Cover the full litellm.rerank()/arerank() path with HTTP mocked, asserting
metadata.requester_metadata reaches the Discovery Engine :rank body as
userLabels (and stays absent when no metadata is set). Catches plumbing
regressions that unit tests on transform_rerank_request alone would miss.
- Introduced optional environment variables `PYROSCOPE_GRAFANA_USER` and `PYROSCOPE_GRAFANA_API_TOKEN` for Grafana Cloud integration.
- Updated documentation to reflect new configuration options for Pyroscope profiling.
- Enhanced error handling to ensure both credentials are provided when using Grafana Cloud.
- Added tests to validate Grafana Cloud authentication scenarios.
CI's license check fails on the new dev dep because liccheck cannot read
the PEP 639 'License-Expression' field that pytest-recording uses. Add
the package to the manually-verified allowlist (MIT, confirmed via PyPI
classifier).
Also addresses greptile P2 review comments:
- Add 'anthropic-version' to the request-header filter list so live and
mock recordings produce structurally identical cassettes.
- Replace the indentation-sensitive regex in
'_strip_nondeterministic_headers' with a YAML parse-and-rewrite so the
helper keeps working if vcrpy ever changes its serialization style.
Co-authored-by: Mateo Wang <mateo-berri@users.noreply.github.com>
Per Yuneng's feedback, use a single @pytest.mark.vcr marker so one record
sweep populates cassettes for every marked test across all providers,
instead of forcing each test to bind to a hard-coded cassette path.
Changes vs. the initial scaffolding:
- Add 'pytest-recording==0.13.4' on top of vcrpy. Adopt its layout:
cassettes live at 'cassettes/<test_module>/<test_name>.yaml', resolved
automatically. New tests just decorate with '@pytest.mark.vcr' — no
imports or path bookkeeping.
- Move the shared filter/match config into a 'vcr_config' fixture in
'tests/llm_translation/conftest.py' (consumed by pytest-recording for
every marked test in the dir). Drop the standalone 'vcr_config.py'.
- Bulk record / replay via the standard '--record-mode' CLI flag:
'make test-llm-translation-record' now sweeps every '@pytest.mark.vcr'
test under tests/llm_translation in one shot. Optional 'TARGET=' var
scopes to a single file.
- Move existing cassettes to the per-test paths and update the local
in-process Anthropic regenerator to write to the same paths.
- Refresh README + Makefile target docs to match the sweep workflow.
Co-authored-by: Mateo Wang <mateo-berri@users.noreply.github.com>
The wrapper had no production callers after transform_parsed_response
was refactored to call _resolve_json_mode_non_streaming directly.
Updated the parametrized test to call the underlying method.
Add explicit warnings when Presidio apply_to_output skips masking for mixed/unknown stream event shapes, and add regression coverage for the mixed stream flush path to ensure chunk order is preserved and warnings are emitted.
Made-with: Cursor
Keep Presidio apply-to-output streaming in passthrough mode when responses-style events are emitted, preventing dropped lifecycle events that caused Codex stream disconnects. Add a regression test to assert unknown streaming events are preserved in order.
Made-with: Cursor
The default-allow-GET fix in route_checks unblocked the route layer, but a
second class of bug remained: handlers that gate on `user_role !=
PROXY_ADMIN` (or a private `_require_proxy_admin` helper) reject admin
viewer at the handler before the route's HTTP method even matters.
Backend: relax handler role checks on read endpoints to allow
PROXY_ADMIN_VIEW_ONLY (same `_user_has_admin_view` helper used elsewhere).
- /v1/access_group GET (list) + /v1/access_group/{id} GET — split
`_require_proxy_admin` into a parallel `_require_admin_view` for the
two read handlers; writes (POST / PUT / DELETE) keep the strict gate.
- /cloudzero/settings GET, /vantage/settings GET — read-only views.
- /config_overrides/hashicorp_vault GET — read-only config view.
- /team/permissions_list GET — let admin viewer see permissions like
a Proxy Admin would.
- /jwt/key/mapping/list, /jwt/key/mapping/info — JWT mapping reads.
- /v1/mcp/discover, /v1/mcp/openapi-registry — MCP picker views.
- /schedule/anthropic_beta_headers_reload/status — read-only status.
- /adaptive_router/state — read-only live snapshot.
UI: hide write buttons that admin viewer should not see (button click
would fail the backend write gate, but the UX expectation is no button).
- Internal Users: hide "Invite User" button.
- Access Groups: hide "Create Access Group" + Delete row action.
- Budgets: hide "+ Create Budget" + Edit/Delete row actions.
- Prompts: hide "+ Add New Prompt" / "Upload .prompt File"; gate the
prompt-table Edit/Delete actions on `isProxyAdminRole` (was
`isAdminRole` which incorrectly included admin viewer).
- Router Settings → Fallbacks: hide AddFallbacks panel + per-row Test
+ Delete actions.
- AI Hub: hide "Select Models / Agents / MCP Servers / Skills to Make
Public" + "Useful Links Management" (writes).
These pages remain VISIBLE for admin viewer (read parity); only the
write entry points are hidden.
The redirect-following added to async_safe_get checks response.is_redirect
on every hop. Two vertex batch tests stub AsyncHTTPHandler.get with a bare
MagicMock, whose default-truthy is_redirect made the redirect path fire,
then crashed in httpx.URL().join() because headers.get('location') was
also a MagicMock instead of a string. Set is_redirect=False explicitly so
the mocked response models a non-redirect terminal response.
Also tighten _extract_redirect_url to raise SSRFError on non-string
Location values (defense-in-depth — a real httpx Response always returns
str|None, but this avoids a confusing TypeError if anything else ever
slips through).
This is an unrelated CI fix piggybacked on the SCIM PR to unblock the
batches test suite.
Root cause: admin_viewer_routes was an explicit allowlist, so every newly-added
GET endpoint anywhere in the codebase silently 403'd for admin viewer until
someone remembered to add it. We had whacked /spend/logs/ui, /customer/list,
/guardrails/list, /policies/attachments/list, /invitation/info, and several
others in serial — but the next round still surfaced /in_product_nudges,
/health/latest, /credentials, /v1/mcp/network/client-ip, /claude-code/plugins,
/policy/templates. This pattern keeps repeating because the model is wrong.
Structural fix in `_check_proxy_admin_viewer_access`:
- Default-allow safe HTTP methods (GET / HEAD / OPTIONS) on any
non-inference route. Admin Viewer's principle is read parity with
Proxy Admin; HTTP semantics already mark GET as side-effect-free, so
using the method as the allow signal is the correct primitive.
- Unsafe methods (POST/PUT/PATCH/DELETE) still go through the existing
explicit allowlists + the hard-blocked write set
(/user/new, /team/new, /key/generate, …).
- LLM/inference routes still 403 (cost-incurring).
The existing admin_viewer_routes list is retained as a backstop for the
small set of routes implemented as POST but semantically read (e.g.
/spend/calculate). Adding new GET endpoints no longer requires touching
this list.
Models page tab/panel off-by-one (UI bug for Admin Viewer):
Tremor's TabList filters falsy children but TabPanels does not, so
conditionally hiding "Add Model" with `{!shouldHideAddModelTab && ...}`
left a phantom panel slot — clicking "LLM Credentials" showed nothing,
and clicking "Pass-Through Endpoints" showed the credentials panel.
Refactor to a single source-of-truth `visibleTabs` array; tab and
panel indices now can never desync.
Tests:
- 12 parametrized tests covering the 6 user-reported endpoints + 4
hypothetical-future endpoints + 2 already-fixed ones, all asserting
Admin Viewer GET succeeds via the default-allow path (no allowlist
entry needed).
- 5 parametrized tests for POST writes still 403'ing
(random-future-write, /user/new, /team/new, /key/generate, /model/new).
- All 207 existing route_checks tests still pass — backward-compatible.
User reported six more 403s and "still restricts access to keys + models" after
the first round. Root causes:
1. Six read endpoints were missing from admin_viewer_routes:
- /guardrails/list, /v2/guardrails/list (Guardrails page)
- /guardrails/submissions, /guardrails/submissions/{guardrail_id}
- /guardrails/usage/overview (Guardrails Monitor page)
- /policies/attachments/list (Policies page)
- /get/mcp_semantic_filter_settings (Settings page)
2. /guardrails/submissions handler treated admin viewer as non-admin, filtering
them to only their team submissions. Switch to _user_has_admin_view() so
admin viewer sees all submissions (read parity with Proxy Admin).
3. UI Keys page (user_dashboard.tsx) and Models page (ModelsAndEndpointsView.tsx)
each had a hard "Access Denied" block specifically for "Admin Viewer" — a
leftover from the pre-parity era. Remove the blocks; gate the "Create Key"
button on the Keys page so admin viewer can read keys but not mint them.
Also drop the post-login redirect that forced admin viewers to /usage on
sign-in (page.tsx).
Tests:
- Extend ADMIN_VIEWER_SETTINGS_ROUTES parametrize list to cover all 7 new
routes (route-checks layer is now the layer production traffic actually
hits, vs. the dependency-override-bypass that was masking the gap).
The Prisma schema declares LiteLLM_VerificationToken.blocked as a
nullable Boolean with no default, so virtual keys created via the
key management endpoint persist with blocked=NULL. SQL equality
(`blocked = false`) never matches NULL rows, so the previous
`where={'blocked': not blocked}` filter silently skipped virtually
all real keys when SCIM tried to block them. This made SCIM
deprovisioning a no-op — and especially dangerous in DELETE flows
where the user row is removed afterwards, leaving orphaned but
fully-functional keys.
Match both `False` and `None` when blocking, and only `True`
when unblocking, so the state flip (and cache invalidation) actually
fires for the keys it should.
The openai SDK returns ResponseOutputMessage and ResponseOutputText as
raw Pydantic v2 models that lack .get() (unlike LiteLLM's own wrapper
objects). Add a _to_dict() helper that normalizes plain dicts,
BaseLiteLLMOpenAIResponseObject (has .get()), and raw Pydantic models
(has .model_dump()) into a consistent dict interface.
Two Greptile P2s addressed:
1. (security) The audit-log row for an ``add_team_callbacks`` call would
serialize the entire ``callback_vars`` block — including
``langfuse_secret_key``, ``langsmith_api_key``, and the GCS service
account path — verbatim into ``LiteLLM_AuditLogs``. Anyone with read
access to the audit table could harvest team callback credentials.
Same risk for ``disable_team_logging`` when the team's existing row
has populated ``callback_settings.callback_vars``.
Add ``_redact_callback_secrets``: deep-copies the metadata snapshot
and replaces every ``callback_vars`` value with ``***REDACTED***``.
The keys are kept so an auditor can still see *which* fields
changed. Applied to both before and after snapshots.
2. ``asyncio.create_task`` is fire-and-forget; if the audit-log write
raises (transient DB error etc.) the exception is silently
discarded by the event loop and the audit row is just missing —
the exact gap this PR is closing. Attach a ``done_callback`` that
logs the exception at warning level via ``verbose_proxy_logger`` so
the operator sees there's a gap.
Tests assert that callback values are not present in the serialized
audit payload (both for ``add_team_callbacks`` and for
``disable_team_logging`` when the team's existing row has
populated secrets).
The two mutating endpoints in team_callback_endpoints.py
(``/team/{id}/callback`` POST and ``/team/{id}/disable_logging``) wrote
team metadata without emitting an audit-log row. The disable variant
is the worst case: a logging-control action that itself isn't logged,
so an admin (or compromised admin) could zero out a team's
observability with no forensic trail.
Add ``_emit_team_callback_audit_log`` mirroring the
``store_audit_logs``-gated pattern already used in team_endpoints.py
for /team/new and /team/update. When ``litellm.store_audit_logs`` is
True, both endpoints now emit an ``LiteLLM_AuditLogs`` row capturing
the calling user, the API key, and the before/after team metadata.
When the flag is False the helper is a no-op, so non-Enterprise
deployments are unaffected.
The ``litellm_changed_by`` header is now also accepted on
``/team/{id}/disable_logging`` to match the existing
``add_team_callbacks`` shape; the header is optional so existing
callers are unaffected.
Variant scope: the file has three endpoints — both mutating variants
are now logged. The read-only ``GET /team/{id}/callback`` is unchanged.
Other unlogged callback / logging-control admin endpoints elsewhere in
the proxy (e.g. ``/cache/settings``, ``/config_overrides/hashicorp_vault``)
are out of scope here and would be addressed in a separate PR.
Tests cover both endpoints in both ``store_audit_logs`` states and
verify that the captured before/after metadata reflects the actual
mutation, plus that the ``litellm-changed-by`` header overrides the
auth user_id when supplied.
