A budget bypass that ships off by default stays open for every deployment
that does not know to look for the flag, so `enforce_fallback_budget` now
defaults to true and `general_settings.enforce_fallback_budget: false` is
the opt-out for anyone who wants the old unguarded behaviour back.
BREAKING CHANGE: a paid fallback target is now refused for callers who are
over their key or user `max_budget`. Deployments relying on fallbacks to
keep serving over-budget callers must set enforce_fallback_budget: false.
The failure hook received data without the resolved id when body parsing or
add_litellm_data_to_request raised, so proxy-only spend logging minted a fresh id
that did not match the error log or the x-litellm-call-id header. The id is now
part of the request data from the start and merged over the parsed body, which
also removes the post-hoc in-place assignment
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Budget is enforced once during auth, against the requested model group.
`_is_model_cost_zero` waives every budget check for a zero-cost group, and the
router then picks a fallback target afterwards, inside `run_async_fallback`,
where nothing re-checks budget. A free model with a paid fallback therefore
bills with no budget gate at all.
Add `fallback_budget_check`, the budget sibling of the existing
`fallback_access_check`: a predicate awaited per fallback target that skips
targets the caller cannot pay for. The primary attempt is untouched, so a
zero-cost model is never blocked by budget and only the paid fallback is
refused.
Counter reads pass `max_budget` so `get_current_spend` verifies against
authoritative recorded spend, matching the auth-time key and user checks; a
counter restored from an older snapshot reads as a hit rather than a clean
miss, so without it a stale-low value would keep admitting paid fallbacks.
A zero-cost fallback target is always allowed, and a team key does not inherit
the key owner's personal budget unless `apply_user_budget_to_team_keys` is set,
matching `_PROXY_MaxBudgetLimiter`.
Scope is key and user budgets. Team, team-member, end-user, org, global and
per-model budgets are not covered yet: those auth-path functions enforce rather
than report, so reusing them would fire threshold alerts and take spend
reservations for a target that is then skipped. Two limitations of that scope
are documented in the module docstring: the check reads the spend counter
rather than reserving against it, so concurrent fallbacks can cross a cap
together; and a request reaching the router without
`metadata["user_api_key_auth"]` is not restricted. Both are shared with
`fallback_model_access.py`.
Opt-in via `general_settings.enforce_fallback_budget`.
Relates to #41344
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Already shaped ProxyException and HTTPException errors passing through the moderations, audio speech, Anthropic Messages, and handle_exception_on_proxy paths now answer with the x-litellm-call-id header the route logged under, without overwriting a header the exception was raised with. The GET /v1/batches failure hook receives the resolved request data so the spend log request_id matches the response header and the error log
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Drop the model_access_denied_message setting, its {model} template, the DB
override entry and the Admin UI field. Model access denials now always return
the fixed client message while the allowlist diagnostic is logged at the final
HTTP, realtime and MCP boundaries
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Post-auth denials from can_key_call_resolved_model (per-request alias
rewrite, MCP sampling, realtime) never reach the auth exception handler,
so the internal allowlist reason was dropped when
model_access_denied_message was set. Log it once from the ProxyException
response handler and the realtime rejection path instead, and convert
JWT ModelAccessDeniedHTTPException into the specialized ProxyException so
the same boundary covers it
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Add litellm_settings.model_access_denied_message, a template ({model} placeholder) returned to clients instead of the detailed "can only access models=[...]" text on key/team/user/org/project and team-member model access denials. The full denial reason is still written to the proxy logs at WARNING. Unset keeps the existing detailed message, status codes and error types are unchanged.
Expose the new setting and the existing expose_router_debug_in_errors flag in the Admin UI general settings (String editor, Boolean toggle with an explicit True default) and allow both as safe DB overrides so they persist and propagate across workers.
Resolves LIT-5283
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Releasing the slot unconditionally from the route raced the limiter's own
success handler on the logging worker: both could read the same stashed
acquisition before either cleared it, and under the integer in-memory
fallback that double-decrements the counter. The route now releases only on
exits without a success callback (pre-call rejection, pre-call cancellation,
and Phase 2 exits without the success stamp), matching the HTTP disconnect
path's ownership rule.
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf(auth): prefetch user, team, membership, org and project in one MGET, one query and one pipeline
Auth read each object with its own Redis GET and, on a miss, its own DB
query, then the admission spend counters with one GET each. The prefetch
warms every entry the checks read with one MGET, one raw query for the
Redis misses and one pipeline write, and a per-request batch serves the
spend counter reads from one MGET. The per-object getters stay the
readers and the fallback, so enforcement does not depend on the prefetch
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(auth): keep prefetch and spend batch collections immutable
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf(auth): let the cold spend-counter reseed reuse the admission MGET instead of one GET per counter
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf(auth): prefetch referenced auth objects only after the key's model access check passes
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(auth): give the prefetch-ordering test's patches their test-quality reasons
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(auth): move the real-Postgres prefetch join test to the proxy_behavior shard
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(auth): read NULL nested permission and budget lists as [] in the prefetch join
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf(proxy): batch post-call spend counter reads and carry budget state through the request
Post-call warm checks, reservation reads and reconcile reads for one request now go through a task-local spend counter batch: one MGET answers every counter, successful increments write their result back into the batch so no second Redis read follows, and invalidation forgets the key. RedisCache.async_increment sends INCRBYFLOAT and its TTL command in one pipeline round trip.
Auth pins frozen team, user and org budget snapshots on UserAPIKeyAuth, the pre-call setup writes them into the request metadata, and Prometheus reads them back instead of calling get_key_object, get_team_object, get_user_object and get_org_object on the response path. The getters stay as the fallback for requests that carried nothing (custom auth, unauthenticated routes, skipped checks).
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf(proxy): reconcile the budget reservation and the post-call warm checks from one MGET and one pipeline
A scope opened inside an open spend counter batch binds into it instead of starting its own, so the reservation reconcile and the post-call warm checks share the request's single MGET. The reconcile reads every reserved counter concurrently, sends the consistent adjustments in one INCRBYFLOAT+EXPIRE pipeline and settles a flushed or reseeded counter on its own afterwards, keeping the pre-call resize fail-closed. PendingSpendIncrement moves to spend_counter_batch so budget_reservation can build a pipeline without importing a private name
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* chore(proxy): drop the dataclass import left behind by the PendingSpendIncrement move
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(types): import Self from typing_extensions so the proxy imports on Python 3.10
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): use a neutral organization alias in the carried budget state tests
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): cover recorded and forgotten spend counter values in the request batch
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(caching): assert async_set_cache_pipeline_with_ttls keeps per-entry TTLs
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* refactor(proxy): type the reservation entry carried through reconcile adjustments
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(auth): map the model table's aliases column to model_aliases in the prefetch join and read user memberships the way get_user_object does
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* feat(proxy): make the in-memory management cache capacity configurable
Add general_settings.user_api_key_cache_max_size (positive int, default 200) to resize the
in-memory tier of the shared user_api_key_cache at startup and on DB config reloads, expose it
in the Admin UI general settings, and cover it with behavioral tests. Prior art: #34726
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(caching): resize the in-memory tier from DualCache so any cache instance honours the cap
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style(proxy): wrap the cache capacity field description to the 120 col limit
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* perf(proxy): pipeline spend counter increments into one redis call
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): apply surviving spend increments before raising scope error
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* style(proxy): ruff format spend counter helpers
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): settle inner spend counter gathers and fall back per key on pipeline failure
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): suppress BLE001 on pipeline fallback catch
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): invalidate all batched spend counters on pipeline failure
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): keep team member budget enforced at the cap and across Redis counter expiry
The cached-key auth path admitted a request when the team member counter sat exactly at max_budget, and a Redis counter that expired during a long stream was reconciled against this pod's stale in-memory copy, driving the shared counter negative and reopening the budget. Reject at >= like every other budget check, read Redis before the per-pod copy when judging the reconcile delta, and add the settled request cost after a DB reseed since reserved keys skip the normal increment
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* test(proxy): reseeded reservation counter also carries the settled request cost
The reseed test asserted counter == DB floor. The floor is read before the async spend flush writes this request, so the counter now lands at floor plus settled cost, matching the after leg in the PR proof
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: yassin <yassin@berri.ai>
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): eagerly initialize string callbacks
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): materialize string callbacks after load_config so later litellm_settings keys are applied
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
* fix(proxy): drop casts when snapshotting string callbacks so LIT006 stays at base
Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
---------
Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com>
Co-authored-by: yucheng <yucheng@berri.ai>
* fix(proxy): load db credentials inside the model reconcile so a worker never serves a model before its credential
* fix(proxy): load db credentials in the model read-through so a request miss never adds a model before its credential
* fix(proxy): read credentials from the writer db before the router update and look a credential up once
* test(proxy): assert the credential is loaded when db models reach the router instead of the call order
The failure hook's usage estimate and the project ITPM reservation both
called litellm.token_counter inline on the event loop, so a large request
that failed or hit the limiter stalled the gateway the same way the
count_tokens endpoints did. Both now run through asyncify. The loop-lag
probe the existing tests used moves into a shared helper that warms the
tokenizer first, and two new tests fail when either count runs inline
Both proxy token counting endpoints already count in a worker thread, but the
HuggingFace tokenizer's encode holds the GIL for the whole call, so a 600k-token
count on a Claude model still froze the event loop for up to 0.8 s and every
other request with it. Count through encode_batch_fast, which releases the GIL,
and tokenize at most TOKEN_COUNTER_MAX_EXACT_CHARS characters of any one string
(default 4,000,000), scaling the exact count of that prefix by the string's
length above it so the largest payloads stay bounded.
* feat(proxy): resolve root_path per request from SERVER_ROOT_PATHS
One deployment can encode exactly one client-visible URL path prefix
today: SERVER_ROOT_PATH is a scalar stamped onto the app at startup, so
a pod fronting several ingress prefixes 404s every prefix but one before
any handler runs, and MCP OAuth discovery can emit only one prefix's
URLs (RFC 9728 section 3 exact-match fails for the rest).
Add an opt-in outermost ASGI middleware that matches the request path
against a configured prefix list (SERVER_ROOT_PATHS, comma-separated) on
a segment boundary and sets scope["root_path"] for that request only.
Everything downstream is stock Starlette: route matching strips
root_path so routes stay registered root-relative, and request.base_url
re-includes it, so the discovery documents' resource and the 401
challenges' resource_metadata land under the prefix the client actually
called — with no discovery-builder changes.
LazyFeatureMiddleware now strips the scope root_path (falling back to
the cached SERVER_ROOT_PATH scalar) before feature prefix matching, so
lazily-registered routers — the MCP OAuth discovery router among them —
load under per-request prefixes.
Follow-up to the routing discussion on #35226; composes with, but does
not depend on, #35576.
* fix(proxy): import Sequence from collections.abc (ruff UP035 strict-budget gate)
* review(greptile): trim implementation commentary; fixture-own MCP registry state in tests
Addresses both P2s from the first Greptile pass:
- per_request_root_path_middleware.py (and the related _lazy_features /
proxy_server comments) cut down to the constraints the code cannot
express, per repo comment guidance
- the new discovery tests no longer clear/repopulate the shared MCP
registry inline; a fixture snapshots it, hands the test an empty
registry, and restores it afterwards so no state leaks between cases
* fix(lint): mutable-ok marker on the prefix accumulator (LIT002 type-discipline gate)
* fix(proxy): tie 401 challenges and get_custom_url to the per-request root_path
The per-request root_path middleware sets scope["root_path"] to the
prefix the client actually called, but the OAuth 401 challenges
(raise_user_oauth_challenge / raise_token_exchange_challenge) still
built their resource_metadata from SERVER_ROOT_PATH. On a pod fronting
several prefixes, the challenge advertised a discovery URL under a
different prefix than the discovery document served — the two
disagreed on where the resource metadata lives, and a strict RFC 9728
client refused the challenge. Route the challenges through a small
ContextVar the middleware populates so they read the same effective
root_path Starlette resolves the request under.
The same accessor fixes get_custom_url: when a request lives under a
SERVER_ROOT_PATHS-matched prefix, request.base_url already carries it,
so appending the SERVER_ROOT_PATH scalar on top produced e.g.
/tenant-a/legacy/sso/callback — a path that does not exist. Reading
the per-request prefix instead (and relying on join_paths's tail-dedup)
keeps SSO login/callback URLs under one prefix — the one the request
actually arrived on.
Fallback: outside a request (module-load-time UI URL builders,
background tasks) the ContextVar is unset and the accessor reads
SERVER_ROOT_PATH, matching get_server_root_path() so scalar-only
deployments are byte-identical.
* fix(mcp): challenge URL under per-request prefix must route, and mock parity
Two follow-ups to the review fix that made the 401 challenge use the
per-request root_path:
1. oauth_protected_resource_path must pick the URL structure that
actually routes for the mechanism in use:
- The scalar SERVER_ROOT_PATH deployment registers the well-known
routes with the prefix INSERTED (via well_known_root_suffix at
import time), matching RFC 8414 §3. The challenge URL must use the
same insertion or a client fetching it 404s.
- The per-request SERVER_ROOT_PATHS deployment can't register routes
per prefix; PerRequestRootPathMiddleware strips the prefix from
scope["path"] and the router matches the un-inserted route. The
URL must place the prefix BEFORE .well-known so the strip leaves a
matching path.
The previous fix used the insertion form for both, which 404'd the
discovery fetch on the per-request path — the discovery doc and the
challenge would then disagree on where the resource metadata lives,
the very failure the review flagged. End-to-end verified: the URL
the challenge advertises routes and the doc's `resource` field
equals the URL the client originally called (RFC 9728 §3).
2. get_request_root_path now delegates its fallback through
get_server_root_path() instead of reading the env directly, so every
existing `monkeypatch.setattr("litellm.proxy.utils.get_server_root_path"`
test override keeps working. This unstubbed the mock on the /v2/login
test that failed on the last CI run.
Plus the lint budget: annotate the local accumulator Final, tag the
scope["root_path"] rewrite as an intentional ASGI-contract mutation,
tag the reused `path`/`root_path` rebinds in LazyFeatureMiddleware, and
add reason strings to the two new PLC0415 lazy-import noqas.
* test(mcp): pin the reviewer's expected end-state — challenge URL routes, resource matches called URL
End-to-end regression test that mounts the discoverable router + the
per-request root_path middleware, hits an MCP endpoint that raises
raise_user_oauth_challenge, fetches the resource_metadata URL the
challenge advertises, and checks the returned document's `resource`
equals the URL the client originally called (RFC 9728 §3 exact match).
Covers /tenant-a, /tenant-b, and the unprefixed path on the same app so
a regression on any prefix — challenge URL 404s, or doc emits a
different prefix than the client called — fails at this test rather
than in a strict MCP client's discovery.
---------
Co-authored-by: gym-cmd <186399764+gym-cmd@users.noreply.github.com>
A slow or unreachable counter store made a pre-relay rejection wait behind
the reservation release before the client saw the error event and the close.
Close first and release in finally, mirroring the relay's own failure path,
so a client that already hung up still gets its reservation released.
The three pre-relay exits of realtime_websocket_endpoint (missing model,
key/model access denied, pre-call rejection such as a rate limit or a
guardrail) returned before the finally that releases the auth-time budget
reservation, so a rejected session pinned the key at the reserved amount
until the counter TTL expired and its next requests got budget_exceeded
while /key/info showed spend 0. A single _reject_realtime_session helper
now releases the reservation before sending the error event and closing,
and release_or_invalidate_budget_reservation shields the release from a
second cancellation and logs, rather than raises, a failing invalidate
fallback so it can never mask the session's own outcome.
The blanket finally release from the previous commit also zeroed the reservation
of successful sessions. Success settlement is enqueued on the logging worker, not
awaited, so the endpoint's finally ran first and released the reservation the cost
callback still had to reconcile, dropping the real spend from the key/team/user
counters.
The relay now stamps a synchronous marker (REALTIME_SESSION_SUCCESS_LOGGED_KEY) on
the shared logging object at the single success-dispatch site, and the endpoint
releases the reservation only when that marker is absent. Refused or failed
sessions, which never log success, still release; successful sessions leave the
reservation for the cost callback to settle to actual spend. Exactly one settler
touches each reservation, so the idempotent reconcile never double-adjusts.
A refused or failed /v1/realtime session never ran the success cost callback
or a failure hook, so its pre-call budget reservation stayed open and kept the
key/team/user spend counters pinned above real spend, 429ing later requests on
the same key until the counter's TTL expired. The endpoint now reconciles the
reservation in a finally, reusing a shared release_or_invalidate_budget_reservation
helper that mirrors the success/failure paths (release to zero, else invalidate
the reserved counters and finalize).
The relayed upstream close message and reason also go through the proxy's
client-facing redaction, so a credential, internal hostname, private IP, or
server path echoed by the upstream never reaches the client verbatim.
