_PATH_LENGTH, so a stored icon can
never exceed the cap that a later edit re-validates against.
* fix: resolve CSS escapes before stripping external SVG references
A literal url(/@import matcher missed CSS-escaped references: a value
like style="fill:u\72l(https://attacker/x)" or an escaped @import
survived sanitization because the browser un-escapes \72 to r at
tokenization time while the regex never saw a url(...). The external
reference was then persisted and served to other clients.
Un-escape CSS the way a browser does (single pass over \hex and
\char escapes) before the url()/@import matchers run, on both the client
and server sanitizers. Also scrub the multi-declaration style attribute
in place instead of dropping it, so co-located local paint rules survive
while escaped external references are neutralized.
* fix: re-sanitize after scrubbing style blocks to block markup reintroduction
scrubStyleBlocks runs after sanitizeHtml and splices un-escaped CSS back
as raw markup, so an escaped sequence like \3c/style\3e\3cimage/\3e
(inert text during the first pass) became a real </style><image href>
element after unescaping, past the allowlist. Re-run the allowlist over
the scrubbed result whenever a <style> block was rewritten, stripping any
element the un-escaping surfaced while leaving legit local rules intact.
The client sanitizer is unaffected: it sets the scrubbed CSS as a DOM
text node, which serializes with < and > escaped, so no element is
reintroduced. Added a regression test pinning that behavior too.
* fix: make forced-colors icon tint override the inline background color
The forced-colors rule for .custom-icon-tint had no !important, so it
lost the cascade to the element's inline `background-color: currentColor`
and the masked glyph still blanked out in Windows High Contrast — the
exact regression the rule was meant to prevent. Verified in a real
forced-colors render: without !important the tinted icon is invisible;
with it the glyph paints in CanvasText and stays visible.
* fix: enforce icon length cap in the sanitizer, not the schema
The schema `.max()` on iconPath rejected the whole update at parse time.
A server whose stored icon predates the cap (previous versions allowed
large data-URI images) re-submits that value from the edit dialog, so
the user was locked out of changing any field or clearing the bad icon.
Drop the schema `.max()` and enforce MAX_MCP_ICON_PATH_LENGTH in
sanitizeMcpIconPath for every value type: an over-cap SVG is compacted
or dropped, an over-cap raster/URL is dropped. Editing a server with a
pre-existing oversized icon now succeeds and clears the icon instead of
failing validation.
* fix: decode XML entities and parse quoted URLs in SVG CSS scrubbing
Two more bypasses of the CSS reference scrubber, both live once the
stored image/svg+xml is parsed by a viewer:
- A <style> block reaches the scrubber as raw text, so an entity-encoded
reference like @import (or @import) was never seen as @import;
the browser decodes it at render time. Decode XML character references
(numeric + the five predefined named) before the matchers run.
- The url() regex excluded ')' even inside quotes, so a valid quoted URL
like url("https://x/a)b") slipped through. Parse quoted and unquoted
URL tokens separately so a ')' inside quotes stays part of the target.
Applied to both the server (sanitizeMcpIconPath) and client (sanitizeSvg)
scrubbers. Entity decoding also catches an entity-encoded </style> markup
breakout, which the re-sanitize pass then strips.
* fix: close custom icon review threads
* fix: transpile sanitize-html's ESM dependency chain in api jest runs
* fix: preserve marker definitions and fragment-only marker references
* fix: keep object-cover sizing when tinting masked icons
* fix: preserve textPath labels and currentColor color scopes in icon sanitizer
* fix: preserve pattern coordinate-system attributes in icon sanitizer
Keep patternUnits, patternContentUnits, and patternTransform so patterned
SVGs do not fall back to objectBoundingBox after server-side sanitize.
* refactor: sanitize icons with one shared DOMPurify policy
The server re-implemented DOMPurify's SVG profiles as two hand-written
sanitize-html allowlists, so every element and attribute the client kept had to
be transcribed by hand, and each miss surfaced as an icon that previewed
correctly and then lost its markers, patterns, filters, or xlink binding once
stored. Run DOMPurify on jsdom instead and share the config and reference hook
with the client, which drops both allowlists.
Restrict url() values to same-document fragments on every attribute rather than
on the marker properties alone, so external references in fill, stroke, filter,
mask, and clip-path are stripped as well.
Restore feDropShadow casing after sanitizing. DOMPurify has to parse as HTML,
its XML mode drops filter primitives and gradients, and the HTML tag-name
adjustment table predates that primitive, so it survived lowercased and left its
filter empty once the stored icon was re-parsed as image/svg+xml.
* fix: defer jsdom load until MCP icon sanitization
jsdom and dompurify were imported at module scope, so every
require('@librechat/api') paid their startup cost. Load them on first
SVG sanitize instead.
* fix: sort type import in MCP icon sanitizer
The static-check import sorter wants value imports before type-only
imports. Moving the DOMPurify type import after librechat-data-provider
clears the drift.
* fix: address PR review bot findings
chatgpt-codex-connector:
- Resolve icon URLs through new URL() instead of trusting a leading slash, so
/\attacker.example/icon.svg is no longer treated as same-origin
- Check every url() token in an SVG attribute rather than only the first, and
reject values carrying a CSS escape that would decode into one
- Forbid animateTransform, animateMotion, animateColor, and mpath, which the
SVG profile admitted despite the policy intending to remove animation
- Preserve the radial-gradient fr attribute the profile omits
- Reject multi-tone grayscale in scanMonochrome, which a mask would flatten
- Sample at 128px so a small chromatic accent is not averaged away
- Settle a stalled icon load after 10s so its in-flight entry is released
- Bound the decoded SVG before building a jsdom tree for it
- Reject an oversized icon upload with a visible error instead of accepting it
and silently clearing it on save
* fix: address Codex findings and trim icon sanitizer comments
- Bound icon files by size before reading or sanitizing them
- Track tone spread across every painted pixel, not only solid ones
- Settle monochrome detection on any sampling throw
- Use named DOMPurify/JSDOM types and shorten PR comments
* fix: percent-decode SVG data URI bodies before base64 decoding
* fix: declare the SVG namespace on sanitized icons and keep icon rejections out of form validity
`create_file` reads its target before writing so it can tell a create from
an overwrite, so `readSandboxFile` missing the path is the ordinary case for
a create — yet it was reported at error level, through `logAxiosError`, which
rendered the sandbox's own `cat: …: No such file or directory` as "An error
occurred while setting up the request" plus a stack. Every ordinary file
creation emitted an error-level entry claiming a transport fault about a file
that was written half a second later, and an operator reading it concluded the
file had been lost.
Split the two failures `readSandboxFile` was conflating. A request that never
completed still goes to `logAxiosError`; a request the sandbox answered with
stderr is now classified on that stderr — an absent path logs at debug, and
anything else (permission, a broken interpreter) keeps error level with the
sandbox's own wording. The thrown error is unchanged, so the authoring flow's
create/overwrite decision still reads it the same way.
The classifier moves to `packages/api/src/files/code/errors.ts` and is now
shared: `handlers.ts` had two near-duplicate copies of it, the narrower of
which documented itself as narrower than the broader one without either
deriving from the other.
Also carry the cause on `[preflightCodeOutputBatch] … could not be inspected`.
That warning follows each `create_file` in the same request, and the batch
discarded the error entirely — so a Code API that refused the download, one
routed to the wrong execution profile, and a file too large to inspect all
arrived as the same sentence, and neither this warning nor the one above could
be told apart from the failures around them. It goes through
`getSafeErrorMetadata`, which is what keeps the upstream message and the
artifact's name — both of which can echo submitted content — out of the log.
`DELETE /api/files` looked the assistant up by `id`, but the Assistant schema
stores it as `assistant_id`. `api/db/connect.js` sets `strictQuery`, under which
Mongoose drops filter keys absent from the schema — so the lookup degraded to
`findOne({})` and returned whichever assistant happened to be first in the
collection, regardless of owner.
The unlink path then read that stranger's `tool_resources` and intersected it
with the caller's requested files, so a file the caller named could be unlinked
on the strength of another user's assistant record. With an empty collection the
same lookup returned `null` and the unguarded `assistant.tool_resources` threw,
answering 400 instead of completing the request.
Filter on `assistant_id` and guard the dereference. Adds three route tests that
mirror the server's `strictQuery` setting, without which the harness cannot
reproduce either failure.
`GET /api/convos` read its page size as `parseInt(req.query.limit, 10) || 25`
and handed it to `getConvosByCursor` unclamped, where it becomes
`.limit(limit + 1)`. `?limit=-1` therefore issued `.limit(0)`, which MongoDB
reads as no limit at all, and the route returned the caller's entire
conversation list in one response; `?limit=999999999` forced the same unbounded
fetch. `attachSharedFlags` then amplified it with a second query whose `$in`
carried every returned id.
Every other list route already clamps, so the fix reuses that shape: the
projects handler's local `normalizeLimit` moves to `packages/api/src/utils/query`
and both routes share it, and `getConvosByCursor` bounds its own page size so
the invariant holds for any caller.
Exports application logs as OTel log records when OTEL_LOGS_ENABLED=true,
correlated with the active trace, and gates every signal explicitly so a
disabled signal never falls back to the SDK's default OTLP exporter.
Verified end to end over OTLP/gRPC against an in-process collector.
* fix: Detect silently dead Redis sockets with deadline-bounded heartbeats
A Redis peer can vanish without a FIN or RST (a dropped NAT entry, a proxy
failover, a migrated VM). The socket stays `ready` from ioredis's point of
view, no error fires, and every command on it waits for the kernel's
retransmission timeout — about fifteen minutes at Linux defaults.
Observed in production: while the main client's socket was dead,
POST /api/agents hung until Cloudflare returned a 524; while the subscriber's
socket was dead, every attach failed with "Timed out synchronizing Redis
subscription" for sixteen minutes across five conversations.
`REDIS_PING_INTERVAL` could not help — it awaits ping() with no deadline, so
the probe just joins everything else that is waiting.
- Add startRedisHeartbeat: races a PING against a deadline and destroys the
socket when the probe goes unanswered, so ioredis reconnects, replays queued
commands, and re-subscribes a subscriber's channels.
- Destroy the stream rather than disconnect(true), whose FIN a vanished peer
never acknowledges; cluster clients fall back to the regular reconnect.
- Add createIoRedisSubscriber so every dedicated subscriber gets an error
listener and its own heartbeat, and route the generation-stream and both
subagent-routing subscribers through it.
- Set ioredis keepAlive from REDIS_KEEP_ALIVE (default 10s).
- Add REDIS_PING_TIMEOUT, REDIS_SUBSCRIBER_PING_INTERVAL and REDIS_KEEP_ALIVE,
documented in .env.example.
- Give the fake RESP server a silent mode and a SUBSCRIBE reply so the
dead-peer case can be tested against real clients.
* fix: Probe each cluster node and refuse a non-positive heartbeat deadline
A cluster-level PING is routed to one node, so a healthy reply proved nothing
about the other sockets. The heartbeat now resolves its targets each tick —
every node for a cluster, the client itself otherwise — and races an
independent deadline per target, tearing down only the socket that stops
answering. A non-positive deadline disables the heartbeat with a warning
instead of forcing a reconnect on every tick, and the single-probe test no
longer depends on wall-clock slack.
* 🚰 fix: Record the Provider Drain When Shutdown Cuts a Generation Short
* fix: Await generation settlement on shutdown instead of publishing the drain
* fix: Spend the real shutdown budget and cover the initialization window
* fix: Keep the shutdown registration literal the index spec asserts on
* fix: Keep draining settlements registered while shutdown is waiting
* fix: Track provider executions at the begin/drain chokepoints instead of per caller
* fix: Open the provider tracker before the begin CAS and cap cluster workers to the primary deadline
* fix: Propagate the cluster deadline and release trackers on reconfigure and abandonment
* fix: Hold SIGTERM until the worker acknowledges the cluster deadline
* fix: Refuse provider starts after shutdown begins and harden the cluster deadline handoff
* test: Assert the multi-execution shutdown wait with both executions begun before shutdown
* 🧊 feat: Persist Context Fading Tier Across Agent Runs
Carry the pruner's latched context-fading tier from `@librechat/agents`
through `contextMeta` the way `calibrationRatio` already travels, so a
historical tool result keeps the same truncated bytes from one run to the
next and Anthropic's prefix-based prompt cache survives across turns
(danny-avila/agents#497).
- Add `IAgentFadingTier` to the message and conversation `contextMeta`
types, schemas and the data-provider zod schema
- Add `isAgentFadingTier`, `resolvePersistableFadingTier` and
`resolveRunContextMeta` in `packages/api`; a tier is persisted once it
carries information (masking active or a budget below the window), and
a latched tier is kept even at a neutral calibration ratio
- Seed `createRun({ fadingTier })` from the parent response's contextMeta
regardless of encoding, since caps are character-based; the field rides
the existing forward-compatible `runConfig` variable and older SDK
versions ignore it, as does `Run.getFadingTier` being absent
- Accept a valid `fading` field in event actor context and reject a
malformed one, matching the calibration checks
- Capture run context meta through one module-level helper for both
completion paths instead of two inline copies
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Seed Fading Tier Across HITL Resume, Validate at the DB Layer
Address review findings on fading tier persistence:
- Leave the "is this tier worth persisting" decision to the SDK, whose
`Run.getFadingTier()` now returns only informative tiers; comparing
against the client's window misclassified untouched conversations
whenever a reserve ratio was configured
- Carry `contextMeta` through the HITL pause projection (staged approval,
job metadata, resume state) and seed the rebuilt client from it, so a
paused turn resumes with the same tier and calibration instead of
re-deriving a shallower one and rewriting the prefix
- Share run seeding between `chatCompletion` and `resumeCompletion`, and
make capture tolerate partial runs and bare resume contexts
- Validate `contextMeta.fading` in `commitAgentEventActorState` with a
shared `isAgentFadingTier` guard in data-schemas, so a malformed tier is
rejected at the DB layer instead of silently cold-starting the actor
- Reuse `IAgentEventActorContextMeta` for `IMessage.contextMeta` instead of
extending an inline duplicate
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 test: Cover Resume Seeding From Paused Context Meta
Assert the resume controller seeds the rebuilt client from the context meta
captured at the pause before rebuilding the run, and move the capture
helper's JSDoc back above its function.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧹 style: Sort Imports in Fading Tier Plumbing
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Persist Fading Tier Through Redis Jobs and Branches
Address Codex and Copilot review findings on fading tier persistence:
- Deserialize `contextMeta` in `RedisJobStore`, whose explicit read mapper
rebuilt every other pause field but this one, so Redis-backed
deployments resumed with no tier to seed after every HITL pause
- Copy server-private `contextMeta` onto the assistant message created by
`POST /api/messages/branch`, so the next turn seeds its pruner the same
way it would from the parallel source response
- Constrain the `fading` subdocuments in both Mongoose schemas (version
enum, positive budget, required fields) to match the zod schema
- Throw a distinct "fading tier is invalid" error from both context meta
validators instead of reporting a calibration failure
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Carry Context Meta Through Stop and Re-Pause
Addresses the second Codex pass on the fading-tier persistence:
- A re-pause whose resumed segment has nothing persistable now clears the
job's `contextMeta` in the same transition instead of leaving the first
pause's calibration and tier for the next resume to seed from.
- The client publishes the run's live calibration and fading tier onto the
job after each pre-invoke context snapshot (deduplicated on value), and
the Stop path copies `jobData.contextMeta` onto the stopped response, so
a follow-up from a stopped turn keeps the prompt prefix stable.
- Mid-run captures (HITL pause, Stop) read the graph's live state; `Run`
refreshes its own getters only after the stream settles, so the pause
used to persist the seeds it started from.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Fence Context Meta Publication Ahead of Stop
A Stop handled by another replica reads the job before the owner is
signaled, so a fire-and-forget publish could lose the race, and a run
stopped before its first context snapshot never published the tier it
inherited. The client now publishes the inherited seed before the run
streams or resumes, and the context-usage handler awaits each
snapshot's publish so the write lands before the model call it
describes begins. Failures still only log; the pre-run publish is
optional-called so bare resume contexts keep working.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Read Back Context Meta After the Abort Claim
The run owner publishes its context meta ahead of each model call and
awaits the write, so a publish landing between the abort's initial job
read and its terminal claim describes the call whose partial output the
abort snapshot carries. abortJob now reads the same-epoch job back after
the claim and content re-read and takes its context meta, so the stopped
response persists the tier that produced its bytes. The refresh is
best-effort and logs on failure like the content refresh beside it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Share In-Flight Context Meta Writes and Clear Neutral Live State
Two snapshot callbacks in a parallel-agent run can capture the same tier
before the first write settles; the publisher now caches the in-flight
promise per serialized value so the second caller awaits the same
durable write instead of treating an uncommitted value as published.
A live snapshot with nothing left to persist after an earlier publish now
writes a neutral record (ratio 1, no tier), since a running job's fields
cannot be deleted through the metadata writer, so a Stop no longer
persists an earlier non-neutral state onto the response. The pre-run
seed publish is unchanged and is told apart from live snapshots by a
flag from the context-usage sink.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 chore: Drop Unused Rest Parameter in Context Meta Test
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 feat: Persist Per-Agent Fading Tiers as Compact Context Meta
Reconciles the host with @librechat/agents at 9aacad1a, where graph
history stays canonical and each Run derives a provider-only projection
from the latched tiers. LibreChat therefore persists only compact state:
the calibration ratio, the default agent's tier, and now the per-agent
tiers from Run.getFadingTiers(), stored as validated entries so agent IDs
never become MongoDB field names and restored onto a null-prototype
record for RunConfig.fadingTiers. Truncated messages, canonical tool
content, projection provenance and per-message truncation state are
never persisted; the SDK's latched flag is stripped on capture.
The per-agent map travels everywhere the single tier already did:
response contextMeta, event-actor checkpoint state, HITL pause and
resume job metadata in both stores, the Stop path, and branch creation.
A shared Mongoose definition keeps the message and conversation schemas
identical, the zod schema validates entries, and commitAgentEventActorState
rejects malformed maps. Comments no longer describe graph messages as
truncated; only the provider projection is.
Tests cover: tool results and inputs reach the persisted content parts
at full size; contextMeta carries exactly the compact tier and
calibration fields; a persisted snapshot round-trips into the seeded
RunConfig unchanged and prototype-safe; per-agent tiers survive pause,
Redis job round-trips, and event-actor context validation.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Restore Event Actor Types and Publish Seeds Before Abortable Setup
The previous commit's doc rewording in the conversation types dropped the
event-actor interfaces that precede the fading tier, which broke the
declaration bundle in CI; they are restored unchanged.
The inherited context meta is now published onto the job as soon as the
parent's state is loaded and awaited at the top of chatCompletion and
resumeCompletion, ahead of run creation and the other abortable setup
stages, so a Stop during setup still persists the parent's tiers onto the
stopped response. The partial response saved when every subscriber
disconnects now carries the job's context meta like the Stop and pause
paths do. The branch route keeps the source's context meta in the saved
message but strips it from the HTTP response, matching client reads.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Serialize Distinct Context Meta Publications
Overlapping snapshots from parallel agents could carry different tiers;
each started its own write and both job stores keep whichever write
finished last, so an older snapshot could overwrite a newer one. Each
distinct publication is now issued after the previous one settles, in
call order, so the newest snapshot is what a Stop reads back. Equal
values still share one write.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Project Root Message Reads and Keep Inherited Meta on Setup Failure
The root GET /api/messages paths returned raw documents: the
single-message query read called getMessages without a projection and the
cursor page performed an unprojected find. Both now apply
CLIENT_MESSAGE_SELECT, with getMessagesByCursor accepting a select option,
so the persisted fading tiers never reach the client through the
paginated endpoint; the search-hydration read projects the same way.
A turn that loads a parent's context meta but fails before its run exists
used to replace the inherited meta with the capture of a missing run, so
the persisted error response lost the last valid tiers. Both finalizers
now keep the inherited meta when no run was created while a created run's
neutral state may still clear it.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 refactor: Move Context Meta Publication Into packages/api
The publication coordinator (ordering, deduplication, in-flight sharing,
failure state) now lives in packages/api as createContextMetaPublisher,
with selectRunContextMetaToPublish deciding what a publication carries;
the legacy client keeps only the wiring to the job metadata writer. A
transiently failing write is retried with a short backoff before the
publication is reported and forgotten, so a snapshot's model call no
longer proceeds on the first rejection while the job holds the previous
record. The SDK swallows handler errors, so retrying is the only way to
strengthen the fence without failing the turn.
A terminal save now supplies contextMeta: null when the finished run has
nothing to carry, and saveMessage unsets a previously stored value in
that case, so a disconnect snapshot's record cannot outlive the run that
completed neutrally.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Widen the saveMessage Method Type for the Null Unset
The exported method interface still typed contextMeta as the message
field, so the spec exercising the null unset failed typechecking; it now
mirrors the implementation signature.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Publish Every Agent's Snapshot, Sanitize Search Hits, Unbound Agent IDs
The context-usage handler now awaits the run's context meta publish for
every ON_CONTEXT_USAGE event, hidden sequential agents included, while
the visible-snapshot bookkeeping and the client emission keep their gate;
a Stop before the next visible snapshot finds the tier a hidden agent's
model call latched. Search results are built from the hydrated hit minus
its server-private contextMeta, since Meilisearch hydration projects every
schema field. The per-agent tier entry guard no longer caps the agent ID
length: ephemeral agent IDs encode endpoint, model and sender without a
bound, and dropping such an entry would let that agent alone re-derive
its tier on the next turn. Also formats the publication spec so the
static checks pass.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Unset Context Meta Atomically, Keep Publication State After a Failed Write
- `saveMessage` folds the `contextMeta` unset into the same update that
persists the terminal response, on both the plain and the provenance-merge
paths, so a failure between two writes can no longer leave a completed row
carrying a disconnect snapshot's state.
- `createContextMetaPublisher` tracks whether any record has committed
separately from the retryable latest publication, so `hasPublished` stays
true after a later publication exhausts its retries and a following neutral
snapshot still overwrites the earlier record on the job.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Carry Context Meta Through the Resumable Stop Route and Neutral Resumed Completions
- The resumable Stop route copies the refreshed `jobData.contextMeta` onto the
terminal assistant message it persists, as the abort middleware already
does, so a stopped resumable response keeps the tiers that produced its
partial output.
- A resumed run that completes with neutral context state now saves
`contextMeta: null`, which unsets the paused segment's record instead of
letting an omitted field keep it on the completed row.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Trust Only Server-Authored Context Meta, Unset It on a Neutral Stop, Read It From the Job on Disconnect
- `POST /api/messages/:conversationId` and conversation imports drop any
client-supplied `contextMeta`, and the run seeds its calibration and fading
tiers only from a server-authored parent response, so a forged tier can never
shape a provider projection.
- Both Stop paths save `contextMeta: jobData.contextMeta ?? null`, so a job that
re-paused with neutral state unsets what an earlier pause stored on the row
instead of leaving it behind.
- The disconnect partial save reads the same-epoch job record for the run's
published context meta, since the client-facing resume snapshot never carries
server-private state.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧪 test: Mock the Job Store in the Resume Metadata Spec
The disconnect partial save now reads the same-epoch job record through
`GenerationJobManager.getJobStore()`, so the resume metadata spec's manager
mock provides it as the other disconnect spec already does.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 🧊 fix: Keep Context Meta Off the POST Message Response
`POST /api/messages/:conversationId` returned the saved row verbatim, so a
client write against an existing server-authored message echoed the row's
stored context meta. The route now projects the row through the same client
projection the branch route uses, which drops the server-private field.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
* 📦 chore: Bump @librechat/agents to 3.7.16 for Persisted Fading Tiers
3.7.16 is the first release that carries danny-avila/agents#497, so
`Run.getFadingTier()` / `getFadingTiers()` now report the latched tier
this branch persists and `RunConfig.fadingTier` / `fadingTiers` seed the
next run from it. The dependency set is unchanged from 3.7.15; only the
package entry moves.
Adds an SDK-backed round trip to the fading spec: a first pruner latches
an informative tier, `resolveRunContextMeta` reduces it to the persisted
shape, and a second pruner seeded from that shape reproduces the first
run's projected tool-exchange bytes under instruction and calibration
drift, while an unseeded pruner under the same drift relaxes its tier and
rewrites them.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01WNbgMeR2aZDyhsqAy4L8gF
---------
Co-authored-by: Claude <noreply@anthropic.com>
* ⏭️ feat: Honor an Interrupt Before the Model Has Answered
An interrupt armed while the model is still thinking sat as "Interrupting"
until the entire turn finished, then landed as a terminal continuation. The
cause is upstream: the SDK reads `shouldPreempt` once per streamed chunk, and
its seal gate requires non-empty text — so a silent provider is never polled at
all, and a reasoning-only turn is never sealable.
`@librechat/agents` gains a discard-and-restart path for exactly that window
(`HOOK_PREEMPT_RESTART_CAPABLE`). This wires the host half:
- `GenerationJobManager.subscribePreempt` registers run-scoped wake listeners,
notified after an ACCEPTED arm — local or cross-replica, since both land in
`armPreemptIds`. Fenced on `createdAt` like every other preempt entry point,
and a throwing listener never fails the arm a queued steer depends on.
- `createSteerPreemptPoll` hands that channel to the SDK as
`StreamPreemption.subscribe`, gated on a new `isSteerPreemptRestartSupported`
probe so an SDK that cannot act on a wake is never handed one.
The wake is a hint only: `isPreemptRequested` stays the single authority, so
the level-triggered contract the seal path depends on is unchanged.
Inert until `@librechat/agents` publishes the restart contract — the probe
returns false against the pinned 3.7.11 and the run keeps today's behavior.
* 🩹 fix: Replay an Already-Armed Preempt on Subscribe
Preempt requests are level-triggered, and the window between a job becoming
steerable and the SDK installing its model-attempt listener is real. An
interrupt landing there was recorded with no callbacks to notify, and on a
silent or reasoning-only turn no later chunk poll may ever run — leaving the
request to wait out the whole turn, the exact stall this channel removes.
`subscribePreempt` now replays an existing arm to the newly registered listener
only, after registration: waking the whole set would re-notify runs that
already looked, and waking before registration would leave a concurrent arm
with nowhere to land. The `createdAt` fence is unchanged and still gates the
replay.
The SDK also reads the level flag once at each attempt's start, so this is
belt-and-braces rather than the sole cover — but it makes the host correct on
its own terms instead of resting on when the SDK happens to look.
* 🔧 chore: Update @librechat/agents dependency to version 3.7.14 in package-lock.json and related package.json files
* 🔧 chore: Update dependencies in package-lock.json and package.json
- Bump versions for @humanfs/core (0.19.1 to 0.19.2), @humanfs/node (0.16.6 to 0.16.8), @xmldom/xmldom (0.8.13 to 0.8.15), and qs (6.15.2 to 6.16.0).
- Add @humanfs/types (0.15.0) as a new dependency for @humanfs/node.
- Update dependencies for side-channel and side-channel-list to their latest versions.
* 🛑 fix: Forward Run Abort Signal to Foreground Tool Calls
Stopping a generation left in-flight tool calls running on the far side.
For MCP over streamable-http the external server kept burning CPU until
the tool finished or timed out, with no `notifications/cancelled` ever
sent.
The MCP layer was never the problem: `MCPManager.callTool` already
spreads `options.signal` into `client.request`, and the SDK's
`Protocol.request` sends `notifications/cancelled` the moment that signal
aborts. The signal simply never arrived.
Agent tools do not execute inside LangGraph's ToolNode — `eventDrivenMode`
dispatches `on_tool_execute` and `createToolExecuteHandler` invokes them.
The agents SDK puts the run's abort signal on the batch request
(`ToolExecuteBatchRequest.signal`, documented as one handlers should
forward), but the handler destructured every field except `signal` and
built its invoke config without it. `config.signal` was therefore
`undefined` in every foreground tool, so `createMCPTool` derived no
signal and the SDK registered no abort listener.
Forward it. This restores cancellation for all signal-aware foreground
tools, not just MCP. The detached background invoke keeps its own
controller, since that work deliberately outlives the turn.
An aborted call now rejects promptly and resolves an error result rather
than hanging, so the abort is logged at debug instead of as a tool error
— a user pressing Stop should not spray the error log.
* 🔇 fix: Keep Cancelled Tool Calls on the Filtered Path and Out of Error Logs
Addresses both findings from the Codex review of 48fded564a.
P1 — the cancellation branch returned before `filteredToolOutputResult`,
so with tool-output filtering configured an aborted call's error text
reached the turn uninspected. Worse, `runSignal.aborted` says the run is
over, not that this rejection was the cancellation: an unrelated failure
racing the Stop took the same unfiltered exit. Cancellation now only
selects the log level; filtering and the result shape are unchanged.
P2 — `createMCPTool` logs its own error before wrapping and rethrowing,
so every Stop during an MCP call still emitted an error-level MCP entry
and could pollute operational alerts. That catch now recognizes the abort
and logs at debug; the wrapped message still reaches the turn.
Both are covered by tests that fail against the previous commit.
* 🎯 fix: Require an Abort-Shaped Error and Spare Shared OAuth Flows
Addresses both findings from the Codex review of b3c67082cd.
An aborted run signal proves the turn is over, not that the rejection in
hand was the cancellation — a permission, OAuth, or upstream failure can
reject in the same tick a user presses Stop. Both quiet-log branches now
require the error to look like an abort as well, so a real failure racing
the Stop stays at error level and visible to operational alerts.
`isAbortError` moves to `@librechat/api` and the copy in
`abortMiddleware` is retired, keeping one implementation. Its spec pulls
the real function through its partial package mock rather than asserting
a stub.
Forwarding the run signal also reached ActionService, where two of the
three OAuth flows are keyed `userId:action_id` and so outlive the run
that opened them: a second run for the same action joins the very same
record, and the browser's OAuth callback reads its metadata to exchange
the code. `monitorFlow` deletes the key when a waiter's signal aborts, so
one Stop could strand a concurrent run and discard an authorization the
user had already granted. Those two flows no longer take the signal; the
run-scoped login flow, keyed by thread and run, still does.
Every guard is covered by a test that fails when the guard is removed.
* 🧷 fix: Detach the Stopped OAuth Waiter Instead of Dropping Its Signal
Addresses the Codex review of d53ff170ab.
Withholding the run signal from the shared Action OAuth flows protected
concurrent runs and the browser callback, but left this invocation
attached to the flow after the Stop. A callback landing late then resumed
`_call` straight into `preparedExecutor.execute` — a consequential API
request running after the user had already stopped the turn. Handing the
signal back is not the answer either, since `monitorFlow` deletes the
shared `userId:action_id` key on abort.
`detachOnAbort` separates the two: the caller stops waiting the moment
the signal aborts, while the shared flow runs on for whoever else needs
it — another run, or the callback exchanging its code. Late settlement of
the detached work is swallowed rather than surfacing as an unhandled
rejection.
Both shared flows now use it. The run-scoped login flow keeps the signal
directly, as nothing outside its run observes it.
* 🪪 fix: Keep an Action OAuth Abort an Abort
Addresses two of the three findings from the Codex review of 70fc0782b1.
Detaching the stopped waiter left the rejection to be misread on the way
out. The refresh catch treated it as a failed refresh and called
`requestLogin()`, emitting an OAuth prompt and opening pending
authorization state for a turn that had already ended. `requestLogin` and
the surrounding auth catch relabelled it `Failed to authenticate OAuth
tool` / `Authentication failed`, and the outer catch handed that to
`logAxiosError` — so an ordinary Stop produced two error-level entries
and returned failure text as the tool's result. Each boundary now lets an
abort through unchanged, and the outer catch logs it at debug and
rethrows so it stays a cancellation end to end.
Deferred: waiter-only cancellation in `FlowStateManager`, so a detached
`monitorFlow` stops polling instead of running out the flow's TTL. That
poller is bounded at three minutes and lives exactly as long as it does
for a waiter that never left, so it changes no lifetime this branch
introduced; giving `FlowStateManager` a non-destructive abort mode is its
own change, shared with MCP and indexSync.
* 🤫 fix: Classify the Abort at the MCPManager Boundary Too
Addresses the Codex review of 9bcf1aa57e.
`MCPManager.callTool` logs every rejection at error level before
rethrowing to `createMCPTool`, so a user Stop still emitted an MCP error
from the inner boundary even after the outer one learned to recognize it.
The same signal-plus-abort-shape classification now applies here.
The review also noted, correctly, that the existing test could not have
caught this: it replaces `getMCPManager` wholesale, so it only ever
exercised the outer logger. The new cases drive the real manager with a
connection whose `client.request` rejects the way the SDK does once a
request signal aborts, and cover both the cancellation and a genuine
failure racing the Stop.
* 🧹 chore: Restore Import Order in the Agents Tool Handler
* feat: Handle reaching the agent tool-call limit gracefully
When the graph exhausts its per-turn step budget, persist the turn as
unfinished with finish_reason tool_call_limit instead of a red error.
Warn the model as the budget runs low, and offer Keep going or Answer now.
* docs: Add screenshots of the tool-call-limit notice
Light and dark captures of the unfinished-turn card for the PR description.
* style: Tighten the tool-call-limit notice card
Pin dismiss to the top-right with uniform rounding, use a slightly
smaller hit target, and round the card to xl. Refresh the PR screenshots.
* fix: Address Copilot review findings on the step-budget path
Skip empty and subagent-scoped PostToolBatch events when counting
rounds. Drop the double-cast in the notice test helper and coerce the
tool-call-limit finish reason in the Continue-gate spec.
* fix: Address Codex review findings on budget and dismiss
Pass the resolved recursion limit into the OpenAI-compatible
processStream path so the step-budget hook matches the graph.
Reset the notice dismiss state when the displayed message id changes.
* fix: Unblock CI typecheck and import sorting
Declare recursion_limit on the OpenAI service agent type, sort that
file's local imports, and route the notice test context through unknown
so tsc accepts the partial ChatContext fixture.
* fix: Sort remaining PR imports for static checks
* chore: Remove tool-call-limit notice screenshots
* fix: Keep recovery actions from consuming composer drafts
Pass empty file, skill, and quote overrides on Keep going and Answer
now so a staged compose is not attached or drained.
* fix(api): align message search with page size
Pass a positive, capped page size to MeiliSearch and use the same conversation limit so requested message search pages are not silently truncated.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* fix(data-schemas): expand shared-link search results
Request the MeiliSearch result ceiling using only conversationId so cursor pagination is not limited to the default 20 hits or burdened with unused fields.
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
---------
Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
* 🚧 fix: Narrow Chat MCP Selections to What the Picker May Offer
The chat MCP picker offers a server only when the operator has not hidden it
(`chatMenu: false`) and the user reaches it directly rather than through an
agent (`consumeOnly`). Nothing enforced either condition on the request, so a
stale client, a replayed body, or a hand-written one could name any server and
get its tools — `consumeOnly` in particular is a permission the registry derives
per user, so ignoring it let a user reach in plain chat a server they had only
been granted through an agent.
Narrow the picker's selection in the two ephemeral loaders, against the same
predicate the client's `selectableServers` uses. The resolver is injected, so
`packages/api` stays free of the registry and the loaders keep their existing
`deps` shape.
Provenance is why this lives in the loaders rather than at the tool-resolution
chokepoint: only here is a server name known to have come from the picker.
Servers a model spec pins are the operator's own choice and are added after the
narrowing, so pinning a chat-hidden server to a spec keeps working.
`chatMenu` only ever comes from the config tier, so the overlay settles it with
no lookup at all. Lookups for the rest run concurrently and are cached per
server and user. A name the registry cannot resolve — request-tier, declared on
the body and never registered — is kept, as is one whose lookup fails: this
narrows an already-authenticated selection and is not the authorization
boundary.
* 📝 docs: Say What chatMenu Now Bounds
The schema still described `chatMenu` as dropdown visibility alone, which is
misleading now that it also bars a server from the chat selection a request
carries. Spell out both halves, and what it deliberately does not restrict:
agents, and servers a model spec pins.
* 🛡️ fix: Close Four Gaps in the Chat MCP Narrowing
Dedupe before resolving. The selection is request-supplied and the loaders used
to build a `Set` immediately; mapping the raw array first let one body repeat a
name enough times to launch a resolver — registry, cache, and ACL reads, with no
single-flight — per occurrence.
Resolve names through the tool-loading aliases. A configured `private server` is
addressable as `private_server` downstream, so an exact-key lookup let the
normalized spelling walk past a `chatMenu: false` check and still load tools.
Names are returned as sent; only the decision is canonicalized.
Let the request overlay settle `chatMenu`. The config tier is where the flag
lives and is what the picker itself was built from, so an overlay that re-enables
a server now outranks the durable config the registry returns, which was silently
dropping a valid selection. `consumeOnly` still comes from the registry.
Narrow the request body, not just the loader's copy. `applyContextToAgent` reads
`req.body.ephemeralAgent.mcp` directly and prefers it over the agent's tools, so
a hidden server's `serverInstructions` reached the model context even after its
tools were dropped.
* 🧭 fix: Ask the Registry Which MCP Servers the Picker May Offer
Three rounds of review found the same mistake in three disguises: this narrowing
re-derived the registry's tier precedence instead of asking for it. A config
entry was treated as authoritative over a user-DB or process-backed server of the
same name; the alias map was built from config names only, so a DB server named
`private_server` was misrouted to a config `private server`; and the overlay
short-circuit ran before any of that could be consulted.
Resolve the accessible set through `resolveAllMcpConfigs` instead — the same call
behind `GET /api/mcp/servers`, so what the request may use and what the picker
was offered come from one place and cannot drift. Precedence, cross-tier name
collisions, and the config overlay are the registry's job again, and the
selectability test is a single flag check over what it returns.
The result is smaller: no overlay branch, no per-server lookups, one catalog read
regardless of how long the selection is, and aliases built over the whole
accessible set so an exact name always outranks another server's normalized form.
Publish the servers actually in play back onto the request body, spec pins
included. `applyContextToAgent` prefers that list over the agent's tools, so
narrowing to the picker selection alone had started dropping a spec-pinned
server's `serverInstructions` while suppressing a hidden one's.
* 🔑 fix: Decode Action OAuth Credentials After Decryption
Action client credentials are URL-encoded before encryption, but the OAuth
token exchange decrypted them with `decryptV2` alone, sending values like
`abc%40domain.com` to the provider and causing 401s for secrets containing
`@`, `+`, `=`, `/` or `:`. Decoding falls back to the raw decrypted value so
legacy secrets stored before 299cabd6e (which contain an unescaped `%`) still work.
Fixes#14636
* 🔑 fix: Share Action Credential Codec Between Read Paths
The encode-before-encrypt convention had two readers with different
behaviour. `ActionService.decryptSensitiveValue` decoded unconditionally,
so a credential stored before encoding was introduced (299cabd6e, March
2025) containing a stray `%` threw `URIError` and failed every read of the
action, not just the token exchange.
Extract the pair into `packages/api/src/actions/crypto.ts` as the single
implementation and consume it from both `ActionService` and the OAuth token
exchange, so the two paths cannot drift again.
Tests cover the codec against real crypto: reserved characters and literal
percent escapes round-trip, pre-encoding values are returned unchanged, and
a pre-encoding value with a stray `%` no longer throws.
A version marker on newly encrypted values was prototyped and dropped: it
changes the stored format for every action credential while fixing nothing,
since values written by the encoding convention already decode correctly
and pre-encoding values stay unmarked either way.
---------
Co-authored-by: Danny Avila <danny@librechat.ai>
* 🖼️ fix: Stop Reporting a Failed Sandbox Image Read as an Unreadable Format
`read_file` on a `/mnt/data` image collapsed every failure — rate limit,
missing path, truncated transfer, transport error — into the binary-file
hint, whose image branch reads as a permanent capability limit ("is an
image file and cannot be read as text"). The model stopped retrying, and
the operator had no way to tell which failure had actually occurred.
The failure that surfaced it: the reader windows the bytes through
`/exec` stdout, one call per window, against the Code API's per-user
execution limiter (20 requests / 30s in the reference deployment). At a
fixed 32KB window a 122KB chart cost 4 calls and a 1MB image cost 33 —
more than a whole limiter window, so anything over ~640KB could never
complete. A turn that generates several charts and reads them back
exhausts the budget partway through and reports the 429s as unreadable
images.
- `handlers.ts`: failed reads name the real cause and the affordance that
matches it — rate-limited waits and retries, a missing path lists the
directory, a truncated transfer retries. The "cannot be read as text"
phrasing is now reserved for bytes that are not a supported image. An
image past the inline budget says to downscale and read the smaller
copy rather than pointing at `bash_tool`, which cannot show an image.
- `process.js`: the window is derived from the runner's stdout budget
(48,960B vs 32,768B, so a 122KB chart is 3 calls and 1MB is 22), halves
and re-reads the same offset when a runner truncates anyway (remembered
per Code API, with the env remedy logged), waits out a rate limit it
can afford instead of discarding the bytes already pulled, and stops at
a 24-call ceiling so no read stalls a turn across limiter windows.
- `utils/code.ts`: shared `getCodeApiRetryAfterMs` (header, then the
`retry_after_seconds` body); `skillFiles.ts` drops its private copy.
* ♻️ refactor: Move the Sandbox Image Reader Into TypeScript
Addresses the review round on #15453.
- P1: windowing, window sizing, the in-sandbox reader script, response
parsing and byte assembly move to `packages/api/src/files/code/image.ts`,
and the Code API rate-limit policy to `packages/api/src/utils/code.ts`.
`process.js` keeps only the `/exec` transport it owns. Both new modules
are tested against their real implementations (19 + 13 cases), including
the round-trip ceiling, the changed-file guard and the shell-injection
guard on the reader script.
- P2: a configured stdout budget below the old 8KB window floor produced a
window the runner could never emit, and narrowing was pinned to the same
floor, so no retry could succeed. The window now aligns to the base64
group size instead, and the round-trip ceiling bounds what a small budget
costs.
- P2: `.env.example` documented a 32,768-byte default that no longer holds
and never mentioned the stdout budget. It now describes
`LIBRECHAT_CODE_SANDBOX_OUTPUT_MAX_SIZE` as the setting to tune, with the
chunk override as the escape hatch.
- P3: `process.spec.js` no longer reimplements the retry parser in its
module mock. Behaviour is tested where it lives; the mock is a declared
passthrough and the spec covers what the transport still owns — session
forwarding, the profile header, parser hand-off and error logging.
- P3: `getCodeApiRetryAfterMs` reads the response body through `in`
narrowing rather than an `as` assertion.
- P3: the window-sizing JSDoc no longer claims a multiple of 3 in the one
case that is not — an explicit `LIBRECHAT_CODE_IMAGE_CHUNK_BYTES`.
* 🎯 fix: Size Sandbox Image Reads to What the Limiter Can Deliver
Addresses review round 2 on #15453.
- Narrowing now halves the caller's own failing window instead of the
shared learned value, and the learned value only moves down to the
narrowest size any read needed. Reads run concurrently (one per tool
call), so two reads failing at the same size were compounding into a
window two steps smaller than the runner would have accepted, and
persisting it for every later read.
- Narrowing continues until a runner accepts a window or a floor below
which no runner could serve an image at all. Two halvings stopped at
12,240 bytes — still ~16KB of base64 — so an 8-10KB stdout cap that was
not configured up front stayed unreachable.
- The round-trip ceiling is now one limiter window (20), and a read that
cannot finish within it says so from the FIRST response rather than
spending the whole budget rediscovering it: the first window reveals
both the file size and what a call really delivers. A near-limit image
no longer drains the turn's executions before failing, which is what
made the wait budget look too small.
- The round-trip message names the size this deployment can actually
inline, so the model has a downscale target instead of a guess.
- `python3: not found` from the sandbox reader was matching the generic
"not found" in `isSandboxMissingFileError` and being reported as a
missing image path, sending the model to `ls /mnt/data` while hiding a
runner dependency. Image reads now classify only path-specific ENOENT
forms as a missing file.
* 📐 fix: Size a Truncated Read's Retry From the Cap It Revealed
Addresses review round 3 on #15453.
A ~1KB sandbox stdout cap is a real runner default, and the 1,536-byte
narrowing floor sat above it: discovery stopped one step short and
declared every image unreadable, even though small ones fit comfortably
in the 576-byte window the configured path already derives.
Narrowing now runs until no smaller window exists. It also no longer has
to grope its way down: a truncated response emitted exactly as much as
the runner allows, so its length measures the cap this deployment never
declared, and the retry is derived from it directly — one round-trip
instead of the eight blind halvings it takes to fall from ~49KB to ~1KB.
Blind halving remains the fallback and bounds the result either way, so
every retry is strictly smaller than the window that just failed.
* 🗂️ fix: Resolve Code Interpreter Destination Collisions
Codeapi mounts each `/exec` input at a destination derived from its `name`
and rejects the whole request when two entries land on one. LibreChat never
enforced that: `primeFiles` sent `name: file.filename` verbatim, and the
graph seed deduped by `(storage_session_id, id)` — identity, not destination.
Two conversation files sharing a filename therefore killed code execution for
the rest of the conversation. The rejected request never reaches the sandbox,
so no result comes back to collapse the pair, and every later turn re-primes
both and fails identically. Only code-generated outputs are covered by the
`(filename, conversationId, context, tenantId)` partial unique index; uploads
carry `context: message_attachment`, so one conversation can hold several
records sharing a filename — two uploads of `image.png`, or an upload whose
name a later execution wrote back in place.
`packages/api/src/files/code/destinations.ts` now owns the name → destination
mapping for every contributor to an `/exec` file list, mirroring codeapi's
conflict rule including directory-prefix conflicts:
- `primeFiles` claims per pushed file, disambiguating with a `-2`, `-3`
counter and echoing the resolved path into the tool context, so the model
is told where each file actually mounts. Claim order comes from immutable
`createdAt`, not `getFiles`'s `updatedAt` sort, which usage accounting and
re-upload both bump — order-dependent destinations would silently repoint
a path an earlier turn had already handed the model. Newest wins the bare
name, matching `ToolNode.updateCodeSession` and giving an in-place rewrite
precedence over the original it superseded.
- `seedCodeFilesIntoSessions` drops a later file whose destination is already
claimed. Renaming there would invent a path nothing told the model about,
and a conflict at that layer means one logical file arrived twice under
different storage pointers.
Fixes#15443
* 🧹 chore: Tolerate Holes in the Destination Priority Sort
`primeFiles` sorts the raw `getFiles` result before its own per-entry guard
runs, so the comparator has to survive a null row rather than throw ahead of
the loop that was already defending against one.
Also tightens the tool-context assertion in the upload-vs-output case, which
was passing on a trailing space rather than on the path.
* 🔁 fix: Address Codex Review on Destination Assignment
Four findings from the review of `b0f61d0`.
**Identity-derived suffixes instead of a collision counter.** A counter is
assigned from whatever is in the set, so a third file arriving under the
literal name a displaced file had been given pushed that file along, and code
the model wrote in an earlier turn would silently start reading the newcomer.
The fallback now hashes the file's own `file_id`, so a displaced file keeps
its destination for the life of the conversation and only the winner of the
bare name can change. A trailing counter still guards a hash collision so the
search terminates on any input.
**Rank by last content write, not record creation.** `processCodeOutput`
claims one row per `(filename, conversationId)` and rewrites it in place, so
a repeatedly written output keeps its original `createdAt` while recording the
latest writer in `metadata.sourceDispatchedAt`. Ranking on `createdAt` alone
handed the bare path to an upload the output had since been rewritten over.
Recency is now `max(sourceDispatchedAt, createdAt)` — still never `updatedAt`,
which usage accounting bumps without any content changing.
**Advertise a displaced generated output.** Generated files are left out of
the tool context because the model already knows what it wrote — which stops
being true once a newer same-named file takes the bare path and the output
mounts somewhere the model has never seen. Displaced outputs are now listed
with the name they were written as, and displaced uploads name their original
too.
**Rank conversation files above an agent's private ones.** Each agent claims
destinations over its own candidate set — the conversation's files plus its
own — so a private file taking a bare name from a shared file in one agent but
not in another left two agents advertising different paths for the same file
into one shared mount namespace. Shared files now rank first, so the majority
that every agent sees resolves identically. The private-versus-private case
still drops at the merge and is documented there; closing it needs one
assignment across contributors, made before any agent renders its context.
* 📌 test: Pin Alias Ownership on an In-Place Rewrite
Codex round 2 flagged that the identity alias still falls back to a counter
when the alias itself is literally claimed, so a displaced file can move after
all. The mechanism is real and my JSDoc overstated the guarantee, which this
corrects.
The behaviour stays as it is. The realistic way a file arrives under another
file's alias is the model rewriting the displaced file in place: it was told
`<stem>-<hash>.ext`, wrote back to it, and `processCodeOutput` registered an
output under that literal name. Keeping the alias with the newest holder is
what makes the model's familiar path resolve to its own edit; reserving it
against the newcomer would hand that path back to the bytes it replaced. The
test pins that ordering so a later change cannot invert it silently.
* 🔐 fix: Redirect OAuth Browser Navigations Instead of Rendering JSON
The /oauth/* routes are reached by top-level browser navigation, so a
middleware that rejects one with a JSON body makes the browser render
that body as the page. Hitting the login rate limit during a social
sign-in dropped the user on a blank page showing:
{"message": "Too many login attempts, please try again after 5 minutes."}
Mark those requests with markOAuthNavigation and have the two middlewares
that can reject them answer with a redirect to the login page carrying an
error code the client localizes. The login limiter stays a single instance
so the attempt budget is still shared with POST /api/auth/login, and that
route keeps its JSON 429 for the SPA.
* 🧪 fix: Address PR Review Bot Findings
chatgpt-codex-connector:
- Split the inline `type TranslationKeys` specifier out of the value
import, per the standalone `import type` convention in CLAUDE.md. This
also clears the Static checks import-sort failure.
- Cover the new OAuth error branches in Login.spec.tsx: each mapped
`?error=` code raises its localized toast and the query is cleaned,
and unmapped codes including prototype members raise nothing.
* 🧹 fix: Strip LangChain Attribution From User-Facing Provider Errors
LangChain classifies provider failures by mutating `error.message` with a
docs URL, and that message is forwarded verbatim into the error content part
we persist, so users saw `docs.langchain.com/...` credited for a provider's
own 429. The client only intercepted `MODEL_NOT_FOUND`, leaving every other
code to render raw.
Resolve the classification server-side instead: codes with localized copy
become the typed payload the client already knows how to localize, and every
other message keeps the provider's wording with the docs URL removed before
it reaches the database. The untouched error still reaches the logs through
`getSafeErrorMetadata`.
The client keeps a code-driven lookup for messages persisted before this, and
gains `com_error_model_rate_limit` alongside the existing not-found copy. The
OpenAI-compatible and Responses controllers now share one provider-error
helper rather than three copies of it, so the strip applies on those surfaces
too.
* 🛡️ fix: Harden Error Sanitization For Non-String Messages And Legacy Text
Codex review follow-up.
`stripLangChainTroubleshootingUrl` assumed a string, so an error-like object
with a truthy non-string `message` threw `.replace is not a function` inside
the error handler that was meant to report it — the template interpolation it
replaced had coerced such values safely. It now coerces at the boundary, as
does the code lookup.
Legacy messages whose LangChain code has no localized copy still rendered the
docs URL, which is exactly the historical case the client-side path exists to
clean up. The rendered fallback is now stripped too.
Both helpers move to `librechat-data-provider`, which is shared by the client
and the backend, so one implementation serves both instead of a second regex
drifting in the component.
* 🖼️ fix: Apply an Agent's Configured Image Detail
The Agent Builder exposes an Image Detail slider and saves it to
`model_parameters.imageDetail`, but nothing ever read it back: `encodeAndFormat`
resolves the level from `req.body.imageDetail`, which the agents route never
populates, so every image was encoded as `detail: "auto"`. Providers that reject
`auto` — MiniMax's OpenAI-compatible API among them — fail every image upload
through an agent, and the setting that would fix it has no effect.
The root cause is a classification error rather than a missing argument.
`extractLibreChatParams` splits LibreChat's own settings out of an agent's model
parameters, and its key set is derived from the `librechat` group in
`parameterSettings`. `imageDetail` sat in `baseDefinitions` instead, so it was
neither extracted for LibreChat's use nor stripped from the options handed to the
provider.
It is now grouped with `resendFiles` and the other LibreChat-level parameters,
which gives it the same path those already take: extracted during agent
initialization, carried on the initialized agent, passed to the client, and read
by `encodeAndFormat`. `AgentClient.getSaveOptions` already expected
`options.imageDetail` — nothing had ever populated it.
The request body still wins where it is set, so the classic endpoints keep
resolving detail from conversation settings exactly as before.
Fixes#15359
* 🔤 fix: Type Image Detail as Its Enum, Not a Bare String
`TConversation.imageDetail` is the `ImageDetail` enum, so the extractor's return
type and the initialized agent's field are now declared as that enum rather than
widened to `string`, and the tests pass enum members instead of raw literals.
Caught by CI's type check, which I had run before adding the tests and not after —
jest compiles the specs through babel, so the suite stayed green over a type error.
