Commit graph

9 commits

Author SHA1 Message Date
Danny Avila
5ca49b5f4e
🪢 fix: Resolve the Races Behind Four Intermittent Session and Scenario E2E Tests (#18061)
* 🔑 fix: Replay a Request Rejected for a Superseded Bearer Instead of Refreshing Again

A burst of requests leaves with one bearer. The first 401 starts a refresh and the
rest share it only while it is in flight: a 401 delivered after the refresh settled
found no recovery running and started a second one, although the session had already
recovered and the response described a bearer that was no longer in use.

Compare the bearer the rejected request carried with the current one. When it has
been replaced and no recovery is running, replay the request with the current
bearer. A replay that is rejected again carries the current bearer, so it takes the
ordinary refresh path. The fetch transport follows the same rule.

This is what failed "recovers from an expired bearer during app bootstrap without
redirect looping": 9 of the 21 bootstrap 401s were delivered after the first
recovery finished, so the test counted a third refresh and a second recovery cycle.

* 🔊 test: Play Decodable Audio and Hold the Reply's Button in the Read-Aloud Replay Scenario

The fixture answered the speech request with four bytes the MP3 parser rejects, so
every playback failed. A failed playback cancels its request, and the audio is
cached only once the response has been read to the end, so the replay was served
from memory only when the response finished before the decoder gave up. When the
decoder won, the replay issued the second request the scenario forbids.

The test also located the button by its name. While a message is read aloud its
button is named "Stop", so the locator resolved to the user message's read-aloud
button: the wait for playback to end passed at once, and the second click could
land on a different message.

Serve a third of a second of silent MPEG frames so playback succeeds and ends on
its own, and locate the reply's button by its test id so the wait and the replay
address the same control.

* 📅 test: Wait for the Schedule Save to Land Before Reloading

The edit scenario clicked Save and reloaded the page in the next step. A reload
discards requests still in flight, so when it won the race the PATCH was cancelled
before the server answered and the edited schedule never appeared. Await the PATCH
response, and require it to succeed, before reloading.

* 🔗 fix: Grant a Legacy Shared Link's Permissions Once When Its First Views Race

The share page requests a link's config and its messages together, and each request
migrates a legacy link on the way in. Both find no grant and both write one. The
grant is an upsert matched on the principal and resource, which no unique index
covers, so the two writes inserted two rows: a second public entry, a second owner
entry, or both.

`grantPermission` now accepts the `_id` to give the entry if the grant creates it.
The migration derives that id from the link and the principal, so the second insert
collides on `_id` and is replayed as an update of the first entry. It holds across
processes because the database enforces it. Grants that pass no id are unchanged.

"manages a shared-link snapshot and preserves legacy public links through runtime
migration" asserts one public entry and saw two.

* 🔑 fix: Refresh When a Replayed Fetch Is Rejected With the Current Bearer

The fetch transport replayed a request whose bearer had been superseded and
returned whatever came back. A second 401 there means the current bearer is not
accepted either, which is the case the refresh exists for, and the axios path
already handles it that way. Replay first, then fall through to the ordinary
recovery when the replay is rejected too.

* 🔗 refactor: Derive a Unique Grant's Entry Id in the Data Layer

`AccessControlService.grantPermission` took the entry's `_id` as a Mongoose
`ObjectId`, which put a storage type in an exported `packages/api` signature. The
service now takes `uniqueEntry: true`, and `grantPermission` in data-schemas
derives the id from the principal and resource it already matches on. A caller can
no longer pass an id that names a different grant.
2026-11-11 16:22:39 -05:00
Danny Avila
9089d8cba8
🚮 fix: Require Sign-In After Missing OpenID Sessions (#15655) 2026-09-06 12:47:58 -04:00
MarcAmick
ade02054c8
🛟 fix: Keep File Uploads Alive With SSE Heartbeats (#14295)
Some checks are pending
Docker Dev Images Build / build (Dockerfile, librechat-dev, node) (push) Waiting to run
Docker Dev Images Build / build (Dockerfile.multi, librechat-dev-api, api-build) (push) Waiting to run
GitNexus Index / index (push) Waiting to run
GitNexus Index / post-index (push) Blocked by required conditions
Sync Locize Translations & Create Translation PR / Sync Translation Keys with Locize (push) Waiting to run
Sync Locize Translations & Create Translation PR / Create Translation PR on Version Published (push) Blocked by required conditions
Sync Helm Chart Tags / Ignore non-main push (push) Waiting to run
Sync Helm Chart Tags / Sync chart tags (push) Waiting to run
* fix: Use SSE to upload files in order to avoid idle timeouts.  Idle timeouts can occur for example from gateways and other services like cloudfare when uploading large files.  For example during rag processing the file is uploaded to librechat which then sends it to rag.  While librechat is waiting for the embeddings to come back from rag the file upload is sitting idle.  Gateways tend to want to cancel the upload with an http 408 , 504, or 524.  This change uses SSE to perform the upload so that while librechat is sending the file to rag, it consistently sends back a heartbeat event to the client to keep the connection alive.  This is especially useful when utilizing  EMBEDDING_BATCH_SIZE in librechat rag which will allow rag to process signifigantly larger files without running out of memory.

* added tests to packages\api\src\files\sse.spec.ts in order to test the new sse.ts

* fix: Harden SSE file upload lifecycle

* style: Sort data provider imports

---------

Co-authored-by: Marc Amick <MarcAmick@jhu.edu>
Co-authored-by: Danny Avila <danny@librechat.ai>
2026-07-21 19:27:09 -04:00
Marco Beretta
61016e328a
🔄 feat: Continue Shared Conversations as Personal Copies (#13714)
Adds a "Continue this chat" button to the shared conversation view that forks
the shared conversation into a new conversation owned by the viewer and opens it
to continue (issue #13001).

- POST /api/share/:shareId/fork, gated by requireJwtAuth, the fork rate
  limiters, and the canAccessSharedLink ACL (view access = fork access).
- forkSharedConversation clones from the anonymized getSharedMessages payload,
  so only share-visible data is copied.
- Strips file ids from cloned files/attachments so a fork grants no more file
  access than viewing the read-only share, and honors the global shared-file
  kill switch via the snapshotFiles option.
- Reduces the clone to the viewer's active branch, located by its index in the
  shared payload (shared ids are re-anonymized per request and createdAt can
  collide, while the payload order is stable).
- Resolves config/retention, persists, and reads back under the requesting
  user's tenant, not the share owner's; canAccessSharedLink also falls back to
  a system-wide share lookup so cross-tenant public shares resolve (ACL still
  enforced under the share's own tenant).
- Resolves a usable endpoint/model from the viewer's models config instead of
  hard-coding OpenAI, so deployments without OpenAI can send the first message.
- Routes the fork's 401s (logged-out or cold-loaded viewers) through login,
  including when the refresh itself is rejected for a stale session.
- Hides the Temporary Chat toggle once a conversation has a real id, and
  portals the share-settings theme/language dropdowns above the dialog.

Rebased onto dev; collapses the share-fork feature and its review fixes into a
single commit.
2026-06-24 16:27:01 -04:00
Danny Avila
c9180d1ad6
🎯 fix: Narrow Public Share 401 Bypass to the Share Endpoint Only (#12905)
Some checks are pending
Docker Dev Branch Images Build / build (Dockerfile, lc-dev, node) (push) Waiting to run
Docker Dev Branch Images Build / build (Dockerfile.multi, lc-dev-api, api-build) (push) Waiting to run
GitNexus Index / index (push) Waiting to run
GitNexus Index / post-index (push) Blocked by required conditions
2026-06-20 13:57:21 -04:00
Danny Avila
753e53eddd
🛬 fix: Coalesce Auth Recovery into a Single Refresh Flight (#13618)
* fix auth recovery singleflight

* add auth recovery e2e coverage

* handle invalid auth redirect timestamp
2026-06-09 12:04:12 -04:00
Danny Avila
9956a72694
🧭 fix: Subdirectory Deployment Auth Redirect Path Doubling (#12077)
* fix: subdirectory redirects

* fix: use path-segment boundary check when stripping BASE_URL prefix

A bare `startsWith(BASE_URL)` matches on character prefix, not path
segments. With BASE_URL="/chat", a path like "/chatroom/c/abc" would
incorrectly strip to "room/c/abc" (no leading slash). Guard with an
exact-match-or-slash check: `p === BASE_URL || p.startsWith(BASE_URL + '/')`.

Also removes the dead `BASE_URL !== '/'` guard — module init already
converts '/' to ''.

* test: add path-segment boundary tests and clarify subdirectory coverage

- Add /chatroom, /chatbot, /app/chatroom regression tests to verify
  BASE_URL stripping only matches on segment boundaries
- Clarify useAuthRedirect subdirectory test documents React Router
  basename behavior (BASE_URL stripping tested in api-endpoints-subdir)
- Use `delete proc.browser` instead of undefined assignment for cleanup
- Add rationale to eslint-disable comment for isolateModules require

* fix: use relative path and correct instructions in subdirectory test script

- Replace hardcoded /home/danny/LibreChat/.env with repo-root-relative
  path so the script works from any checkout location
- Update instructions to use production build (npm run build && npm run
  backend) since nginx proxies to :3080 which only serves the SPA after
  a full build, not during frontend:dev on :3090

* fix: skip pointless redirect_to=/ for root path and fix jsdom 26+ compat

buildLoginRedirectUrl now returns plain /login when the resolved path
is root — redirect_to=/ adds no value since / immediately redirects
to /c/new after login anyway.

Also rewrites api-endpoints.spec.ts to use window.history.replaceState
instead of Object.defineProperty(window, 'location', ...) which jsdom
26+ no longer allows.

* test: fix request-interceptor.spec.ts for jsdom 26+ compatibility

Switch from jsdom to happy-dom environment which allows
Object.defineProperty on window.location. jsdom 26+ made
location non-configurable, breaking all 8 tests in this file.

* chore: update browser property handling in api-endpoints-subdir test

Changed the handling of the `proc.browser` property from deletion to setting it to false, ensuring compatibility with the current testing environment.

* chore: update backend restart instructions in test subdirectory setup script

Changed the instruction for restarting the backend from "npm run backend:dev" to "npm run backend" to reflect the correct command for the current setup.

* refactor: ensure proper cleanup in loadModuleWithBase function

Wrapped the module loading logic in a try-finally block to guarantee that the `proc.browser` property is reset to false and the base element is removed, improving reliability in the testing environment.

* refactor: improve browser property handling in loadModuleWithBase function

Revised the management of the `proc.browser` property to store the original value before modification, ensuring it is restored correctly after module loading. This enhances the reliability of the testing environment.
2026-03-05 01:38:44 -05:00
Danny Avila
619d35360d
🔒 fix: Request interceptor for Shared Link Page Scenarios (#12036)
* ♻️ refactor: Centralize `buildLoginRedirectUrl` in data-provider

Move `buildLoginRedirectUrl` from `client/src/utils/redirect.ts` into
`packages/data-provider/src/api-endpoints.ts` so the axios 401
interceptor (and any other data-provider consumer) can use the canonical
implementation with the LOGIN_PATH_RE guard and BASE_URL awareness.

The client module now re-exports from `librechat-data-provider`, keeping
all existing imports working unchanged.

* 🔒 fix: Shared link 401 interceptor bypass and redirect loop (#12033)

Fixes three issues in the axios 401 response interceptor that prevented
private shared links (ALLOW_SHARED_LINKS_PUBLIC=false) from working:

1. `window.location.href.includes('share/')` matched the full URL
   (including query params and hash), causing false positives. Changed
   to `window.location.pathname.startsWith('/share/')`.

2. When token refresh returned no token on a share page, the
   interceptor logged and fell through without redirecting, causing an
   infinite retry loop via React Query. Now redirects to login using
   `buildLoginRedirectUrl()` which preserves the share URL for
   post-login navigation.

3. `processQueue` was never called in the no-token branch, leaving
   queued requests with dangling promise callbacks. Added
   `processQueue(error, null)` before the redirect.

* ✅ test: Comprehensive 401 interceptor tests for shared link auth flow

Rewrite interceptor test suite to cover all shared link auth scenarios:

- Unauthenticated user on share page with failed refresh → redirect
- Authenticated user on share page with failed refresh → redirect
- share/ in query params does NOT bypass the auth header guard
- Login path guard: redirect to plain /login (no redirect_to loop)
- Refresh success: assert exact call count (toBe(3) vs toBeGreaterThan)

Test reliability improvements:
- window.location teardown moved to afterEach (no state leak on failure)
- expect.assertions(N) on all tests (catch silent false passes)
- Shared setWindowLocation helper for consistent location mocking

* ♻️ refactor: Import `buildLoginRedirectUrl` directly from data-provider

Update `AuthContext.tsx` and `useAuthRedirect.ts` to import
`buildLoginRedirectUrl` from `librechat-data-provider` instead of
re-exporting through `~/utils/redirect.ts`.

Convert `redirect.ts` to ESM-style inline exports and remove the
re-export of `buildLoginRedirectUrl`.

* ✅ test: Move `buildLoginRedirectUrl` tests to data-provider

Tests for `buildLoginRedirectUrl` now live alongside the implementation
in `packages/data-provider/specs/api-endpoints.spec.ts`.

Removed the duplicate describe block from the client redirect test file
since it no longer owns that function.
2026-03-03 12:03:33 -05:00
Danny Avila
b18915a96b
🚪 fix: Complete OIDC RP-Initiated Logout With id_token_hint and Redirect Race Fix (#12024)
* fix: complete OIDC logout implementation

The OIDC logout feature added in #5626 was incomplete:

1. Backend: Missing id_token_hint/client_id parameters required by the
   RP-Initiated Logout spec. Keycloak 18+ rejects logout without these.

2. Frontend: The logout redirect URL was passed through isSafeRedirect()
   which rejects all absolute URLs. The redirect was silently dropped.

Backend: Add id_token_hint (preferred) or client_id (fallback) to the
logout URL for OIDC spec compliance.

Frontend: Use window.location.replace() for logout redirects from the
backend, bypassing isSafeRedirect() which was designed for user-input
validation.

Fixes #5506

* fix: accept undefined in setTokenHeader to properly clear Authorization header

When token is undefined, delete the Authorization header instead of
setting it to "Bearer undefined". Removes the @ts-ignore workaround
in AuthContext.

* fix: skip axios 401 refresh when Authorization header is cleared

When the Authorization header has been removed (e.g. during logout),
the response interceptor now skips the token refresh flow. This
prevents a successful refresh from canceling an in-progress OIDC
external redirect via window.location.replace().

* fix: guard against undefined OPENID_CLIENT_ID in logout URL

Prevent literal "client_id=undefined" in the OIDC end-session URL
when OPENID_CLIENT_ID is not set. Log a warning when neither
id_token_hint nor client_id is available.

* fix: prevent race condition canceling OIDC logout redirect

The logout mutation wrapper's cleanup (clearStates, removeQueries)
triggers re-renders and 401s on in-flight requests. The axios
interceptor would refresh the token successfully, firing
dispatchTokenUpdatedEvent which cancels the window.location.replace()
navigation to the IdP's end_session_endpoint.

Fix:
- Clear Authorization header synchronously before redirect so the
  axios interceptor skips refresh for post-logout 401s
- Add isExternalRedirectRef to suppress silentRefresh and useEffect
  side effects during the redirect
- Add JSDoc explaining why isSafeRedirect is bypassed

* test: add LogoutController and AuthContext logout test coverage

LogoutController.spec.js (13 tests):
- id_token_hint from session and cookie fallback
- client_id fallback, including undefined OPENID_CLIENT_ID guard
- Disabled endpoint, missing issuer, non-OpenID user
- post_logout_redirect_uri (custom and default)
- Missing OpenID config and end_session_endpoint
- Error handling and cookie clearing

AuthContext.spec.tsx (3 tests):
- OIDC redirect calls window.location.replace + setTokenHeader
- Non-redirect logout path
- Logout error handling

* test: add coverage for setTokenHeader, axios interceptor guard, and silentRefresh suppression

headers-helpers.spec.ts (3 tests):
- Sets Authorization header with Bearer token
- Deletes Authorization header when called with undefined
- No-op when clearing an already absent header

request-interceptor.spec.ts (2 tests):
- Skips refresh when Authorization header is cleared (the race fix)
- Attempts refresh when Authorization header is present

AuthContext.spec.tsx (1 new test):
- Verifies silentRefresh is not triggered after OIDC redirect

* test: enhance request-interceptor tests with adapter restoration and refresh verification

- Store the original axios adapter before tests and restore it after all tests to prevent side effects.
- Add verification for the refresh endpoint call in the interceptor tests to ensure correct behavior during token refresh attempts.

* test: enhance AuthContext tests with live rendering and improved logout error handling

- Introduced a new `renderProviderLive` function to facilitate testing with silentRefresh.
- Updated tests to use the live rendering function, ensuring accurate simulation of authentication behavior.
- Enhanced logout error handling test to verify that auth state is cleared without external redirects.

* test: update LogoutController tests for OpenID config error handling

- Renamed test suite to clarify that it handles cases when OpenID config is not available.
- Modified test to check for error thrown by getOpenIdConfig instead of returning null, ensuring proper logging of the error message.

* refactor: improve OpenID config error handling in LogoutController

- Simplified error handling for OpenID configuration retrieval by using a try-catch block.
- Updated logging to provide clearer messages when the OpenID config is unavailable.
- Ensured that the end session endpoint is only accessed if the OpenID config is successfully retrieved.

---------

Co-authored-by: cloudspinner <stijn.tastenhoye@gmail.com>
2026-03-02 21:34:13 -05:00