* fix(proxy): infer team from DB when JWT has no team and user has one team
- When team_id is unset after JWT auth but the user row has exactly one
team, set team_id, team_object, and team_membership from DB.
- Skip when zero or multiple teams (ambiguous).
- Add parametrized unit tests in test_handle_jwt.py.
Made-with: Cursor
* fix(proxy): JWT single-team DB fallback: catch errors, tests match get_team_object
- Wrap get_team_object + get_team_membership in one try/except; log and skip on failure (stale/missing team id no longer fails auth).
- Parametrize tests: HTTP 404/500, membership error; use side_effect not return_value=None for missing team row.
Made-with: Cursor
* refactor(jwt): extract single-team fallback into _resolve_single_team_fallback helper
Made-with: Cursor
Greptile flagged a regression introduced in the previous commit's merged
exception handler: ``ProxyException.__init__`` normalizes ``code`` via
``str(code)``, so a ``code=None`` (valid per the type signature) becomes
the string ``"None"``. Coercing that with ``int(...)`` raises
``ValueError``, which propagates uncaught and rewrites the auth error as
an unhandled 500 — degrading security posture compared to the pre-merge
``str(e.code) in ("401", "403")`` shape.
Compare against both int and str forms of the auth-error codes instead
of coercing. Adds a regression test for the ``code=None`` case.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Two related issues in `MCPRequestHandler.process_mcp_request`:
1. Public-route detection used `".well-known" in str(request.url)`, a
substring match against the full URL. Attackers could smuggle the
marker via the query string, hostname, or a deeper path segment to
bypass authentication on any MCP route. Replaced with an exact path
prefix on `request.url.path` (`startswith("/.well-known/")`).
2. The OAuth2 passthrough fallback (added in #20602 to support
`auth_type=oauth2` upstream MCP servers like Atlassian) caught any
401/403 from `user_api_key_auth` and replaced the result with an
anonymous `UserAPIKeyAuth()`. That fallback fired regardless of the
target server's configured `auth_type`, so an attacker presenting a
garbage `Authorization` header could exchange a failed LiteLLM auth
for an anonymous session against any server. The fallback now runs
only when EVERY MCP server the request targets is operator-configured
for `auth_type=oauth2`. For any non-oauth2 server (api_key,
bearer_token, basic, etc.), the auth error propagates as before.
Target resolution prefers the `x-mcp-servers` header when present
(including the explicitly-empty case, which fails closed) and otherwise
parses the standard `/mcp/{server_name}` and `/{server_name}/mcp`
transport URL patterns. Routes that don't match either form fail closed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* feat(guardrails): add LLM_AS_A_JUDGE to SupportedGuardrailIntegrations
* feat(types): add EvalVerdict, StandardLoggingEvalInformation; wire eval_information into SpendLogsMetadata
* feat(guardrails): add self-contained llm_as_a_judge guardrail hook
* fix(a2a): filter agent-only litellm_params from acompletion kwargs; pass agent_id into body
* feat(ui): add LLMJudgeFields criteria builder component
* feat(ui): wire LLM-as-a-Judge into add guardrail form
* feat(ui): update EvalViewer — title 'LLM Judge Results', weighted score column, summary row
* fix(ui): wire EvalViewer into LogDetailContent to show LLM judge results on logs page
* fix(guardrails-ui): route llm_as_a_judge to criteria builder step; rename to LiteLLM LLM as a Judge; add litellm logo
* fix(guardrail-viewer): stack lifecycle + eval details vertically to avoid badge overflow in narrow drawer
* fix(guardrail-create): surface config validation errors on create instead of silently orphaning guardrail in DB
* fix(guardrail-registry): hardcode llm_as_a_judge in initializer registry so it loads regardless of package install path
* fix(llm-as-a-judge): fix P1 code quality issues - validate weights/on_failure, guard pre_call, handle multimodal, move imports to module level, fix spurious finally logging
* fix(guardrail_endpoints): use correct PK field in rollback delete and log rollback failure
* fix(llm_as_a_judge): support Pydantic object in _get_litellm_param fallback chain
* fix(LLMJudgeFields): replace @tremor/react Button with antd Button
* fix(llm_as_a_judge): remove dead registry dicts, fix KeyError in prompt builder, set correct status on judge failure
* test(llm_as_a_judge): add unit tests for guardrail hook
* fix(llm_as_a_judge): remove @log_guardrail_information decorator to fix duplicate guardrail_information entries
The decorator and the manual finally block both called add_standard_logging_guardrail_information_to_request_data, producing two entries per request. The decorator also misclassified HTTPException(422) blocks as guardrail_failed_to_respond (it checks for 400). The finally block correctly tracks status throughout, so removing the decorator is sufficient.
* fix(test_gcs_pub_sub): ignore metadata.eval_information in comparison
* fix(test_spend_management): ignore metadata.eval_information in payload comparison
* fix(types/guardrails): add input_type and messages to ApplyGuardrailRequest
* fix(guardrail_endpoints): pass input_type and messages through apply_guardrail endpoint
* fix(guardrail_endpoints): auto-detect post_call guardrails and use input_type=response
* fix(a2a_endpoints): merge agent litellm_params guardrails into data before post_call hooks
* fix(llm_as_a_judge): use float sum with tolerance for weight validation
* fix(guardrail_registry): split long import line for black formatting
* fix(llm_as_a_judge): guard guardrail_name Optional for mypy
* fix(llm_as_a_judge): set guardrail_status=guardrail_intervened when score fails, regardless of on_failure mode
* fix(a2a_endpoints): use try/finally so deferred spend log fires even when guardrail blocks with 422
* fix(litellm_logging): declare _defer_async_logging and _enqueue_deferred_logging on Logging class for mypy
* fix(logging_worker): restore queue.join() in flush() to wait for in-flight callbacks
User-configured pass-through endpoints with ``auth: false`` are
explicitly unauthenticated — the builder short-circuits at
check_api_key_for_custom_headers_or_pass_through_endpoints and returns
a fresh empty UserAPIKeyAuth() without an api_key, user_id, or role.
Pre-refactor, that empty token never reached common_checks. After the
centralization, it does — and common_checks rejects it as admin-only,
breaking every Langfuse / custom unauthenticated pass-through.
This is the same regression class as the public-routes one: a
builder fast-path whose return value cannot survive common_checks.
Honor the same contract here — when the matched endpoint config has
auth != True, skip the centralized gate. auth=True endpoints still
run the full gate (covered by a companion test).
No security regression: ``auth: false`` is the operator's explicit
opt-out from LiteLLM auth on this path. The original commit closed
seven authenticated bypasses; this exemption applies only to a path
the operator has already declared unauthenticated.
The passthrough helper copied the upstream provider's Server: header
(e.g. "cloudflare" from Anthropic) onto the FastAPI response. uvicorn
then added its own Server: header on top, producing two Server: lines
in the wire response. Strict HTTP parsers (aiohttp's, used in CI's
passthrough tests) reject this with "Duplicate 'Server' header found"
and the request fails with a 400.
Same risk for Date, Content-Length, Connection, Keep-Alive: the ASGI
server writes its own copy at serialization time. Forwarding the
upstream's value either duplicates the header or lies about the
re-serialized body length.
Drop these from the forwarded set. Application/business headers
(content-type, x-request-id, anthropic-ratelimit-*, etc.) still pass
through unchanged.
Two regressions introduced by 3737d6a1f3 (centralized common_checks):
1. Public routes (e.g. /health/readiness, /metrics) are exempted by the
builder fast-path but the wrapper then ran common_checks on the
synthetic INTERNAL_USER_VIEW_ONLY token, which has no user_id, no
team, no scopes — so common_checks rejected the request as admin-
only. This broke every k8s readiness probe when master_key is set
(helm chart job confirmed: pod never goes Ready, service has no
endpoints).
2. The admin user_object synthesis only triggered when
user_object is None. After any team-creation flow runs, the row
for litellm_proxy_admin_name (default "default_user_id") exists
in litellm_usertable with the default user_role=internal_user.
get_user_object then returned that row, the synthesis was skipped,
and master_key requests were demoted to internal_user — failing
/team/update, /team/block, etc. The token's user_role is the
source of truth for these paths (set inside the authenticated
master_key / JWT-admin builders); a stale DB row must not override
it.
Fix:
- Short-circuit _run_centralized_common_checks for routes already in
LiteLLMRoutes.public_routes (or general_settings.public_routes).
Same exemption surface the builder already trusts.
- When the token's user_role is PROXY_ADMIN, force the synthesized
admin user_object regardless of what get_user_object returned.
Preserves the spend value from the DB row.
Neither change reopens any of the seven bypasses the original commit
closed: OAuth2, JWT non-admin, DB-fallback, /user/auth, pass-through
headers, etc., still go through the gate. Only paths that were
already admin or already public skip it.
Adds two regression tests:
- test_centralized_common_checks_skips_public_routes
- test_centralized_common_checks_master_key_admin_overrides_db_user_role
* fix(vertex passthrough): log :embedContent and :batchEmbedContents responses
* test(vertex passthrough): add unit tests for :embedContent and :batchEmbedContents logging
* fix(vertex passthrough): extract input text from request body for embedContent token counting
* fix(vertex passthrough): add embedContent and batchEmbedContents to TRACKED_VERTEX_ROUTES
* fix(vertex passthrough): detect Google AI Studio URLs in embedContent handler
* test(vertex passthrough): add unit test for Google AI Studio URL embedContent provider detection
* style: black format vertex_passthrough_logging_handler
The strict 'additionalProperties: false' schema in
test_aaamodel_prices_and_context_window_json_is_valid rejected the new
flag added in this PR's earlier commit. Register it alongside the other
supports_*_reasoning_effort entries so the schema validation passes.
gpt-5.5-pro only accepts reasoning_effort in {medium, high, xhigh}
(verified live against OpenAI's API on 2026-04-24). LiteLLM previously
had no way to express this constraint — the existing JSON schema
covered none/minimal/xhigh but not low. Result: drop_params=true users
saw an avoidable 400 from OpenAI.
Add supports_low_reasoning_effort following the existing opt-out
pattern (default-allow, explicit false to block). Mirror the minimal
branch in OpenAIGPT5Config.map_openai_params so 'low' goes through the
same _is_reasoning_effort_level_explicitly_disabled gate.
Set the flag to false on gpt-5.5-pro and gpt-5.5-pro-2026-04-23 in
both model_prices JSON files (kept in sync). Other models leave the
key absent so behavior is unchanged.
Tests cover: rejection on pro variants (no drop_params), drop on pro
with drop_params=True, passthrough on gpt-5.5 chat, passthrough on
unknown models, and the helper-level _is_reasoning_effort_level_explicitly_disabled
contract.
Verified against OpenAI's live Chat Completions API on 2026-04-24:
POST /v1/chat/completions
{"model": "gpt-5.5", "reasoning_effort": "minimal", ...}
-> 400 Unsupported value: 'reasoning_effort' does not support 'minimal'
with this model. Supported values are: 'none', 'low', 'medium',
'high', and 'xhigh'.
POST /v1/chat/completions
{"model": "gpt-5.5-pro", "reasoning_effort": "minimal", ...}
-> 400 Unsupported value: 'minimal' is not supported with the
'gpt-5.5-pro' model. Supported values are: 'medium', 'high', and
'xhigh'.
Set supports_minimal_reasoning_effort=false on all four entries
(gpt-5.5, gpt-5.5-2026-04-23, gpt-5.5-pro, gpt-5.5-pro-2026-04-23) so
OpenAIGPT5Config._is_reasoning_effort_level_explicitly_disabled fires
and LiteLLM either drops the param (drop_params=True) or raises a
local UnsupportedParamsError, instead of round-tripping to OpenAI for
a 400.
Adds a parametrized test_gpt55_reasoning_effort_flags_match_live_openai_api
test that pins supports_{none,minimal,xhigh}_reasoning_effort on each
entry to OpenAI's actual API contract.
Note: gpt-5.5-pro additionally rejects 'none' and 'low'. 'none' is
already handled (supports_none_reasoning_effort=false). 'low' is not
representable in the current JSON schema (no supports_low flag);
filing separately.
* feat(openai): day-0 support for GPT-5.5 and GPT-5.5 Pro
Add pricing + capability entries for the new GPT-5.5 family launched by
OpenAI on 2026-04-24:
- gpt-5.5 / gpt-5.5-2026-04-23 (chat): $5/$30/$0.50 per 1M
input/output/cached input
- gpt-5.5-pro / gpt-5.5-pro-2026-04-23 (responses-only): $60/$360/$6
per 1M input/output/cached input
Other fees (long-context >272k, flex, batches, priority, cache
discounts) follow the same ratios as GPT-5.4, with context window
retained at 1.05M input / 128K output.
No transformation / classifier code changes are required:
OpenAIGPT5Config.is_model_gpt_5_4_plus_model() already matches 5.5+ via
numeric version parsing, and model registration is driven from the
JSON. The existing responses-API bridge for tools + reasoning_effort
(litellm/main.py:970) already covers gpt-5.5-pro.
Tests:
- GPT5_MODELS regression list now covers gpt-5.5-pro and dated variants
- New test_generic_cost_per_token_gpt55_pro cost-calc test
- Updated test_generic_cost_per_token_gpt55 for long-context fields
* fix(openai): mirror reasoning_effort flags onto gpt-5.5 dated variants
gpt-5.5-2026-04-23 and gpt-5.5-pro-2026-04-23 were missing the
supports_none_reasoning_effort, supports_xhigh_reasoning_effort, and
supports_minimal_reasoning_effort flags that their non-dated
counterparts define. Reasoning-effort routing in OpenAIGPT5Config is
fully capability-driven from these JSON flags — since an absent flag
is treated as False for opt-in levels (xhigh), users pinning to a
dated snapshot would silently lose xhigh support and diverge from the
base alias on logprobs + flexible temperature handling.
Copy the flags onto both dated variants so every dated snapshot
inherits the base model's reasoning-effort capability profile.
Adds a parametrized regression test that asserts
supports_{none,minimal,xhigh}_reasoning_effort parity between each
dated variant and its non-dated counterpart, preventing future drift
when new snapshots are added.
create_model does not inherit the base class docstring, so once an
extension registered a field the effective class had no description.
The UI renders schema.description as a header paragraph — losing it
broke the 'Configuration for UI-specific flags' text. Pass __doc__
through explicitly and add a regression test.
Extract the admin team-header attachment into a helper so
auth_builder stays under the 50-statement lint threshold; apply
black formatting to the two files flagged on the prior commit.
No behavior change.
Add _depth/_max_depth guards (default 20) so the nested dict masking
cannot run away, and allowlist the function in the recursive_detector
CI check alongside the other bounded recursive helpers.
`_check_byok_credential` previously returned silently when `prisma_client`
was None, bypassing BYOK ownership validation during database-outage
windows. Any proxy-authenticated user could invoke BYOK-protected MCP
tools without a stored credential during the outage window.
Now raises HTTP 503 with a structured error so the flow fails closed.
Regression test asserts 503 is raised when `prisma_client` is None.
Reported by @brodmart in GHSA-6762-2m23-5mxp.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Adds three new UISettings flags gated by check_org_admin_feature_access:
- disable_key_generate_for_org_admin
- disable_team_create_for_org_admin
- disable_model_add_for_org_admin
When enabled by a proxy admin, users with the ORG_ADMIN role receive a 403
on /key/generate, /team/new, and /model/new respectively. All other roles
(proxy admin, internal user, team) are unaffected and continue through
their existing auth checks. Flags are persisted to litellm_uisettings and
synced into general_settings via _RUNTIME_GENERAL_SETTINGS_FLAGS so the
enforcement helper can read them at request time.
Scope the header-driven team fetch to LLM API routes so admin
management routes keep the pre-existing bypass behavior (no
phantom teams, no 404s on mgmt calls). Team context is threaded
onto UserAPIKeyAuth so spend logs, rate limits, and team_models
attribution are correctly applied when admins act on behalf of
a team via x-litellm-team-id.
* fix(proxy): honor object_permission for managed vector store access
* perf(proxy): preload team object_permission on UserAPIKeyAuth
Populate team_object_permission during virtual-key and JWT auth when the
team is loaded, so can_user_access_vector_store uses it in memory first
and only falls back to get_object_permission by id when missing.
Made-with: Cursor
Three concerns raised by bot reviewers, all addressed:
1. CodeQL cyclic-import warning
``experimental_pass_through/transformation.py`` imported from the
parent ``..transformation`` module, which CodeQL flagged as a
potential cycle. Extracted the helper into a new leaf module
``vertex_ai_partner_models/anthropic/output_params_utils.py`` that
has no heavy imports of its own. Both transformation files now
import from it cleanly. Renamed the helper from the underscore-
prefixed ``_sanitize_vertex_anthropic_output_params`` to the
public ``sanitize_vertex_anthropic_output_params`` since it is now
shared across modules.
2. Greptile P2: redundant ``None`` guard on ``extra_kwargs``
``handler.py`` had two ``extra_kwargs = extra_kwargs if ... else {}``
coercions; the second was a no-op because line 220 already
coerced. Removed the second one and added a NOTE comment so future
readers understand ``extra_kwargs`` is guaranteed non-None at the
point of use.
3. Greptile P2: misleading "already translated" docstring
The docstring claimed the translator above mapped
``output_config.format`` to ``response_format``, but Greptile
correctly traced the code and found that only the legacy top-level
``output_format`` was being translated — ``output_config.format``
was being silently dropped on the adapter path. Two-part fix:
a. Code: extended ``_translate_output_format_to_openai`` to accept
both shapes (top-level ``output_format`` AND
``output_config.format`` sub-key). Top-level still takes
precedence when both are supplied. This means callers using the
newer Anthropic Structured Outputs API now have their schema
properly forwarded to non-Anthropic backends as
``response_format``.
b. Tests: rewrote the misleading docstring to describe what
actually happens, plus added two new tests:
* ``test_output_format_top_level_still_translates`` —
regression guard for the legacy path
* ``test_output_format_takes_precedence_over_output_config_format``
— documents the precedence rule explicitly
Tests: 28/28 pass (was 26/26 before; +2 for the new translation
behavior + precedence). All run in ~0.5s, no real network calls.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Resolves the silent strip of Anthropic Structured Outputs across the
Vertex AI Claude transformation paths and the Anthropic-adapter
re-merge. Consolidates and supersedes four stalled community PRs
addressing overlapping aspects of the same root bug:
- #23475 (Vertex AI Claude blanket-strip removal)
- #23396 (Vertex AI Claude conditional passthrough)
- #23706 (Anthropic adapter exclude output_config from non-Anthropic
backends)
- #22727 (Anthropic adapter strip output_config for non-Anthropic
backends)
Closes / addresses: #23380 (Vertex AI Claude output_config drop),
related: #26423, #25079, #24549, #25971, #25957, #26163, #24856.
What was broken
---------------
* Vertex AI Claude paths called ``data.pop("output_config")`` and
``data.pop("output_format")`` unconditionally even when Vertex
accepted those fields. Callers asking for Structured Outputs got a
200 with prose and never knew the schema constraints had been
silently dropped (often masked for months by permissive fallback
parsers).
* The ``/v1/messages`` -> ``/chat/completions`` adapter
(``LiteLLMMessagesToCompletionTransformationHandler``) re-merged the
raw Anthropic-shaped ``output_config`` into ``completion_kwargs``
AFTER the translator already mapped its meaningful parts to
``response_format`` / ``reasoning_effort``. Non-Anthropic backends
(Azure OpenAI, Fireworks, Bedrock Nova, etc.) then 400'd with
"Extra inputs are not permitted".
Approach
--------
Vertex AI Claude (chat-completion + experimental_pass_through paths):
Replace the unconditional pop with a sanitizer
``_sanitize_vertex_anthropic_output_params`` that strips only the
Vertex-unsupported keys (today: ``effort``) from ``output_config``
while forwarding ``format`` and the legacy top-level
``output_format``. Defensive: non-dict ``output_config`` values are
dropped to avoid sending malformed payloads downstream.
Greptile P1 from PR #23396 addressed: when ``output_config`` carries
both ``format`` and ``effort``, the prior conditional pass-through
forwarded ``effort`` and reproduced the 400. The new helper filters
per-key.
Anthropic ``/v1/messages`` adapter:
Add ``output_config`` to a named module-level constant
``ANTHROPIC_ONLY_REQUEST_KEYS`` and wire it into ``excluded_keys`` so
the post-translation re-merge skips re-adding the raw key. This
fixes the 400 on non-Anthropic backends and avoids the conflicting
duplicate (``response_format`` + raw ``output_config``) on
Anthropic-family backends.
Greptile P2 from PR #23706 addressed: the constant gives reviewers
one grep target instead of an inline literal that silently grows.
Greptile P2 from PR #22727 addressed: ``extra_kwargs or {}`` is
replaced with explicit ``is None`` checks so empty-dict callers no
longer skip the fallback path.
Tests
-----
* tests/test_litellm/llms/vertex_ai/vertex_ai_partner_models/anthropic/
test_vertex_ai_partner_models_anthropic_transformation.py:
- 5 new/updated cases plus a direct unit test for
``_sanitize_vertex_anthropic_output_params``.
- Updated ``test_vertex_ai_claude_sonnet_4_5_structured_output_fix``
so its mock-injected ``output_format`` is asserted to FLOW THROUGH
(the original test asserted the now-buggy strip behavior).
* tests/test_litellm/llms/anthropic/experimental_pass_through/
adapters/test_handler_output_config_passthrough.py (new):
- Constant export sanity, output_config strip with ``effort`` only,
output_config strip with ``format`` only, regression guard that
unrelated extras still flow, explicit-empty-dict path, and the
``extra_kwargs=None`` no-crash path.
Test-quality fixes incorporated from Greptile review on the
superseded PRs:
* No ``inspect.getsource`` source-text assertions (PR #24114 / #23475).
* ``sys.path`` insertion is anchored to ``__file__`` (PR #23706).
* Assertion messages are positional, not tuple (PR #24114-class bug).
* No ``or {}`` masking explicit empty dicts in helper signatures
(PR #22727).
Verified locally: 26/26 pass with this commit. The new tests
fail (or fail to import) on ``main`` without it.
Out of scope
------------
* The ``max_tokens`` capping logic from PR #22727 — independent
concern, deserves its own PR with a focused test plan.
* Architectural rework of the ``excluded_keys`` mechanism (Greptile
P2 on PR #23706 noted point-fix growth). The named constant gives
maintainers a clear place to extend; a registry-based approach
would be a follow-up.
Co-Authored-By: netbrah <netbrah>
Co-Authored-By: s-zx <s-zx>
Co-Authored-By: invoicepulse <invoicepulse>
Co-Authored-By: cfdude <cfdude>
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(team_endpoints): auto-add SSO team members to org for proxy admins
* test: proxy_admin vs team_admin security boundary for team→org move
* screenshots: before/after for team-org SSO fix
* fix(team_endpoints): restore staging security features dropped in SSO commit
Co-Authored-By: Ishaan Jaff <ishaan@berri.ai>
* style: black formatting for team_endpoints
Unit Tests: Proxy DB Operations / proxy-db (auth-checks, tests/proxy_unit_tests/test_auth_checks.py tests/proxy_unit_tests/test_user_api_key_auth.py, 20, 8) (push) Has been cancelled
Unit Tests: Proxy DB Operations / proxy-db (remaining, tests/proxy_unit_tests --ignore=tests/proxy_unit_tests/test_key_generate_prisma.py --ignore=tests/proxy_unit_tests/test_auth_checks.py --ignore=tests/proxy_unit_tests/test_user_api_key_auth.py --ignore=tests/proxy_unit_tests/test_p… (push) Has been cancelled
