From 7015bf37bb3830dab79b3efd886eee9b5b7ffad6 Mon Sep 17 00:00:00 2001 From: ryan-crabbe-berri Date: Sat, 5 Sep 2026 17:32:17 -0700 Subject: [PATCH 01/26] fix(proxy): apply team model aliases on the JWT auth path Team reads on the auth path never loaded the team's alias table, and the team-based JWT branch copied a hand-picked subset of team fields onto UserAPIKeyAuth, so aliases (and a few other team grants) never reached JWT callers: restricted teams 403'd alias requests and open teams 400'd them Load the alias relation where the team row is read and cached, project the team onto every team_* token field through one shared team_grants helper used by both JWT returns, and keep the relation when team model add/delete rewrites the cached team. ui_sso reuses the shared alias table model Resolves LIT-5858 Claude-Session: https://claude.ai/code/session_01EX13mWex6RaBo9PYnkAtFW --- basedpyright-code-budget.json | 2 +- litellm/proxy/auth/auth_checks.py | 9 +- litellm/proxy/auth/handle_jwt.py | 5 +- litellm/proxy/auth/team_grants.py | 122 +++++++++++++++++ litellm/proxy/auth/user_api_key_auth.py | 22 +-- .../management_endpoints/team_endpoints.py | 4 +- litellm/proxy/management_endpoints/ui_sso.py | 23 +--- .../proxy/auth/test_auth_checks.py | 64 +++++++++ .../proxy/auth/test_handle_jwt.py | 52 +++++++ .../proxy/auth/test_team_grants.py | 129 ++++++++++++++++++ .../proxy/auth/test_user_api_key_auth.py | 116 ++++++++++++++++ .../test_team_endpoints.py | 44 ++++++ type-discipline-budget.json | 2 +- 13 files changed, 547 insertions(+), 47 deletions(-) create mode 100644 litellm/proxy/auth/team_grants.py create mode 100644 tests/test_litellm/proxy/auth/test_team_grants.py diff --git a/basedpyright-code-budget.json b/basedpyright-code-budget.json index 0b0a61192e6..8bdc251c684 100644 --- a/basedpyright-code-budget.json +++ b/basedpyright-code-budget.json @@ -105,7 +105,7 @@ "limit": 109 }, "reportUnknownMemberType": { - "limit": 38271 + "limit": 38269 }, "reportUnknownParameterType": { "limit": 19584 diff --git a/litellm/proxy/auth/auth_checks.py b/litellm/proxy/auth/auth_checks.py index dc693317de0..a71a1993064 100644 --- a/litellm/proxy/auth/auth_checks.py +++ b/litellm/proxy/auth/auth_checks.py @@ -475,6 +475,7 @@ def _is_model_cost_zero(model: str | list[str] | None, llm_router: Router | None _NO_MODEL_INFO: Final[Mapping[str, object]] = MappingProxyType({}) +_TEAM_GRANT_RELATIONS: Final[Mapping[str, object]] = MappingProxyType({"litellm_model_table": True}) def _has_ptu_flat_cost(model: str, llm_router: "Router") -> bool: @@ -2858,7 +2859,9 @@ class TeamNotFoundError(HTTPException): async def _get_team_db_check( team_id: str, prisma_client: PrismaClient, team_id_upsert: bool | None = None ) -> "_PrismaTeamRow | None": - response = await _team_table(TeamRepository(prisma_client)).find_unique(where={"team_id": team_id}) + response = await _team_table(TeamRepository(prisma_client)).find_unique( + where={"team_id": team_id}, include=_TEAM_GRANT_RELATIONS + ) if response is None and team_id_upsert: from litellm.proxy.management_endpoints.team_endpoints import new_team @@ -3158,7 +3161,9 @@ async def get_team_object_by_alias( # Query database by team_alias try: - teams: Final = await _team_table(TeamRepository(prisma_client)).find_many(where={"team_alias": team_alias}) + teams: Final = await _team_table(TeamRepository(prisma_client)).find_many( + where={"team_alias": team_alias}, include=_TEAM_GRANT_RELATIONS + ) if not teams: raise HTTPException( diff --git a/litellm/proxy/auth/handle_jwt.py b/litellm/proxy/auth/handle_jwt.py index 0795cee7409..69091ee8344 100644 --- a/litellm/proxy/auth/handle_jwt.py +++ b/litellm/proxy/auth/handle_jwt.py @@ -53,6 +53,7 @@ from litellm.proxy._types import ( ) from litellm.proxy.auth.auth_checks import can_team_access_model from litellm.proxy.auth.route_checks import RouteChecks +from litellm.proxy.auth.team_grants import team_model_aliases from litellm.proxy.common_utils.user_api_key_cache import ( UserApiKeyCache, get_management_object_ttl, @@ -1595,7 +1596,7 @@ class JWTAuthManager: model=requested_model, team_object=team_object, llm_router=llm_router, - team_model_aliases=None, + team_model_aliases=team_model_aliases(team_object), ) ): is_allowed = allowed_routes_check( @@ -2132,7 +2133,7 @@ class JWTAuthManager: model=requested_model, team_object=team_object, llm_router=llm_router, - team_model_aliases=None, + team_model_aliases=team_model_aliases(team_object), ) except ProxyException: continue diff --git a/litellm/proxy/auth/team_grants.py b/litellm/proxy/auth/team_grants.py new file mode 100644 index 00000000000..1196011dcdd --- /dev/null +++ b/litellm/proxy/auth/team_grants.py @@ -0,0 +1,122 @@ +"""Project a team row (plus the caller's membership in it) onto the ``team_*`` fields of ``UserAPIKeyAuth``. + +The virtual-key path gets these fields for free from the combined-view SQL join. Every other auth path +starts from a ``LiteLLM_TeamTable`` object instead and has to copy them over by hand, which is how JWT +callers kept losing grants (aliases, permissions, limits) one field at a time. Build the badge through +``team_grants`` and the two paths cannot drift. +""" + +from collections.abc import Mapping, Sequence +from types import MappingProxyType +from typing import Annotated, Final + +from pydantic import BaseModel, BeforeValidator, ConfigDict, TypeAdapter, ValidationError +from pydantic.main import IncEx +from typing_extensions import ReadOnly, TypedDict + +from litellm.proxy._types import ( + LiteLLM_ObjectPermissionTable, + LiteLLM_TeamMembership, + LiteLLM_TeamTable, + Member, +) + +_MODEL_ALIASES_ADAPTER: Final = TypeAdapter(dict[str, str]) +_JSON_COLUMNS: Final[Mapping[str, IncEx | bool]] = MappingProxyType( + {"metadata": True, "litellm_model_table": MappingProxyType({"model_aliases": True})} +) + + +def _decode_model_aliases(value: object) -> object: + """``LiteLLM_ModelTable.model_aliases`` is typed ``str | dict``; writers hand Prisma ``json.dumps(...)``, so take both.""" + if not isinstance(value, str): + return value + try: + return _MODEL_ALIASES_ADAPTER.validate_json(value) + except ValidationError: + return None + + +class TeamModelAliasTable(BaseModel): + model_config = ConfigDict(protected_namespaces=()) + + model_aliases: Annotated[Mapping[str, str] | None, BeforeValidator(_decode_model_aliases)] = None + + +class _TeamJsonColumns(BaseModel): + """The two loosely typed columns on ``LiteLLM_TeamTable``, re-read with the shape the badge needs.""" + + metadata: Mapping[str, object] | None = None + litellm_model_table: TeamModelAliasTable | None = None + + +class TeamGrants(TypedDict, total=False): + """Keyword arguments for ``UserAPIKeyAuth``. Empty when the caller has no team, so the model's own defaults apply.""" + + team_alias: ReadOnly[str | None] + team_tpm_limit: ReadOnly[int | None] + team_rpm_limit: ReadOnly[int | None] + team_max_budget: ReadOnly[float | None] + team_soft_budget: ReadOnly[float | None] + team_spend: ReadOnly[float | None] + team_models: ReadOnly[Sequence[str]] + team_blocked: ReadOnly[bool] + team_metadata: ReadOnly[Mapping[str, object] | None] + team_model_aliases: ReadOnly[Mapping[str, str] | None] + team_object_permission_id: ReadOnly[str | None] + team_object_permission: ReadOnly[LiteLLM_ObjectPermissionTable | None] + team_member: ReadOnly[Member | None] + team_member_spend: ReadOnly[float | None] + team_member_tpm_limit: ReadOnly[int | None] + team_member_rpm_limit: ReadOnly[int | None] + + +def _json_columns(team_object: LiteLLM_TeamTable) -> _TeamJsonColumns: + try: + return _TeamJsonColumns.model_validate(team_object.model_dump(include=_JSON_COLUMNS)) + except ValidationError: + return _TeamJsonColumns() + + +def team_model_aliases(team_object: LiteLLM_TeamTable | None) -> Mapping[str, str] | None: + if team_object is None: + return None + alias_table: Final = _json_columns(team_object).litellm_model_table + return alias_table.model_aliases if alias_table is not None else None + + +def team_grants( + team_object: LiteLLM_TeamTable | None, + team_membership: LiteLLM_TeamMembership | None, + user_id: str | None, +) -> TeamGrants: + if team_object is None: + return TeamGrants() + json_columns: Final = _json_columns(team_object) + return TeamGrants( + team_alias=team_object.team_alias, + team_tpm_limit=team_object.tpm_limit, + team_rpm_limit=team_object.rpm_limit, + team_max_budget=team_object.max_budget, + team_soft_budget=team_object.soft_budget, + team_spend=team_object.spend, + team_models=tuple(team_object.models), + team_blocked=team_object.blocked, + team_metadata=json_columns.metadata, + team_model_aliases=( + json_columns.litellm_model_table.model_aliases if json_columns.litellm_model_table is not None else None + ), + team_object_permission_id=team_object.object_permission_id, + team_object_permission=team_object.object_permission, + team_member=next( + (m for m in team_object.members_with_roles if user_id is not None and m.user_id == user_id), + None, + ), + team_member_spend=team_membership.spend if team_membership is not None else None, + team_member_tpm_limit=( + team_membership.safe_get_team_member_tpm_limit() if team_membership is not None else None + ), + team_member_rpm_limit=( + team_membership.safe_get_team_member_rpm_limit() if team_membership is not None else None + ), + ) diff --git a/litellm/proxy/auth/user_api_key_auth.py b/litellm/proxy/auth/user_api_key_auth.py index 93293db24c6..08dbe2508ec 100644 --- a/litellm/proxy/auth/user_api_key_auth.py +++ b/litellm/proxy/auth/user_api_key_auth.py @@ -82,6 +82,7 @@ from litellm.proxy.auth.oauth2_proxy_hook import handle_oauth2_proxy_request from litellm.proxy.auth.resolvers import CredentialRef, Principal from litellm.proxy.auth.resolvers.store import IdentityStore from litellm.proxy.auth.route_checks import RouteChecks +from litellm.proxy.auth.team_grants import team_grants from litellm.proxy.auth.trusted_proxy_utils import get_trusted_proxy_cidrs from litellm.proxy.common_utils.cache_coordinator import EventDrivenCacheCoordinator from litellm.proxy.common_utils.http_parsing_utils import ( @@ -1476,24 +1477,16 @@ async def _user_api_key_auth_builder( user_id=user_id, user_email=user_email, team_id=team_id, - team_alias=(team_object.team_alias if team_object is not None else None), - team_tpm_limit=(team_object.tpm_limit if team_object is not None else None), - team_rpm_limit=(team_object.rpm_limit if team_object is not None else None), - team_models=(team_object.models if team_object is not None else []), - team_metadata=(team_object.metadata if team_object is not None else None), org_id=org_id, end_user_id=end_user_id, parent_otel_span=parent_otel_span, jwt_claims=jwt_claims, + **team_grants(team_object=team_object, team_membership=team_membership, user_id=user_id), ) valid_token = UserAPIKeyAuth( api_key=None, team_id=team_id, - team_alias=(team_object.team_alias if team_object is not None else None), - team_tpm_limit=(team_object.tpm_limit if team_object is not None else None), - team_rpm_limit=(team_object.rpm_limit if team_object is not None else None), - team_models=(team_object.models if team_object is not None else []), user_role=( LitellmUserRoles(user_object.user_role) if user_object is not None and user_object.user_role is not None @@ -1507,17 +1500,8 @@ async def _user_api_key_auth_builder( user_tpm_limit=(user_object.tpm_limit if user_object is not None else None), user_rpm_limit=(user_object.rpm_limit if user_object is not None else None), user_model_max_budget=(user_object.model_max_budget if user_object is not None else None), - team_member_rpm_limit=( - team_membership.safe_get_team_member_rpm_limit() if team_membership is not None else None - ), - team_member_tpm_limit=( - team_membership.safe_get_team_member_tpm_limit() if team_membership is not None else None - ), - team_metadata=(team_object.metadata if team_object is not None else None), jwt_claims=jwt_claims, - ) - valid_token.team_object_permission = ( - team_object.object_permission if team_object is not None else None + **team_grants(team_object=team_object, team_membership=team_membership, user_id=user_id), ) # AUTO_REGISTER deferred from _resolve_jwt_to_virtual_key. diff --git a/litellm/proxy/management_endpoints/team_endpoints.py b/litellm/proxy/management_endpoints/team_endpoints.py index 2ec68a10f65..c050368b3fe 100644 --- a/litellm/proxy/management_endpoints/team_endpoints.py +++ b/litellm/proxy/management_endpoints/team_endpoints.py @@ -5601,7 +5601,7 @@ async def team_model_add( updated_team: Final = await _team_db(prisma_client).update( where={"team_id": data.team_id}, data={"updated_at": datetime.now(timezone.utc)}, - include={"object_permission": True}, + include={"litellm_model_table": True, "object_permission": True}, ) if updated_team is None: raise HTTPException( @@ -5688,7 +5688,7 @@ async def team_model_delete( updated_team: Final = await _team_db(prisma_client).update( where={"team_id": data.team_id}, data={"models": updated_models}, - include={"object_permission": True}, + include={"litellm_model_table": True, "object_permission": True}, ) if updated_team is None: raise HTTPException( diff --git a/litellm/proxy/management_endpoints/ui_sso.py b/litellm/proxy/management_endpoints/ui_sso.py index 3e6434a5afd..c60888e298f 100644 --- a/litellm/proxy/management_endpoints/ui_sso.py +++ b/litellm/proxy/management_endpoints/ui_sso.py @@ -22,7 +22,6 @@ from html import escape from types import MappingProxyType from typing import ( TYPE_CHECKING, - Annotated, Any, Final, Literal, @@ -42,7 +41,7 @@ if TYPE_CHECKING: import jwt from fastapi import APIRouter, Depends, Header, HTTPException, Request, Response, status from fastapi.responses import RedirectResponse -from pydantic import BaseModel, BeforeValidator, ConfigDict, TypeAdapter, ValidationError +from pydantic import BaseModel, TypeAdapter, ValidationError import litellm from litellm._logging import verbose_proxy_logger @@ -95,6 +94,7 @@ from litellm.proxy.auth.auth_utils import ( ) from litellm.proxy.auth.handle_jwt import JWTHandler from litellm.proxy.auth.ip_address_utils import IPAddressUtils +from litellm.proxy.auth.team_grants import TeamModelAliasTable from litellm.proxy.auth.user_api_key_auth import user_api_key_auth from litellm.proxy.common_utils.admin_ui_utils import ( admin_ui_disabled, @@ -209,31 +209,14 @@ def _team_detail_db(repo: TeamRepository) -> "TableActions[_TeamDetailRow]": return repo.table -_MODEL_ALIASES_ADAPTER: Final = TypeAdapter(dict[str, str]) _SSO_TOKEN_CLAIMS_ADAPTER: Final = TypeAdapter(Mapping[str, object]) -def _decode_model_aliases(value: object) -> object: - """``/team/new`` stores team model aliases as a JSON-encoded string in the Json column.""" - if not isinstance(value, str): - return value - try: - return _MODEL_ALIASES_ADAPTER.validate_json(value) - except ValidationError: - return None - - -class _TeamModelAliasTable(BaseModel): - model_config = ConfigDict(protected_namespaces=()) - - model_aliases: Annotated[Mapping[str, str] | None, BeforeValidator(_decode_model_aliases)] = None - - class _TeamRowGrants(BaseModel): team_id: str team_alias: str | None = None models: tuple[str, ...] = () - litellm_model_table: _TeamModelAliasTable | None = None + litellm_model_table: TeamModelAliasTable | None = None class CliSsoTeamDetail(BaseModel): diff --git a/tests/test_litellm/proxy/auth/test_auth_checks.py b/tests/test_litellm/proxy/auth/test_auth_checks.py index 2284a05b2e9..5bfef2b6445 100644 --- a/tests/test_litellm/proxy/auth/test_auth_checks.py +++ b/tests/test_litellm/proxy/auth/test_auth_checks.py @@ -2374,6 +2374,44 @@ def _mock_prisma_for_team_lookup(find_unique): return mock_prisma_client +_TEAM_ALIAS_TABLE_ROW = {"id": 1, "model_aliases": '{"fast": "gpt-4o"}', "created_by": "admin", "updated_by": "admin"} + + +def _prisma_team_row(include): + """Mimics Prisma: the `litellm_model_table` relation rides on the row only when the query `include`s it.""" + columns = {"team_id": "team-aliases", "team_alias": "aliases", "models": ["gpt-4o"]} + row = ( + {**columns, "litellm_model_table": _TEAM_ALIAS_TABLE_ROW} + if (include or {}).get("litellm_model_table") + else columns + ) + return SimpleNamespace(dict=lambda: row, model_dump=lambda: row) + + +@pytest.mark.asyncio +async def test_get_team_object_loads_model_aliases_relation(): + """LIT-5858: the auth path read teams without `include`ing `litellm_model_table`, so every JWT + team came back with `model_aliases=None` and alias requests 403'd.""" + from litellm.proxy.auth.auth_checks import get_team_object + from litellm.proxy.auth.team_grants import team_model_aliases + + async def find_unique(where, include=None): + return _prisma_team_row(include) + + mock_cache = MagicMock() + mock_cache.async_get_cache = AsyncMock(return_value=None) + mock_cache.async_set_cache = AsyncMock() + + team = await get_team_object( + team_id="team-aliases", + prisma_client=_mock_prisma_for_team_lookup(AsyncMock(side_effect=find_unique)), + user_api_key_cache=mock_cache, + check_db_only=True, + ) + + assert team_model_aliases(team) == {"fast": "gpt-4o"} + + @pytest.mark.asyncio async def test_get_team_object_distinguishes_absent_team_from_unreadable_row(): """A deleted team and a database that would not answer both surface as a 404, @@ -6195,6 +6233,32 @@ async def test_get_team_object_by_alias_db_fetch_returns_cached_obj(): assert result.models == ["gpt-4"] +@pytest.mark.asyncio +async def test_get_team_object_by_alias_loads_model_aliases_relation(): + """LIT-5858: same regression as `test_get_team_object_loads_model_aliases_relation`, for the + `team_alias_jwt_field` lookup.""" + from litellm.proxy.auth.auth_checks import get_team_object_by_alias + from litellm.proxy.auth.team_grants import team_model_aliases + + async def find_many(where, include=None): + return [_prisma_team_row(include)] + + mock_prisma_client = MagicMock() + mock_prisma_client.db.litellm_teamtable.find_many = AsyncMock(side_effect=find_many) + + mock_cache = MagicMock() + mock_cache.async_get_cache = AsyncMock(return_value=None) + mock_cache.async_set_cache = AsyncMock() + + team = await get_team_object_by_alias( + team_alias="aliases", + prisma_client=mock_prisma_client, + user_api_key_cache=mock_cache, + ) + + assert team_model_aliases(team) == {"fast": "gpt-4o"} + + @pytest.mark.asyncio async def test_get_org_object_by_alias_db_fetch_returns_validated_org(): from litellm.proxy._types import LiteLLM_OrganizationTable diff --git a/tests/test_litellm/proxy/auth/test_handle_jwt.py b/tests/test_litellm/proxy/auth/test_handle_jwt.py index 99a0a4c0a8b..94226b5404d 100644 --- a/tests/test_litellm/proxy/auth/test_handle_jwt.py +++ b/tests/test_litellm/proxy/auth/test_handle_jwt.py @@ -13,6 +13,7 @@ from litellm.proxy._types import ( DEFAULT_JWKS_STALE_TTL, JWTLiteLLMRoleMap, LiteLLM_JWTAuth, + LiteLLM_ModelTable, LiteLLM_TeamMembership, LiteLLM_TeamTable, LiteLLM_UserTable, @@ -1255,6 +1256,57 @@ async def test_find_team_with_model_access_model_group(monkeypatch): assert team_obj.team_id == "team-1" +@pytest.mark.asyncio +@pytest.mark.parametrize( + "model_aliases", + ['{"fast": "gpt-4o"}', {"fast": "gpt-4o"}], + ids=["json-string", "dict"], +) +async def test_find_team_with_model_access_resolves_team_model_alias(monkeypatch, model_aliases): + """LIT-5858: a JWT team that grants `gpt-4o` under the alias `fast` must resolve a request + for `fast`. The JWT path used to pass `team_model_aliases=None`, so every alias request 403'd.""" + import sys + import types + + from litellm.caching import DualCache + from litellm.proxy.utils import ProxyLogging + from litellm.router import Router + + router = Router(model_list=[{"model_name": "gpt-4o", "litellm_params": {"model": "gpt-4o"}}]) + proxy_server_module = types.ModuleType("proxy_server") + proxy_server_module.llm_router = router + monkeypatch.setitem(sys.modules, "litellm.proxy.proxy_server", proxy_server_module) + + team = LiteLLM_TeamTable( + team_id="team-aliases", + models=["gpt-4o"], + litellm_model_table=LiteLLM_ModelTable(model_aliases=model_aliases, created_by="admin", updated_by="admin"), + ) + + async def mock_get_team_object(*args, **kwargs): + return team + + monkeypatch.setattr("litellm.proxy.auth.handle_jwt.get_team_object", mock_get_team_object) + + jwt_handler = JWTHandler() + jwt_handler.litellm_jwtauth = LiteLLM_JWTAuth() + user_api_key_cache = DualCache() + + team_id, team_obj = await JWTAuthManager.find_team_with_model_access( + team_ids={"team-aliases"}, + requested_model="fast", + route="/chat/completions", + jwt_handler=jwt_handler, + prisma_client=None, + user_api_key_cache=user_api_key_cache, + parent_otel_span=None, + proxy_logging_obj=ProxyLogging(user_api_key_cache=user_api_key_cache), + ) + + assert team_id == "team-aliases" + assert team_obj is team + + @pytest.mark.asyncio async def test_find_team_with_model_access_v1_messages_default_routes(monkeypatch): """Regression for #31189: a single-team JWT that grants the requested model diff --git a/tests/test_litellm/proxy/auth/test_team_grants.py b/tests/test_litellm/proxy/auth/test_team_grants.py new file mode 100644 index 00000000000..447fc1c93a1 --- /dev/null +++ b/tests/test_litellm/proxy/auth/test_team_grants.py @@ -0,0 +1,129 @@ +import pytest + +from litellm.proxy._types import ( + LiteLLM_BudgetTable, + LiteLLM_ObjectPermissionTable, + LiteLLM_TeamMembership, + LiteLLM_TeamTable, + LiteLLM_VerificationTokenView, + Member, + UserAPIKeyAuth, +) +from litellm.models.team import LiteLLM_ModelTable +from litellm.proxy.auth.team_grants import team_grants, team_model_aliases + +TEAM_ID = "team-grants" +USER_ID = "user-in-team" +ALIASES = {"fast": "gpt-4o-mini", "smart": "gpt-4o"} + + +def _alias_table(model_aliases) -> LiteLLM_ModelTable: + return LiteLLM_ModelTable(model_aliases=model_aliases, created_by="admin", updated_by="admin") + + +def _full_team(model_aliases=ALIASES) -> LiteLLM_TeamTable: + return LiteLLM_TeamTable( + team_id=TEAM_ID, + team_alias="grants-team", + tpm_limit=1000, + rpm_limit=10, + max_budget=50.0, + soft_budget=25.0, + spend=12.5, + models=["gpt-4o", "gpt-4o-mini"], + blocked=True, + metadata={"tier": "gold"}, + litellm_model_table=_alias_table(model_aliases), + object_permission_id="op-1", + object_permission=LiteLLM_ObjectPermissionTable(object_permission_id="op-1", mcp_servers=["mcp-a"]), + members_with_roles=[ + Member(user_id="someone-else", role="user"), + Member(user_id=USER_ID, role="admin"), + ], + ) + + +def _membership() -> LiteLLM_TeamMembership: + return LiteLLM_TeamMembership( + user_id=USER_ID, + team_id=TEAM_ID, + spend=3.25, + litellm_budget_table=LiteLLM_BudgetTable(tpm_limit=500, rpm_limit=5), + ) + + +def test_team_grants_cover_every_team_field_the_key_path_gets(): + """Class guard for LIT-5858 and its siblings: every ``team_*`` column the combined-view SQL hands the + virtual-key path must come out of the projection too, with the team's actual value, so adding a column + to ``LiteLLM_VerificationTokenView`` without teaching ``team_grants`` fails here instead of in prod.""" + team = _full_team() + grants = team_grants(team_object=team, team_membership=_membership(), user_id=USER_ID) + token = UserAPIKeyAuth(team_id=TEAM_ID, **grants) + + view_team_fields = {name for name in LiteLLM_VerificationTokenView.model_fields if name.startswith("team_")} + assert view_team_fields - {"team_id"} <= set(grants) + assert all(grants[name] is not None for name in view_team_fields - {"team_id"}) + + assert token.team_alias == "grants-team" + assert token.team_tpm_limit == 1000 + assert token.team_rpm_limit == 10 + assert token.team_max_budget == 50.0 + assert token.team_soft_budget == 25.0 + assert token.team_spend == 12.5 + assert token.team_models == ["gpt-4o", "gpt-4o-mini"] + assert token.team_blocked is True + assert token.team_metadata == {"tier": "gold"} + assert token.team_model_aliases == ALIASES + assert token.team_object_permission_id == "op-1" + assert token.team_object_permission is not None + assert token.team_object_permission.mcp_servers == ["mcp-a"] + assert token.team_member == Member(user_id=USER_ID, role="admin") + assert token.team_member_spend == 3.25 + assert token.team_member_tpm_limit == 500 + assert token.team_member_rpm_limit == 5 + + +def test_team_grants_without_team_leave_token_defaults(): + token = UserAPIKeyAuth(**team_grants(team_object=None, team_membership=None, user_id=USER_ID)) + assert token == UserAPIKeyAuth() + + +@pytest.mark.parametrize( + "stored_aliases", + [ALIASES, '{"fast": "gpt-4o-mini", "smart": "gpt-4o"}'], + ids=["json-object", "json-string-as-written-by-team-new"], +) +def test_team_model_aliases_decode_both_storage_shapes(stored_aliases): + team = _full_team(model_aliases=stored_aliases) + assert team_model_aliases(team) == ALIASES + assert team_grants(team_object=team, team_membership=None, user_id=None)["team_model_aliases"] == ALIASES + + +@pytest.mark.parametrize("stored_aliases", [None, "not json", '["a", "b"]', {"fast": 3}], ids=str) +def test_team_model_aliases_treat_unusable_column_as_no_aliases(stored_aliases): + team = _full_team(model_aliases=stored_aliases) + assert team_model_aliases(team) is None + assert team_grants(team_object=team, team_membership=None, user_id=None)["team_model_aliases"] is None + + +def test_team_model_aliases_none_without_relation_loaded(): + team = _full_team() + team.litellm_model_table = None + assert team_model_aliases(team) is None + assert team_model_aliases(None) is None + + +def test_team_member_is_the_callers_row_only(): + team = _full_team() + assert team_grants(team_object=team, team_membership=None, user_id="someone-else")["team_member"] == Member( + user_id="someone-else", role="user" + ) + assert team_grants(team_object=team, team_membership=None, user_id="stranger")["team_member"] is None + assert team_grants(team_object=team, team_membership=None, user_id=None)["team_member"] is None + + +def test_membership_limits_absent_without_membership_row(): + grants = team_grants(team_object=_full_team(), team_membership=None, user_id=USER_ID) + assert grants["team_member_spend"] is None + assert grants["team_member_tpm_limit"] is None + assert grants["team_member_rpm_limit"] is None diff --git a/tests/test_litellm/proxy/auth/test_user_api_key_auth.py b/tests/test_litellm/proxy/auth/test_user_api_key_auth.py index d44f96d95bf..78ffbb0db23 100644 --- a/tests/test_litellm/proxy/auth/test_user_api_key_auth.py +++ b/tests/test_litellm/proxy/auth/test_user_api_key_auth.py @@ -6872,3 +6872,119 @@ class TestLitellmReceivedAtStamping: assert result == earlier assert request.state.litellm_received_at == earlier + + +@pytest.mark.asyncio +@pytest.mark.parametrize("is_proxy_admin", [False, True], ids=["standard-return", "proxy-admin-return"]) +async def test_jwt_builder_returns_every_team_grant_the_key_path_gets(is_proxy_admin): + """LIT-5858: the team-based JWT path hand-built ``UserAPIKeyAuth`` from a short list of team fields, so the + team's model aliases (and on the admin return, its object permission) never reached the token and alias + requests 403'd. Both returns now go through ``team_grants``; pin the fields that used to be dropped.""" + import litellm.proxy.proxy_server as _proxy_server_mod + from fastapi import Request + from starlette.datastructures import URL + + from litellm.models.team import LiteLLM_ModelTable + from litellm.proxy._types import ( + LiteLLM_ObjectPermissionTable, + LiteLLM_TeamMembership, + LiteLLM_TeamTable, + Member, + ) + + class _AcceptEveryJwt(JWTHandler): + def is_jwt(self, token: str) -> bool: + return True + + jwt_handler = _AcceptEveryJwt() + jwt_handler.litellm_jwtauth = LiteLLM_JWTAuth() + + team = LiteLLM_TeamTable( + team_id="team-jwt-aliases", + team_alias="jwt-aliases", + models=["gpt-4o"], + max_budget=40.0, + spend=4.0, + blocked=False, + metadata={"tier": "gold"}, + litellm_model_table=LiteLLM_ModelTable( + model_aliases='{"fast": "gpt-4o"}', created_by="admin", updated_by="admin" + ), + object_permission_id="op-jwt", + object_permission=LiteLLM_ObjectPermissionTable(object_permission_id="op-jwt", mcp_servers=["mcp-a"]), + members_with_roles=[Member(user_id="jwt-user", role="admin")], + ) + membership = LiteLLM_TeamMembership(user_id="jwt-user", team_id="team-jwt-aliases", spend=1.5) + builder_result = { + "is_proxy_admin": is_proxy_admin, + "team_object": team, + "user_object": None, + "end_user_object": None, + "org_object": None, + "token": "jwt", + "team_id": "team-jwt-aliases", + "user_id": "jwt-user", + "user_email": "jwt-user@example.com", + "end_user_id": None, + "org_id": None, + "team_membership": membership, + "jwt_claims": {"sub": "jwt-user"}, + } + + mock_proxy_logging_obj = MagicMock() + mock_proxy_logging_obj.internal_usage_cache = MagicMock() + mock_proxy_logging_obj.internal_usage_cache.dual_cache = AsyncMock() + mock_proxy_logging_obj.post_call_failure_hook = AsyncMock(return_value=None) + attrs = { + "prisma_client": MagicMock(), + "user_api_key_cache": DualCache(), + "proxy_logging_obj": mock_proxy_logging_obj, + "master_key": "sk-master-key", + "general_settings": {"enable_jwt_auth": True}, + "llm_model_list": [], + "llm_router": None, + "open_telemetry_logger": None, + "model_max_budget_limiter": MagicMock(), + "user_custom_auth": None, + "jwt_handler": jwt_handler, + "premium_user": True, + "litellm_proxy_admin_name": "admin", + } + originals = {a: getattr(_proxy_server_mod, a, None) for a in attrs} + try: + for k, v in attrs.items(): + setattr(_proxy_server_mod, k, v) + request = Request(scope={"type": "http", "headers": [], "method": "POST"}) + request._url = URL(url="/chat/completions") + with patch( # test-quality-ok: auth_builder is the claim-resolution seam; the regression is how its result is projected onto the token + "litellm.proxy.auth.user_api_key_auth.JWTAuthManager.auth_builder", + new_callable=AsyncMock, + return_value=builder_result, + ): + token = await _user_api_key_auth_builder( + request=request, + api_key="Bearer header.payload.signature", + azure_api_key_header="", + anthropic_api_key_header=None, + google_ai_studio_api_key_header=None, + azure_apim_header=None, + request_data={}, + ) + finally: + for k, v in originals.items(): + setattr(_proxy_server_mod, k, v) + + assert token.team_id == "team-jwt-aliases" + assert token.user_role == (LitellmUserRoles.PROXY_ADMIN if is_proxy_admin else LitellmUserRoles.INTERNAL_USER) + assert token.team_model_aliases == {"fast": "gpt-4o"} + assert token.team_object_permission is not None + assert token.team_object_permission.mcp_servers == ["mcp-a"] + assert token.team_object_permission_id == "op-jwt" + assert token.team_alias == "jwt-aliases" + assert token.team_models == ["gpt-4o"] + assert token.team_max_budget == 40.0 + assert token.team_spend == 4.0 + assert token.team_metadata == {"tier": "gold"} + assert token.team_member == Member(user_id="jwt-user", role="admin") + assert token.team_member_spend == 1.5 + assert token.jwt_claims == {"sub": "jwt-user"} diff --git a/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py index 051e6bed4fd..2f6561046b1 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_team_endpoints.py @@ -2205,6 +2205,50 @@ async def test_team_model_add_delete_refresh_team_cache(endpoint_name): assert update_call_kwargs.get("include", {}).get("object_permission") is True +@pytest.mark.asyncio +@pytest.mark.parametrize("endpoint_name", ["team_model_add", "team_model_delete"]) +async def test_team_model_add_delete_keep_model_aliases_in_team_cache(endpoint_name, monkeypatch): + """LIT-5858: Prisma only returns `litellm_model_table` when the `update` asks for it, so the refreshed + cache entry lost the team's model aliases and JWT alias requests 403'd until the next DB read.""" + from litellm.proxy._types import TeamModelAddRequest, TeamModelDeleteRequest + from litellm.proxy.auth.team_grants import team_model_aliases + from litellm.proxy.common_utils.user_api_key_cache import UserApiKeyCache + from litellm.proxy.management_endpoints.team_endpoints import team_model_add, team_model_delete + + columns = {"team_id": "team-1234", "models": ["gpt-4o", "openai/*"]} + alias_table = {"id": 1, "model_aliases": '{"fast": "gpt-4o"}', "created_by": "admin", "updated_by": "admin"} + + async def update(where, data, include=None): + row = {**columns, "litellm_model_table": alias_table} if (include or {}).get("litellm_model_table") else columns + return SimpleNamespace(team_id="team-1234", model_dump=lambda: row) + + prisma_client = MagicMock() + prisma_client.db.litellm_teamtable.find_unique = AsyncMock(return_value=SimpleNamespace(model_dump=lambda: columns)) + prisma_client.db.litellm_teamtable.update = AsyncMock(side_effect=update) + prisma_client.db.execute_raw = AsyncMock(return_value=None) + cache = UserApiKeyCache() + monkeypatch.setattr("litellm.proxy.proxy_server.prisma_client", prisma_client) + monkeypatch.setattr("litellm.proxy.proxy_server.user_api_key_cache", cache) + monkeypatch.setattr("litellm.proxy.proxy_server.proxy_logging_obj", None) + + admin = UserAPIKeyAuth(user_role=LitellmUserRoles.PROXY_ADMIN, user_id="admin") + if endpoint_name == "team_model_add": + await team_model_add( + data=TeamModelAddRequest(team_id="team-1234", models=["team-byok-1"]), + http_request=MagicMock(), + user_api_key_dict=admin, + ) + else: + await team_model_delete( + data=TeamModelDeleteRequest(team_id="team-1234", models=["openai/*"]), + http_request=MagicMock(), + user_api_key_dict=admin, + ) + + cached_team = await cache.async_get_cache(key="team_id:team-1234", model_type=LiteLLM_TeamTableCachedObj) + assert team_model_aliases(cached_team) == {"fast": "gpt-4o"} + + @pytest.mark.asyncio @pytest.mark.parametrize( "endpoint_name", diff --git a/type-discipline-budget.json b/type-discipline-budget.json index e7186dfe186..7db8ed501f8 100644 --- a/type-discipline-budget.json +++ b/type-discipline-budget.json @@ -3,7 +3,7 @@ "limit": 22180 }, "LIT002": { - "limit": 26729 + "limit": 26727 }, "LIT003": { "limit": 261 From c84131b81ab6feb552e23c5996559159e8810066 Mon Sep 17 00:00:00 2001 From: ryan-crabbe-berri Date: Mon, 7 Sep 2026 16:00:20 -0700 Subject: [PATCH 02/26] feat(proxy): price cache and reasoning tokens in /cost/estimate POST /cost/estimate now accepts cache_read_input_tokens, cache_creation_input_tokens and reasoning_tokens, bills them at the model's cache and reasoning rates, and reports each share per request, per day and per month next to the rates it used. Custom-priced deployments also get cache and reasoning lines in the cost breakdown now, so the estimate and the spend logs reconcile with their totals instead of showing zero for those tokens. Requested by a customer (Pylon #7365). Claude-Session: https://claude.ai/code/session_011Tn3657NkV6ojLqewL64Kb --- litellm/cost_calculator.py | 1 + .../litellm_core_utils/llm_cost_calc/utils.py | 68 +++-- litellm/proxy/_types.py | 36 +++ .../cost_tracking_settings.py | 253 +++++++++++------- .../llm_cost_calc/test_llm_cost_calc_utils.py | 64 +++++ .../test_cost_tracking_settings.py | 187 ++++++++++++- tests/test_litellm/test_cost_calculator.py | 50 ++++ ui/litellm-dashboard/src/lib/http/schema.d.ts | 103 ++++++- 8 files changed, 641 insertions(+), 121 deletions(-) diff --git a/litellm/cost_calculator.py b/litellm/cost_calculator.py index 9a9d2ceda03..a1181ee4124 100644 --- a/litellm/cost_calculator.py +++ b/litellm/cost_calculator.py @@ -1746,6 +1746,7 @@ def completion_cost( service_tier=service_tier, data_residency=data_residency, vertex_location=vertex_location, + custom_cost_per_token=custom_cost_per_token, ) _reasoning_cost = _token_type_breakdown.reasoning_cost _cache_read_cost = _token_type_breakdown.cache_read_cost diff --git a/litellm/litellm_core_utils/llm_cost_calc/utils.py b/litellm/litellm_core_utils/llm_cost_calc/utils.py index 68dc27ec25e..8d53c8da3e6 100644 --- a/litellm/litellm_core_utils/llm_cost_calc/utils.py +++ b/litellm/litellm_core_utils/llm_cost_calc/utils.py @@ -19,6 +19,7 @@ from litellm.types.utils import ( CacheCreationTokenDetails, CallTypes, CompletionTokensDetailsWrapper, + CostPerToken, DataResidency, ImageResponse, ModelInfo, @@ -1307,6 +1308,40 @@ class TokenTypeCostBreakdown: cache_creation_cost: float +def _reasoning_token_count(usage: Usage) -> int: + parsed: Final = ( + parse_completion_tokens_details(usage)["reasoning_tokens"] if usage.completion_tokens_details is not None else 0 + ) + return parsed or _coerce_token_count(getattr(usage, "reasoning_tokens", 0)) + + +def _cache_token_counts(usage: Usage) -> tuple[int, int, CacheCreationTokenDetails | None]: + """(cache read tokens, cache creation tokens, cache creation details): read from prompt_tokens_details + first, then the private top-level counters the Usage constructor mirrors cache tokens onto for + providers/callers that bypass the details.""" + parsed: Final = parse_prompt_tokens_details(usage) if usage.prompt_tokens_details is not None else None + parsed_read: Final = parsed["cache_hit_tokens"] if parsed is not None else 0 + parsed_creation: Final = parsed["cache_creation_tokens"] if parsed is not None else 0 + return ( + parsed_read or _coerce_token_count(getattr(usage, "_cache_read_input_tokens", 0)), + parsed_creation or _coerce_token_count(getattr(usage, "_cache_creation_input_tokens", 0)), + parsed["cache_creation_token_details"] if parsed is not None else None, + ) + + +def _custom_pricing_token_type_breakdown(usage: Usage, custom_cost_per_token: CostPerToken) -> TokenTypeCostBreakdown: + """Flat custom pricing has no tiers, uplifts or reasoning rate: cache tokens bill at the configured + cache rates (else the input rate) and reasoning at the output rate, as _cost_per_token_custom_pricing_helper does.""" + input_rate: Final = custom_cost_per_token["input_cost_per_token"] + cache_read_tokens, cache_creation_tokens, _ = _cache_token_counts(usage) + return TokenTypeCostBreakdown( + reasoning_cost=float(_reasoning_token_count(usage)) * custom_cost_per_token["output_cost_per_token"], + cache_read_cost=float(cache_read_tokens) * custom_cost_per_token.get("cache_read_input_token_cost", input_rate), + cache_creation_cost=float(cache_creation_tokens) + * custom_cost_per_token.get("cache_creation_input_token_cost", input_rate), + ) + + def get_token_type_cost_breakdown( model: str, custom_llm_provider: str | None, @@ -1315,6 +1350,7 @@ def get_token_type_cost_breakdown( data_residency: str | None = None, vertex_location: str | None = None, current_time: datetime | None = None, + custom_cost_per_token: CostPerToken | None = None, ) -> TokenTypeCostBreakdown: """ Provider-agnostic cost of reasoning and cache tokens, derived from the usage @@ -1325,9 +1361,13 @@ def get_token_type_cost_breakdown( land on ``prompt_tokens_details`` (via the Usage constructor and provider transformations) and reasoning tokens on ``completion_tokens_details``. It reuses the same rate-resolution primitives as the total-cost path so the breakdown can - never drift from the totals. Returns zeros (never raises) when the model or its - pricing cannot be resolved. + never drift from the totals. A deployment billed by ``custom_cost_per_token`` is + priced from those flat rates instead of the cost map, for the same reason. + Returns zeros (never raises) when the model or its pricing cannot be resolved. """ + if custom_cost_per_token is not None: + return _custom_pricing_token_type_breakdown(usage=usage, custom_cost_per_token=custom_cost_per_token) + try: model_info: Final = get_model_info(model=model, custom_llm_provider=custom_llm_provider) except Exception: @@ -1348,12 +1388,6 @@ def get_token_type_cost_breakdown( threshold_is_inclusive=_uses_inclusive_token_thresholds(custom_llm_provider), ) - reasoning_tokens = ( - parse_completion_tokens_details(usage)["reasoning_tokens"] if usage.completion_tokens_details is not None else 0 - ) - if not reasoning_tokens: - reasoning_tokens = _coerce_token_count(getattr(usage, "reasoning_tokens", 0)) - reasoning_rate: Final = _resolve_billed_reasoning_rate( model_info=model_info, usage=usage, @@ -1361,23 +1395,9 @@ def get_token_type_cost_breakdown( completion_base_cost=completion_base_cost, current_time=billing_time, ) - reasoning_cost = float(reasoning_tokens) * reasoning_rate - - cache_read_tokens = 0 - cache_creation_tokens = 0 - cache_creation_token_details: CacheCreationTokenDetails | None = None - if usage.prompt_tokens_details is not None: - prompt_tokens_details: Final = parse_prompt_tokens_details(usage) - cache_read_tokens = prompt_tokens_details["cache_hit_tokens"] - cache_creation_tokens = prompt_tokens_details["cache_creation_tokens"] - cache_creation_token_details = prompt_tokens_details["cache_creation_token_details"] - # Fall back to the private top-level counters the Usage constructor mirrors cache - # tokens onto, so providers/callers that bypass prompt_tokens_details are covered. - if not cache_read_tokens: - cache_read_tokens = _coerce_token_count(getattr(usage, "_cache_read_input_tokens", 0)) - if not cache_creation_tokens: - cache_creation_tokens = _coerce_token_count(getattr(usage, "_cache_creation_input_tokens", 0)) + reasoning_cost = float(_reasoning_token_count(usage)) * reasoning_rate + cache_read_tokens, cache_creation_tokens, cache_creation_token_details = _cache_token_counts(usage) cache_read_cost = float(cache_read_tokens) * cache_read_cost_rate cache_creation_cost = calculate_cache_writing_cost( cache_creation_tokens=cache_creation_tokens, diff --git a/litellm/proxy/_types.py b/litellm/proxy/_types.py index abce10690e5..65b7d409d7c 100644 --- a/litellm/proxy/_types.py +++ b/litellm/proxy/_types.py @@ -5137,9 +5137,26 @@ class CostEstimateRequest(LiteLLMPydanticObjectBase): model: str = Field(description="Model name (from /model_group/info)") input_tokens: int = Field(description="Expected input tokens per request", ge=0) output_tokens: int = Field(description="Expected output tokens per request", ge=0) + cache_read_input_tokens: int = Field( + default=0, description="Input tokens read from the prompt cache; counted within input_tokens", ge=0 + ) + cache_creation_input_tokens: int = Field( + default=0, description="Input tokens written to the prompt cache; counted within input_tokens", ge=0 + ) + reasoning_tokens: int = Field( + default=0, description="Reasoning tokens the model emits; counted within output_tokens", ge=0 + ) num_requests_per_day: int | None = Field(default=None, description="Number of requests per day", ge=0) num_requests_per_month: int | None = Field(default=None, description="Number of requests per month", ge=0) + @model_validator(mode="after") + def validate_token_subsets(self) -> "CostEstimateRequest": + if self.cache_read_input_tokens + self.cache_creation_input_tokens > self.input_tokens: + raise ValueError("cache_read_input_tokens plus cache_creation_input_tokens cannot exceed input_tokens") + if self.reasoning_tokens > self.output_tokens: + raise ValueError("reasoning_tokens cannot exceed output_tokens") + return self + class CostEstimateResponse(LiteLLMPydanticObjectBase): """Response body for /cost/estimate endpoint.""" @@ -5147,6 +5164,9 @@ class CostEstimateResponse(LiteLLMPydanticObjectBase): model: str input_tokens: int output_tokens: int + cache_read_input_tokens: int = 0 + cache_creation_input_tokens: int = 0 + reasoning_tokens: int = 0 num_requests_per_day: int | None = None num_requests_per_month: int | None = None # Per-request costs @@ -5154,17 +5174,33 @@ class CostEstimateResponse(LiteLLMPydanticObjectBase): input_cost_per_request: float = Field(description="Input token cost per request (before margin)") output_cost_per_request: float = Field(description="Output token cost per request (before margin)") margin_cost_per_request: float = Field(default=0.0, description="Margin/fee added per request") + cache_read_cost_per_request: float = Field(default=0.0, description="Cache-read share of input_cost_per_request") + cache_creation_cost_per_request: float = Field( + default=0.0, description="Cache-write share of input_cost_per_request" + ) + reasoning_cost_per_request: float = Field(default=0.0, description="Reasoning share of output_cost_per_request") # Daily costs (if num_requests_per_day provided) daily_cost: float | None = Field(default=None, description="Total daily cost (includes margin)") daily_input_cost: float | None = Field(default=None, description="Daily input token cost") daily_output_cost: float | None = Field(default=None, description="Daily output token cost") daily_margin_cost: float | None = Field(default=None, description="Daily margin/fee") + daily_cache_read_cost: float | None = Field(default=None, description="Cache-read share of daily_input_cost") + daily_cache_creation_cost: float | None = Field(default=None, description="Cache-write share of daily_input_cost") + daily_reasoning_cost: float | None = Field(default=None, description="Reasoning share of daily_output_cost") # Monthly costs (if num_requests_per_month provided) monthly_cost: float | None = Field(default=None, description="Total monthly cost (includes margin)") monthly_input_cost: float | None = Field(default=None, description="Monthly input token cost") monthly_output_cost: float | None = Field(default=None, description="Monthly output token cost") monthly_margin_cost: float | None = Field(default=None, description="Monthly margin/fee") + monthly_cache_read_cost: float | None = Field(default=None, description="Cache-read share of monthly_input_cost") + monthly_cache_creation_cost: float | None = Field( + default=None, description="Cache-write share of monthly_input_cost" + ) + monthly_reasoning_cost: float | None = Field(default=None, description="Reasoning share of monthly_output_cost") # Pricing info input_cost_per_token: float | None = None output_cost_per_token: float | None = None + cache_read_input_token_cost: float | None = Field(default=None, description="Rate billed per cache-read token") + cache_creation_input_token_cost: float | None = Field(default=None, description="Rate billed per cache-write token") + output_cost_per_reasoning_token: float | None = Field(default=None, description="Rate billed per reasoning token") provider: str | None = None diff --git a/litellm/proxy/management_endpoints/cost_tracking_settings.py b/litellm/proxy/management_endpoints/cost_tracking_settings.py index 204051c3715..51f31a757cd 100644 --- a/litellm/proxy/management_endpoints/cost_tracking_settings.py +++ b/litellm/proxy/management_endpoints/cost_tracking_settings.py @@ -27,7 +27,15 @@ from litellm.proxy._types import ( UserAPIKeyAuth, ) from litellm.proxy.auth.user_api_key_auth import user_api_key_auth -from litellm.types.utils import CostPerToken, LlmProvidersSet, ModelInfo +from litellm.types.utils import ( + CostBreakdown, + CostPerToken, + LlmProvidersSet, + ModelInfo, + ModelResponse, + PromptTokensDetailsWrapper, + Usage, +) router: Final = APIRouter() @@ -46,13 +54,15 @@ def _configured_price(key: str, sources: tuple[Mapping[str, object], ...]) -> fl def _extract_custom_pricing( - litellm_params: Mapping[str, object], model_info: Mapping[str, object] + litellm_params: Mapping[str, object], model_info: Mapping[str, object], builtin: ModelInfo | None ) -> CostPerToken | None: """ Pull per-token pricing configured on a deployment so on-prem / self-hosted models (absent from the public cost map) still estimate a real cost. Pricing may live on ``litellm_params`` or ``model_info``; ``litellm_params`` - wins, matching the router's cost-map registration precedence. + wins, matching the router's cost-map registration precedence. Cache rates the + deployment leaves unset come from the backend model's built-in entry, then its + own input rate, again matching what the router registers for live billing. """ sources: Final = (litellm_params, model_info) input_price: Final = _configured_price("input_cost_per_token", sources) @@ -61,9 +71,15 @@ def _extract_custom_pricing( if input_price is None and output_price is None: return None + input_rate: Final = input_price or 0.0 + cache_sources: Final = sources if builtin is None else (*sources, builtin) + cache_read_price: Final = _configured_price("cache_read_input_token_cost", cache_sources) + cache_creation_price: Final = _configured_price("cache_creation_input_token_cost", cache_sources) return CostPerToken( - input_cost_per_token=input_price or 0.0, + input_cost_per_token=input_rate, output_cost_per_token=output_price or 0.0, + cache_read_input_token_cost=input_rate if cache_read_price is None else cache_read_price, + cache_creation_input_token_cost=input_rate if cache_creation_price is None else cache_creation_price, ) @@ -98,17 +114,14 @@ def _resolve_model_for_cost_lookup(model: str) -> ResolvedCostModel: model_info: Final = first_deployment.get("model_info", {}) custom_llm_provider: Final = litellm_params.get("custom_llm_provider") provider: Final = str(custom_llm_provider) if custom_llm_provider is not None else None - custom_cost_per_token: Final = _extract_custom_pricing(litellm_params, model_info) - - # Check base_model first (needed for Azure custom deployment names) + # base_model wins (needed for Azure custom deployment names) base_model: Final = model_info.get("base_model") or litellm_params.get("base_model") - if base_model: - verbose_proxy_logger.debug("Resolved model '%s' to base_model '%s' from router", model, base_model) - return ResolvedCostModel(str(base_model), provider, custom_cost_per_token) - - resolved_model: Final = litellm_params.get("model") + resolved_model: Final = base_model or litellm_params.get("model") if resolved_model: verbose_proxy_logger.debug("Resolved model '%s' to '%s' from router", model, resolved_model) + custom_cost_per_token: Final = _extract_custom_pricing( + litellm_params, model_info, _lookup_model_info(str(resolved_model)) + ) return ResolvedCostModel(str(resolved_model), provider, custom_cost_per_token) except Exception as e: verbose_proxy_logger.debug("Could not resolve model '%s' from router: %s", model, e) @@ -117,19 +130,97 @@ def _resolve_model_for_cost_lookup(model: str) -> ResolvedCostModel: return ResolvedCostModel(model, None, None) -def _calculate_period_costs(num_requests, cost_per_request, input_cost, output_cost, margin_cost): - """ - Calculate costs for a given number of requests. +@dataclass(frozen=True, slots=True) +class CostLines: + """Cost of one request split the way the spend logs split it: the cache lines are + shares of input_cost and the reasoning line is a share of output_cost.""" - Returns tuple of (total_cost, input_cost, output_cost, margin_cost) or all None if num_requests is None/0. - """ - if not num_requests: - return None, None, None, None - return ( - cost_per_request * num_requests, - input_cost * num_requests, - output_cost * num_requests, - margin_cost * num_requests, + total_cost: float + input_cost: float + output_cost: float + margin_cost: float + cache_read_cost: float + cache_creation_cost: float + reasoning_cost: float + + def times(self, num_requests: int | None) -> "CostLines | None": + if not num_requests: + return None + return CostLines( + total_cost=self.total_cost * num_requests, + input_cost=self.input_cost * num_requests, + output_cost=self.output_cost * num_requests, + margin_cost=self.margin_cost * num_requests, + cache_read_cost=self.cache_read_cost * num_requests, + cache_creation_cost=self.cache_creation_cost * num_requests, + reasoning_cost=self.reasoning_cost * num_requests, + ) + + +def _cost_lines(cost_per_request: float, cost_breakdown: CostBreakdown | None) -> CostLines: + breakdown: Final = cost_breakdown if cost_breakdown is not None else CostBreakdown() + return CostLines( + total_cost=cost_per_request, + input_cost=breakdown.get("input_cost", 0.0), + output_cost=breakdown.get("output_cost", 0.0), + margin_cost=breakdown.get("margin_total_amount", 0.0), + cache_read_cost=breakdown.get("cache_read_cost", 0.0), + cache_creation_cost=breakdown.get("cache_creation_cost", 0.0), + reasoning_cost=breakdown.get("reasoning_cost", 0.0), + ) + + +@dataclass(frozen=True, slots=True) +class EffectiveTokenRates: + input_cost_per_token: float | None + output_cost_per_token: float | None + cache_read_input_token_cost: float | None + cache_creation_input_token_cost: float | None + output_cost_per_reasoning_token: float | None + + +def _custom_token_rates(custom_cost_per_token: CostPerToken) -> EffectiveTokenRates: + input_rate: Final = custom_cost_per_token["input_cost_per_token"] + output_rate: Final = custom_cost_per_token["output_cost_per_token"] + return EffectiveTokenRates( + input_cost_per_token=input_rate, + output_cost_per_token=output_rate, + cache_read_input_token_cost=custom_cost_per_token.get("cache_read_input_token_cost", input_rate), + cache_creation_input_token_cost=custom_cost_per_token.get("cache_creation_input_token_cost", input_rate), + output_cost_per_reasoning_token=output_rate, + ) + + +def _cost_map_token_rates(model_info: ModelInfo | None) -> EffectiveTokenRates: + """Base rates the cost calculator bills flat usage at: a cost-map model without a cache price + bills cache tokens at zero, and one without a reasoning price bills reasoning at the output rate.""" + if model_info is None: + return EffectiveTokenRates(None, None, None, None, None) + sources: Final = (model_info,) + output_rate: Final = _configured_price("output_cost_per_token", sources) + reasoning_rate: Final = _configured_price("output_cost_per_reasoning_token", sources) + return EffectiveTokenRates( + input_cost_per_token=_configured_price("input_cost_per_token", sources), + output_cost_per_token=output_rate, + cache_read_input_token_cost=_configured_price("cache_read_input_token_cost", sources) or 0.0, + cache_creation_input_token_cost=_configured_price("cache_creation_input_token_cost", sources) or 0.0, + output_cost_per_reasoning_token=output_rate if reasoning_rate is None else reasoning_rate, + ) + + +def _usage_for_estimate(request: CostEstimateRequest) -> Usage: + cache_tokens: Final = request.cache_read_input_tokens + request.cache_creation_input_tokens + return Usage( + prompt_tokens=request.input_tokens, + completion_tokens=request.output_tokens, + total_tokens=request.input_tokens + request.output_tokens, + reasoning_tokens=request.reasoning_tokens, + prompt_tokens_details=PromptTokensDetailsWrapper( + cached_tokens=request.cache_read_input_tokens, + cache_creation_tokens=request.cache_creation_input_tokens, + ) + if cache_tokens + else None, ) @@ -530,11 +621,14 @@ async def estimate_cost( - model: Model name (e.g., "gpt-4", "claude-3-opus") - input_tokens: Expected input tokens per request - output_tokens: Expected output tokens per request + - cache_read_input_tokens: Cache-read tokens per request, counted within input_tokens (optional) + - cache_creation_input_tokens: Cache-write tokens per request, counted within input_tokens (optional) + - reasoning_tokens: Reasoning tokens per request, counted within output_tokens (optional) - num_requests_per_day: Number of requests per day (optional) - num_requests_per_month: Number of requests per month (optional) Returns cost breakdown including: - - Per-request costs (input, output, margin) + - Per-request costs (input, output, margin, plus the cache-read, cache-write and reasoning shares) - Daily costs (if num_requests_per_day provided) - Monthly costs (if num_requests_per_month provided) @@ -543,14 +637,15 @@ async def estimate_cost( { "model": "gpt-4", "input_tokens": 1000, + "cache_read_input_tokens": 800, "output_tokens": 500, + "reasoning_tokens": 200, "num_requests_per_day": 100, "num_requests_per_month": 3000 } ``` """ from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj - from litellm.types.utils import ModelResponse, Usage # Resolve model name (handles router aliases like 'e-model-router' -> 'azure_ai/gpt-4') resolved: Final = _resolve_model_for_cost_lookup(request.model) @@ -559,15 +654,7 @@ async def estimate_cost( verbose_proxy_logger.debug("Cost estimate: request.model='%s' resolved to '%s'", request.model, resolved_model) - # Create a mock response with usage for completion_cost - mock_response: Final = ModelResponse( - model=resolved_model, - usage=Usage( - prompt_tokens=request.input_tokens, - completion_tokens=request.output_tokens, - total_tokens=request.input_tokens + request.output_tokens, - ), - ) + mock_response: Final = ModelResponse(model=resolved_model, usage=_usage_for_estimate(request)) # Create a logging object to capture cost breakdown litellm_logging_obj: Final = LiteLLMLoggingObj( @@ -597,75 +684,53 @@ async def estimate_cost( }, ) - # Get cost breakdown from the logging object - cost_breakdown: Final = litellm_logging_obj.cost_breakdown - - input_cost: Final = cost_breakdown.get("input_cost", 0.0) if cost_breakdown else 0.0 - output_cost: Final = cost_breakdown.get("output_cost", 0.0) if cost_breakdown else 0.0 - margin_cost: Final = cost_breakdown.get("margin_total_amount", 0.0) if cost_breakdown else 0.0 + per_request: Final = _cost_lines(cost_per_request, litellm_logging_obj.cost_breakdown) + daily: Final = per_request.times(request.num_requests_per_day) + monthly: Final = per_request.times(request.num_requests_per_month) model_info: Final = _lookup_model_info(resolved_model) - mapped_input_price: Final = model_info.get("input_cost_per_token") if model_info is not None else None - mapped_output_price: Final = model_info.get("output_cost_per_token") if model_info is not None else None + rates: Final = ( + _custom_token_rates(resolved.custom_cost_per_token) + if resolved.custom_cost_per_token is not None + else _cost_map_token_rates(model_info) + ) mapped_provider: Final = model_info.get("litellm_provider") if model_info is not None else None - - input_cost_per_token: Final = ( - resolved.custom_cost_per_token["input_cost_per_token"] - if resolved.custom_cost_per_token is not None - else mapped_input_price - ) - output_cost_per_token: Final = ( - resolved.custom_cost_per_token["output_cost_per_token"] - if resolved.custom_cost_per_token is not None - else mapped_output_price - ) custom_llm_provider: Final = mapped_provider if mapped_provider is not None else resolved_provider - # Calculate daily and monthly costs - ( - daily_cost, - daily_input_cost, - daily_output_cost, - daily_margin_cost, - ) = _calculate_period_costs( - num_requests=request.num_requests_per_day, - cost_per_request=cost_per_request, - input_cost=input_cost, - output_cost=output_cost, - margin_cost=margin_cost, - ) - ( - monthly_cost, - monthly_input_cost, - monthly_output_cost, - monthly_margin_cost, - ) = _calculate_period_costs( - num_requests=request.num_requests_per_month, - cost_per_request=cost_per_request, - input_cost=input_cost, - output_cost=output_cost, - margin_cost=margin_cost, - ) - return CostEstimateResponse( model=request.model, input_tokens=request.input_tokens, output_tokens=request.output_tokens, + cache_read_input_tokens=request.cache_read_input_tokens, + cache_creation_input_tokens=request.cache_creation_input_tokens, + reasoning_tokens=request.reasoning_tokens, num_requests_per_day=request.num_requests_per_day, num_requests_per_month=request.num_requests_per_month, - cost_per_request=cost_per_request, - input_cost_per_request=input_cost, - output_cost_per_request=output_cost, - margin_cost_per_request=margin_cost, - daily_cost=daily_cost, - daily_input_cost=daily_input_cost, - daily_output_cost=daily_output_cost, - daily_margin_cost=daily_margin_cost, - monthly_cost=monthly_cost, - monthly_input_cost=monthly_input_cost, - monthly_output_cost=monthly_output_cost, - monthly_margin_cost=monthly_margin_cost, - input_cost_per_token=input_cost_per_token, - output_cost_per_token=output_cost_per_token, + cost_per_request=per_request.total_cost, + input_cost_per_request=per_request.input_cost, + output_cost_per_request=per_request.output_cost, + margin_cost_per_request=per_request.margin_cost, + cache_read_cost_per_request=per_request.cache_read_cost, + cache_creation_cost_per_request=per_request.cache_creation_cost, + reasoning_cost_per_request=per_request.reasoning_cost, + daily_cost=daily.total_cost if daily is not None else None, + daily_input_cost=daily.input_cost if daily is not None else None, + daily_output_cost=daily.output_cost if daily is not None else None, + daily_margin_cost=daily.margin_cost if daily is not None else None, + daily_cache_read_cost=daily.cache_read_cost if daily is not None else None, + daily_cache_creation_cost=daily.cache_creation_cost if daily is not None else None, + daily_reasoning_cost=daily.reasoning_cost if daily is not None else None, + monthly_cost=monthly.total_cost if monthly is not None else None, + monthly_input_cost=monthly.input_cost if monthly is not None else None, + monthly_output_cost=monthly.output_cost if monthly is not None else None, + monthly_margin_cost=monthly.margin_cost if monthly is not None else None, + monthly_cache_read_cost=monthly.cache_read_cost if monthly is not None else None, + monthly_cache_creation_cost=monthly.cache_creation_cost if monthly is not None else None, + monthly_reasoning_cost=monthly.reasoning_cost if monthly is not None else None, + input_cost_per_token=rates.input_cost_per_token, + output_cost_per_token=rates.output_cost_per_token, + cache_read_input_token_cost=rates.cache_read_input_token_cost, + cache_creation_input_token_cost=rates.cache_creation_input_token_cost, + output_cost_per_reasoning_token=rates.output_cost_per_reasoning_token, provider=custom_llm_provider, ) diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py index 59f0938e338..7065003839b 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py @@ -3874,6 +3874,70 @@ def test_token_type_cost_breakdown_reconciles_with_generic_total(_local_model_co assert text_input_cost + breakdown.cache_read_cost == pytest.approx(prompt_cost) +def _custom_priced_usage() -> Usage: + return Usage( + prompt_tokens=1000, + completion_tokens=500, + total_tokens=1500, + prompt_tokens_details=PromptTokensDetailsWrapper(cached_tokens=800, cache_creation_tokens=100), + completion_tokens_details=CompletionTokensDetailsWrapper(reasoning_tokens=200), + ) + + +def test_token_type_cost_breakdown_prices_custom_pricing_from_its_flat_rates(): + """ + A custom-priced deployment, usually absent from the cost map, used to get zero cache and + reasoning lines while its total already billed cache tokens at the custom cache rates. + The lines must come from the same flat rates: a configured cache rate, else the input + rate for cache tokens and the output rate for reasoning tokens. + """ + from litellm.types.utils import CostPerToken + + breakdown = get_token_type_cost_breakdown( + model="openai/onprem-model", + custom_llm_provider="openai", + usage=_custom_priced_usage(), + custom_cost_per_token=CostPerToken( + input_cost_per_token=1e-6, output_cost_per_token=2e-6, cache_read_input_token_cost=1e-7 + ), + ) + + assert breakdown.cache_read_cost == pytest.approx(800 * 1e-7) + assert breakdown.cache_creation_cost == pytest.approx(100 * 1e-6) + assert breakdown.reasoning_cost == pytest.approx(200 * 2e-6) + + +def test_token_type_cost_breakdown_reconciles_with_custom_pricing_totals(): + from litellm.cost_calculator import cost_per_token + from litellm.types.utils import CostPerToken + + usage = _custom_priced_usage() + custom_cost_per_token = CostPerToken( + input_cost_per_token=1e-6, + output_cost_per_token=2e-6, + cache_read_input_token_cost=1e-7, + cache_creation_input_token_cost=1.25e-6, + ) + + prompt_cost, completion_cost = cost_per_token( + model="openai/onprem-model", + custom_llm_provider="openai", + prompt_tokens=1000, + completion_tokens=500, + usage_object=usage, + custom_cost_per_token=custom_cost_per_token, + ) + breakdown = get_token_type_cost_breakdown( + model="openai/onprem-model", + custom_llm_provider="openai", + usage=usage, + custom_cost_per_token=custom_cost_per_token, + ) + + assert 100 * 1e-6 + breakdown.cache_read_cost + breakdown.cache_creation_cost == pytest.approx(prompt_cost) + assert 300 * 2e-6 + breakdown.reasoning_cost == pytest.approx(completion_cost) + + def test_token_type_cost_breakdown_zero_without_special_tokens(_local_model_cost_map): usage = Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150) diff --git a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py index ec62cc47018..6d3ef2ffea9 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py +++ b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py @@ -8,9 +8,11 @@ from unittest.mock import AsyncMock, MagicMock, patch import pytest from fastapi.testclient import TestClient +from pydantic import ValidationError import litellm +from litellm.proxy._types import CostEstimateRequest from litellm.proxy.management_endpoints.cost_tracking_settings import router from litellm.proxy.proxy_server import app @@ -789,13 +791,13 @@ INPUT_TOKENS = 1000 OUTPUT_TOKENS = 500 -def _router_pricing(**pricing: float) -> MagicMock: +def _router_pricing(model: str = AN_UNDERLYING_MODEL, **pricing: float) -> MagicMock: mock_router = MagicMock() mock_router.get_model_list.return_value = [ { "model_name": AN_ALIAS, "litellm_params": { - "model": AN_UNDERLYING_MODEL, + "model": model, "custom_llm_provider": "openai", **pricing, }, @@ -909,3 +911,184 @@ class TestEstimateCostPeriodTotals: assert response.cost_per_request == pytest.approx(0.0022) assert response.daily_margin_cost == pytest.approx(0.02) assert response.daily_cost == pytest.approx(0.22) + + +CACHE_READ_TOKENS = 800 +CACHE_CREATION_TOKENS = 100 +REASONING_TOKENS = 200 +TEXT_INPUT_TOKENS = INPUT_TOKENS - CACHE_READ_TOKENS - CACHE_CREATION_TOKENS +TEXT_OUTPUT_TOKENS = OUTPUT_TOKENS - REASONING_TOKENS + + +async def _estimate_with_cache_and_reasoning(mock_router: MagicMock | None, model: str = AN_ALIAS, **overrides: int): + return await _estimate( + mock_router, + model=model, + cache_read_input_tokens=CACHE_READ_TOKENS, + cache_creation_input_tokens=CACHE_CREATION_TOKENS, + reasoning_tokens=REASONING_TOKENS, + **overrides, + ) + + +class TestEstimateCostCacheAndReasoningTokens: + @pytest.mark.asyncio + async def test_a_mapped_model_bills_cache_and_reasoning_tokens_at_their_own_rates(self, monkeypatch): + monkeypatch.setitem( + litellm.model_cost, + A_MAPPED_MODEL, + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "cache_read_input_token_cost": 3e-7, + "cache_creation_input_token_cost": 3.75e-6, + "output_cost_per_reasoning_token": 1e-5, + "litellm_provider": "openai", + "mode": "chat", + }, + ) + + response = await _estimate_with_cache_and_reasoning(None, model=A_MAPPED_MODEL, num_requests_per_day=10) + + assert response.cache_read_cost_per_request == pytest.approx(CACHE_READ_TOKENS * 3e-7) + assert response.cache_creation_cost_per_request == pytest.approx(CACHE_CREATION_TOKENS * 3.75e-6) + assert response.reasoning_cost_per_request == pytest.approx(REASONING_TOKENS * 1e-5) + assert response.input_cost_per_request == pytest.approx( + TEXT_INPUT_TOKENS * 3e-6 + CACHE_READ_TOKENS * 3e-7 + CACHE_CREATION_TOKENS * 3.75e-6 + ) + assert response.output_cost_per_request == pytest.approx(TEXT_OUTPUT_TOKENS * 15e-6 + REASONING_TOKENS * 1e-5) + assert response.cost_per_request == pytest.approx( + response.input_cost_per_request + response.output_cost_per_request + ) + assert response.daily_cache_read_cost == pytest.approx(10 * CACHE_READ_TOKENS * 3e-7) + assert response.daily_cache_creation_cost == pytest.approx(10 * CACHE_CREATION_TOKENS * 3.75e-6) + assert response.daily_reasoning_cost == pytest.approx(10 * REASONING_TOKENS * 1e-5) + assert response.monthly_cache_read_cost is None + assert response.cache_read_input_token_cost == pytest.approx(3e-7) + assert response.cache_creation_input_token_cost == pytest.approx(3.75e-6) + assert response.output_cost_per_reasoning_token == pytest.approx(1e-5) + assert ( + response.cache_read_input_tokens, + response.cache_creation_input_tokens, + response.reasoning_tokens, + ) == (CACHE_READ_TOKENS, CACHE_CREATION_TOKENS, REASONING_TOKENS) + + @pytest.mark.asyncio + async def test_a_model_without_cache_or_reasoning_prices_estimates_what_the_proxy_bills(self, monkeypatch): + """The cost calculator bills cache tokens of a cost-map model without cache prices at zero + and its reasoning tokens at the output rate. The estimate reports those effective rates.""" + monkeypatch.setitem( + litellm.model_cost, + A_MAPPED_MODEL, + {"input_cost_per_token": 5e-6, "output_cost_per_token": 6e-6, "litellm_provider": "openai", "mode": "chat"}, + ) + + response = await _estimate_with_cache_and_reasoning(None, model=A_MAPPED_MODEL) + + assert response.cache_read_cost_per_request == 0.0 + assert response.cache_creation_cost_per_request == 0.0 + assert response.reasoning_cost_per_request == pytest.approx(REASONING_TOKENS * 6e-6) + assert response.input_cost_per_request == pytest.approx(TEXT_INPUT_TOKENS * 5e-6) + assert response.cost_per_request == pytest.approx(TEXT_INPUT_TOKENS * 5e-6 + OUTPUT_TOKENS * 6e-6) + assert response.cache_read_input_token_cost == 0.0 + assert response.cache_creation_input_token_cost == 0.0 + assert response.output_cost_per_reasoning_token == pytest.approx(6e-6) + + @pytest.mark.asyncio + async def test_a_request_without_cache_or_reasoning_tokens_estimates_as_before(self, monkeypatch): + monkeypatch.setitem( + litellm.model_cost, + A_MAPPED_MODEL, + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "cache_read_input_token_cost": 3e-7, + "cache_creation_input_token_cost": 3.75e-6, + "output_cost_per_reasoning_token": 1e-5, + "litellm_provider": "openai", + "mode": "chat", + }, + ) + + response = await _estimate(None, model=A_MAPPED_MODEL, num_requests_per_day=10) + + assert response.cost_per_request == pytest.approx(INPUT_TOKENS * 3e-6 + OUTPUT_TOKENS * 15e-6) + assert response.cache_read_cost_per_request == 0.0 + assert response.cache_creation_cost_per_request == 0.0 + assert response.reasoning_cost_per_request == 0.0 + assert response.daily_cache_read_cost == 0.0 + assert response.daily_reasoning_cost == 0.0 + + @pytest.mark.asyncio + async def test_a_custom_priced_deployment_bills_cache_and_reasoning_tokens_from_its_flat_rates(self): + response = await _estimate_with_cache_and_reasoning( + _router_pricing(input_cost_per_token=1e-6, output_cost_per_token=2e-6, cache_read_input_token_cost=1e-7) + ) + + assert response.cache_read_cost_per_request == pytest.approx(CACHE_READ_TOKENS * 1e-7) + assert response.cache_creation_cost_per_request == pytest.approx(CACHE_CREATION_TOKENS * 1e-6) + assert response.reasoning_cost_per_request == pytest.approx(REASONING_TOKENS * 2e-6) + assert response.cost_per_request == pytest.approx( + TEXT_INPUT_TOKENS * 1e-6 + CACHE_READ_TOKENS * 1e-7 + CACHE_CREATION_TOKENS * 1e-6 + OUTPUT_TOKENS * 2e-6 + ) + assert response.cache_read_input_token_cost == pytest.approx(1e-7) + assert response.cache_creation_input_token_cost == pytest.approx(1e-6) + assert response.output_cost_per_reasoning_token == pytest.approx(2e-6) + + @pytest.mark.asyncio + async def test_a_custom_priced_deployment_of_a_mapped_model_inherits_its_built_in_cache_rates(self, monkeypatch): + monkeypatch.setitem( + litellm.model_cost, + A_MAPPED_MODEL, + { + "input_cost_per_token": 5e-6, + "output_cost_per_token": 6e-6, + "cache_read_input_token_cost": 5e-7, + "cache_creation_input_token_cost": 6.25e-6, + "litellm_provider": "openai", + "mode": "chat", + }, + ) + + response = await _estimate_with_cache_and_reasoning( + _router_pricing(model=A_MAPPED_MODEL, input_cost_per_token=1e-6, output_cost_per_token=2e-6) + ) + + assert response.cache_read_cost_per_request == pytest.approx(CACHE_READ_TOKENS * 5e-7) + assert response.cache_creation_cost_per_request == pytest.approx(CACHE_CREATION_TOKENS * 6.25e-6) + assert response.input_cost_per_request == pytest.approx( + TEXT_INPUT_TOKENS * 1e-6 + CACHE_READ_TOKENS * 5e-7 + CACHE_CREATION_TOKENS * 6.25e-6 + ) + assert response.cache_read_input_token_cost == pytest.approx(5e-7) + assert response.cache_creation_input_token_cost == pytest.approx(6.25e-6) + + +class TestCostEstimateRequestTokenSubsets: + def test_cache_tokens_beyond_the_input_tokens_are_rejected(self): + with pytest.raises(ValidationError, match="cannot exceed input_tokens"): + CostEstimateRequest( + model=AN_ALIAS, + input_tokens=INPUT_TOKENS, + output_tokens=OUTPUT_TOKENS, + cache_read_input_tokens=INPUT_TOKENS, + cache_creation_input_tokens=1, + ) + + def test_reasoning_tokens_beyond_the_output_tokens_are_rejected(self): + with pytest.raises(ValidationError, match="cannot exceed output_tokens"): + CostEstimateRequest( + model=AN_ALIAS, + input_tokens=INPUT_TOKENS, + output_tokens=OUTPUT_TOKENS, + reasoning_tokens=OUTPUT_TOKENS + 1, + ) + + def test_the_endpoint_answers_422_when_cache_tokens_exceed_input_tokens(self): + response = client.post( + "/cost/estimate", + headers={"Authorization": "Bearer sk-1234"}, + json={"model": AN_ALIAS, "input_tokens": 1000, "output_tokens": 100, "cache_read_input_tokens": 8000}, + ) + + assert response.status_code == 422 + assert "cannot exceed input_tokens" in response.text diff --git a/tests/test_litellm/test_cost_calculator.py b/tests/test_litellm/test_cost_calculator.py index f8fa2231597..910587d3dc5 100644 --- a/tests/test_litellm/test_cost_calculator.py +++ b/tests/test_litellm/test_cost_calculator.py @@ -3736,6 +3736,56 @@ def test_completion_cost_logs_reasoning_and_cache_breakdown(_local_model_cost_ma assert logging_obj.cost_breakdown["cache_read_cost"] == pytest.approx(100 * 3e-08) +def test_completion_cost_logs_cache_and_reasoning_breakdown_for_custom_pricing(): + """ + A custom-priced deployment bills cache tokens at its custom cache rates, but the + breakdown stored for the spend logs carried no cache or reasoning lines for it. + """ + from datetime import datetime + + from litellm.litellm_core_utils.litellm_logging import Logging + from litellm.types.utils import CompletionTokensDetailsWrapper, CostPerToken + + logging_obj = Logging( + model="openai/onprem-model", + messages=[{"role": "user", "content": "Hello"}], + stream=False, + call_type="completion", + start_time=datetime.now(), + litellm_call_id="custom-pricing-breakdown", + function_id="f", + ) + response = ModelResponse( + model="openai/onprem-model", + usage=Usage( + prompt_tokens=1000, + completion_tokens=500, + total_tokens=1500, + prompt_tokens_details=PromptTokensDetailsWrapper(cached_tokens=800, cache_creation_tokens=100), + completion_tokens_details=CompletionTokensDetailsWrapper(reasoning_tokens=200), + ), + ) + + total = completion_cost( + completion_response=response, + model="openai/onprem-model", + custom_llm_provider="openai", + custom_cost_per_token=CostPerToken( + input_cost_per_token=1e-6, + output_cost_per_token=2e-6, + cache_read_input_token_cost=1e-7, + cache_creation_input_token_cost=1.25e-6, + ), + litellm_logging_obj=logging_obj, + ) + + assert logging_obj.cost_breakdown is not None + assert logging_obj.cost_breakdown["cache_read_cost"] == pytest.approx(800 * 1e-7) + assert logging_obj.cost_breakdown["cache_creation_cost"] == pytest.approx(100 * 1.25e-6) + assert logging_obj.cost_breakdown["reasoning_cost"] == pytest.approx(200 * 2e-6) + assert total == pytest.approx(100 * 1e-6 + 800 * 1e-7 + 100 * 1.25e-6 + 500 * 2e-6) + + def test_cost_per_token_per_second_pricing(monkeypatch): """ Models priced by duration (input/output_cost_per_second) with no per-token rates diff --git a/ui/litellm-dashboard/src/lib/http/schema.d.ts b/ui/litellm-dashboard/src/lib/http/schema.d.ts index 6fb08445aff..6a9c86cc1b5 100644 --- a/ui/litellm-dashboard/src/lib/http/schema.d.ts +++ b/ui/litellm-dashboard/src/lib/http/schema.d.ts @@ -3301,11 +3301,14 @@ export interface paths { * - model: Model name (e.g., "gpt-4", "claude-3-opus") * - input_tokens: Expected input tokens per request * - output_tokens: Expected output tokens per request + * - cache_read_input_tokens: Cache-read tokens per request, counted within input_tokens (optional) + * - cache_creation_input_tokens: Cache-write tokens per request, counted within input_tokens (optional) + * - reasoning_tokens: Reasoning tokens per request, counted within output_tokens (optional) * - num_requests_per_day: Number of requests per day (optional) * - num_requests_per_month: Number of requests per month (optional) * * Returns cost breakdown including: - * - Per-request costs (input, output, margin) + * - Per-request costs (input, output, margin, plus the cache-read, cache-write and reasoning shares) * - Daily costs (if num_requests_per_day provided) * - Monthly costs (if num_requests_per_month provided) * @@ -3314,7 +3317,9 @@ export interface paths { * { * "model": "gpt-4", * "input_tokens": 1000, + * "cache_read_input_tokens": 800, * "output_tokens": 500, + * "reasoning_tokens": 200, * "num_requests_per_day": 100, * "num_requests_per_month": 3000 * } @@ -26392,6 +26397,18 @@ export interface components { * @description Request body for /cost/estimate endpoint. */ CostEstimateRequest: { + /** + * Cache Creation Input Tokens + * @description Input tokens written to the prompt cache; counted within input_tokens + * @default 0 + */ + cache_creation_input_tokens: number; + /** + * Cache Read Input Tokens + * @description Input tokens read from the prompt cache; counted within input_tokens + * @default 0 + */ + cache_read_input_tokens: number; /** * Input Tokens * @description Expected input tokens per request @@ -26417,17 +26434,65 @@ export interface components { * @description Expected output tokens per request */ output_tokens: number; + /** + * Reasoning Tokens + * @description Reasoning tokens the model emits; counted within output_tokens + * @default 0 + */ + reasoning_tokens: number; }; /** * CostEstimateResponse * @description Response body for /cost/estimate endpoint. */ CostEstimateResponse: { + /** + * Cache Creation Cost Per Request + * @description Cache-write share of input_cost_per_request + * @default 0 + */ + cache_creation_cost_per_request: number; + /** + * Cache Creation Input Token Cost + * @description Rate billed per cache-write token + */ + cache_creation_input_token_cost?: number | null; + /** + * Cache Creation Input Tokens + * @default 0 + */ + cache_creation_input_tokens: number; + /** + * Cache Read Cost Per Request + * @description Cache-read share of input_cost_per_request + * @default 0 + */ + cache_read_cost_per_request: number; + /** + * Cache Read Input Token Cost + * @description Rate billed per cache-read token + */ + cache_read_input_token_cost?: number | null; + /** + * Cache Read Input Tokens + * @default 0 + */ + cache_read_input_tokens: number; /** * Cost Per Request * @description Total cost per request (includes margin) */ cost_per_request: number; + /** + * Daily Cache Creation Cost + * @description Cache-write share of daily_input_cost + */ + daily_cache_creation_cost?: number | null; + /** + * Daily Cache Read Cost + * @description Cache-read share of daily_input_cost + */ + daily_cache_read_cost?: number | null; /** * Daily Cost * @description Total daily cost (includes margin) @@ -26448,6 +26513,11 @@ export interface components { * @description Daily output token cost */ daily_output_cost?: number | null; + /** + * Daily Reasoning Cost + * @description Reasoning share of daily_output_cost + */ + daily_reasoning_cost?: number | null; /** * Input Cost Per Request * @description Input token cost per request (before margin) @@ -26465,6 +26535,16 @@ export interface components { margin_cost_per_request: number; /** Model */ model: string; + /** + * Monthly Cache Creation Cost + * @description Cache-write share of monthly_input_cost + */ + monthly_cache_creation_cost?: number | null; + /** + * Monthly Cache Read Cost + * @description Cache-read share of monthly_input_cost + */ + monthly_cache_read_cost?: number | null; /** * Monthly Cost * @description Total monthly cost (includes margin) @@ -26485,10 +26565,20 @@ export interface components { * @description Monthly output token cost */ monthly_output_cost?: number | null; + /** + * Monthly Reasoning Cost + * @description Reasoning share of monthly_output_cost + */ + monthly_reasoning_cost?: number | null; /** Num Requests Per Day */ num_requests_per_day?: number | null; /** Num Requests Per Month */ num_requests_per_month?: number | null; + /** + * Output Cost Per Reasoning Token + * @description Rate billed per reasoning token + */ + output_cost_per_reasoning_token?: number | null; /** * Output Cost Per Request * @description Output token cost per request (before margin) @@ -26500,6 +26590,17 @@ export interface components { output_tokens: number; /** Provider */ provider?: string | null; + /** + * Reasoning Cost Per Request + * @description Reasoning share of output_cost_per_request + * @default 0 + */ + reasoning_cost_per_request: number; + /** + * Reasoning Tokens + * @default 0 + */ + reasoning_tokens: number; }; /** CreateCredentialItem */ CreateCredentialItem: { From 43a02e2dbcd485c3cc5dd31d8b11161da650c356 Mon Sep 17 00:00:00 2001 From: ryan-crabbe-berri Date: Mon, 7 Sep 2026 16:53:41 -0700 Subject: [PATCH 03/26] fix(proxy): report the billed token rates in /cost/estimate The rate fields reported base cost-map prices while the cost lines were billed at the token tier, off-peak window and regional multipliers the calculator picks for the request, so a line did not always equal tokens times its reported rate. get_billed_token_rates now resolves the rates once, the token-type breakdown and the endpoint both read from it, and a tiered-model test asserts every line equals its token count times the rate reported next to it Claude-Session: https://claude.ai/code/session_011Tn3657NkV6ojLqewL64Kb --- .../litellm_core_utils/llm_cost_calc/utils.py | 206 ++++++++++++------ litellm/proxy/_types.py | 6 +- .../cost_tracking_settings.py | 63 ++---- .../llm_cost_calc/test_llm_cost_calc_utils.py | 49 +++++ .../test_cost_tracking_settings.py | 52 ++++- ui/litellm-dashboard/src/lib/http/schema.d.ts | 10 +- 6 files changed, 258 insertions(+), 128 deletions(-) diff --git a/litellm/litellm_core_utils/llm_cost_calc/utils.py b/litellm/litellm_core_utils/llm_cost_calc/utils.py index 8d53c8da3e6..2bb15fb4c48 100644 --- a/litellm/litellm_core_utils/llm_cost_calc/utils.py +++ b/litellm/litellm_core_utils/llm_cost_calc/utils.py @@ -1329,16 +1329,118 @@ def _cache_token_counts(usage: Usage) -> tuple[int, int, CacheCreationTokenDetai ) -def _custom_pricing_token_type_breakdown(usage: Usage, custom_cost_per_token: CostPerToken) -> TokenTypeCostBreakdown: +@dataclass(frozen=True, slots=True) +class BilledTokenRates: + """Per-token rates one request's usage bills at, after token tiers, off-peak windows and the + regional multipliers the totals apply, so each cost line equals its token count times its rate.""" + + input_cost_per_token: float + output_cost_per_token: float + cache_read_input_token_cost: float + cache_creation_input_token_cost: float + cache_creation_input_token_cost_above_1hr: float + output_cost_per_reasoning_token: float + + def scaled(self, multiplier: float) -> "BilledTokenRates": + if multiplier == 1.0: + return self + return BilledTokenRates( + input_cost_per_token=self.input_cost_per_token * multiplier, + output_cost_per_token=self.output_cost_per_token * multiplier, + cache_read_input_token_cost=self.cache_read_input_token_cost * multiplier, + cache_creation_input_token_cost=self.cache_creation_input_token_cost * multiplier, + cache_creation_input_token_cost_above_1hr=self.cache_creation_input_token_cost_above_1hr * multiplier, + output_cost_per_reasoning_token=self.output_cost_per_reasoning_token * multiplier, + ) + + +def _custom_pricing_rates(custom_cost_per_token: CostPerToken) -> BilledTokenRates: """Flat custom pricing has no tiers, uplifts or reasoning rate: cache tokens bill at the configured cache rates (else the input rate) and reasoning at the output rate, as _cost_per_token_custom_pricing_helper does.""" input_rate: Final = custom_cost_per_token["input_cost_per_token"] - cache_read_tokens, cache_creation_tokens, _ = _cache_token_counts(usage) - return TokenTypeCostBreakdown( - reasoning_cost=float(_reasoning_token_count(usage)) * custom_cost_per_token["output_cost_per_token"], - cache_read_cost=float(cache_read_tokens) * custom_cost_per_token.get("cache_read_input_token_cost", input_rate), - cache_creation_cost=float(cache_creation_tokens) - * custom_cost_per_token.get("cache_creation_input_token_cost", input_rate), + output_rate: Final = custom_cost_per_token["output_cost_per_token"] + cache_creation_rate: Final = custom_cost_per_token.get("cache_creation_input_token_cost", input_rate) + return BilledTokenRates( + input_cost_per_token=input_rate, + output_cost_per_token=output_rate, + cache_read_input_token_cost=custom_cost_per_token.get("cache_read_input_token_cost", input_rate), + cache_creation_input_token_cost=cache_creation_rate, + cache_creation_input_token_cost_above_1hr=cache_creation_rate, + output_cost_per_reasoning_token=output_rate, + ) + + +def _cost_map_billed_rates( + model_info: ModelInfo, + usage: Usage, + custom_llm_provider: str | None, + service_tier: str | None, + data_residency: str | None, + vertex_location: str | None, + current_time: datetime | None, +) -> BilledTokenRates: + billing_time: Final = current_time if current_time is not None else datetime.now(timezone.utc) + ( + prompt_base_cost, + completion_base_cost, + cache_creation_cost_rate, + cache_creation_cost_above_1hr_rate, + cache_read_cost_rate, + ) = _get_token_base_cost( + model_info=model_info, + usage=usage, + service_tier=service_tier, + current_time=billing_time, + threshold_is_inclusive=_uses_inclusive_token_thresholds(custom_llm_provider), + ) + reasoning_rate: Final = _resolve_billed_reasoning_rate( + model_info=model_info, + usage=usage, + service_tier=service_tier, + completion_base_cost=completion_base_cost, + current_time=billing_time, + ) + multiplier: Final = ( + _get_regional_uplift_multiplier(model_info, data_residency) + * get_vertex_regional_endpoint_uplift(model_info, vertex_location) + * get_provider_specific_geo_multiplier(model_info=model_info, usage=usage) + ) + return BilledTokenRates( + input_cost_per_token=prompt_base_cost, + output_cost_per_token=completion_base_cost, + cache_read_input_token_cost=cache_read_cost_rate, + cache_creation_input_token_cost=cache_creation_cost_rate, + cache_creation_input_token_cost_above_1hr=cache_creation_cost_above_1hr_rate, + output_cost_per_reasoning_token=reasoning_rate, + ).scaled(multiplier) + + +def get_billed_token_rates( + model: str, + custom_llm_provider: str | None, + usage: Usage, + service_tier: str | None = None, + data_residency: str | None = None, + vertex_location: str | None = None, + current_time: datetime | None = None, + custom_cost_per_token: CostPerToken | None = None, +) -> BilledTokenRates | None: + """Rates the cost calculator bills ``usage`` at, resolved exactly as the totals and the token-type + breakdown resolve them. None when the model's pricing cannot be resolved.""" + if custom_cost_per_token is not None: + return _custom_pricing_rates(custom_cost_per_token) + try: + model_info: Final = get_model_info(model=model, custom_llm_provider=custom_llm_provider) + except Exception: + return None + return _cost_map_billed_rates( + model_info=model_info, + usage=usage, + custom_llm_provider=custom_llm_provider, + service_tier=service_tier, + data_residency=data_residency, + vertex_location=vertex_location, + current_time=current_time, ) @@ -1360,77 +1462,39 @@ def get_token_type_cost_breakdown( cost calculators bypass ``generic_cost_per_token``, because cache tokens always land on ``prompt_tokens_details`` (via the Usage constructor and provider transformations) and reasoning tokens on ``completion_tokens_details``. It reuses - the same rate-resolution primitives as the total-cost path so the breakdown can - never drift from the totals. A deployment billed by ``custom_cost_per_token`` is - priced from those flat rates instead of the cost map, for the same reason. + the same rate resolution as the total-cost path (``get_billed_token_rates``) so the + breakdown can never drift from the totals. A deployment billed by + ``custom_cost_per_token`` is priced from those flat rates instead of the cost map and, + like its totals, bills cache writes flat rather than by their 5m/1h split. Returns zeros (never raises) when the model or its pricing cannot be resolved. """ - if custom_cost_per_token is not None: - return _custom_pricing_token_type_breakdown(usage=usage, custom_cost_per_token=custom_cost_per_token) - - try: - model_info: Final = get_model_info(model=model, custom_llm_provider=custom_llm_provider) - except Exception: + rates: Final = get_billed_token_rates( + model=model, + custom_llm_provider=custom_llm_provider, + usage=usage, + service_tier=service_tier, + data_residency=data_residency, + vertex_location=vertex_location, + current_time=current_time, + custom_cost_per_token=custom_cost_per_token, + ) + if rates is None: return TokenTypeCostBreakdown(0.0, 0.0, 0.0) - billing_time: Final = current_time if current_time is not None else datetime.now(timezone.utc) - ( - _prompt_base_cost, - completion_base_cost, - cache_creation_cost_rate, - cache_creation_cost_above_1hr_rate, - cache_read_cost_rate, - ) = _get_token_base_cost( - model_info=model_info, - usage=usage, - service_tier=service_tier, - current_time=billing_time, - threshold_is_inclusive=_uses_inclusive_token_thresholds(custom_llm_provider), - ) - - reasoning_rate: Final = _resolve_billed_reasoning_rate( - model_info=model_info, - usage=usage, - service_tier=service_tier, - completion_base_cost=completion_base_cost, - current_time=billing_time, - ) - reasoning_cost = float(_reasoning_token_count(usage)) * reasoning_rate - cache_read_tokens, cache_creation_tokens, cache_creation_token_details = _cache_token_counts(usage) - cache_read_cost = float(cache_read_tokens) * cache_read_cost_rate - cache_creation_cost = calculate_cache_writing_cost( - cache_creation_tokens=cache_creation_tokens, - cache_creation_token_details=cache_creation_token_details, - cache_creation_cost_above_1hr=cache_creation_cost_above_1hr_rate, - cache_creation_cost=cache_creation_cost_rate, + cache_creation_cost: Final = ( + float(cache_creation_tokens) * rates.cache_creation_input_token_cost + if custom_cost_per_token is not None + else calculate_cache_writing_cost( + cache_creation_tokens=cache_creation_tokens, + cache_creation_token_details=cache_creation_token_details, + cache_creation_cost_above_1hr=rates.cache_creation_input_token_cost_above_1hr, + cache_creation_cost=rates.cache_creation_input_token_cost, + ) ) - - # Apply the same flat regional-processing uplift the totals get, so per-type - # costs stay reconciled with input_cost/output_cost for regionalized OpenAI hosts. - uplift: Final = _get_regional_uplift_multiplier(model_info, data_residency) - if uplift != 1.0: - reasoning_cost *= uplift - cache_read_cost *= uplift - cache_creation_cost *= uplift - - vertex_uplift: Final = get_vertex_regional_endpoint_uplift(model_info, vertex_location) - if vertex_uplift != 1.0: - reasoning_cost *= vertex_uplift - cache_read_cost *= vertex_uplift - cache_creation_cost *= vertex_uplift - - # Mirror the provider-specific geo uplift (e.g. Anthropic us: 1.1) the totals - # apply, so cache and reasoning line items stay reconciled with them. - geo_multiplier: Final = get_provider_specific_geo_multiplier(model_info=model_info, usage=usage) - if geo_multiplier != 1.0: - reasoning_cost *= geo_multiplier - cache_read_cost *= geo_multiplier - cache_creation_cost *= geo_multiplier - return TokenTypeCostBreakdown( - reasoning_cost=reasoning_cost, - cache_read_cost=cache_read_cost, + reasoning_cost=float(_reasoning_token_count(usage)) * rates.output_cost_per_reasoning_token, + cache_read_cost=float(cache_read_tokens) * rates.cache_read_input_token_cost, cache_creation_cost=cache_creation_cost, ) diff --git a/litellm/proxy/_types.py b/litellm/proxy/_types.py index 65b7d409d7c..3c1cd9bfc69 100644 --- a/litellm/proxy/_types.py +++ b/litellm/proxy/_types.py @@ -5197,9 +5197,9 @@ class CostEstimateResponse(LiteLLMPydanticObjectBase): default=None, description="Cache-write share of monthly_input_cost" ) monthly_reasoning_cost: float | None = Field(default=None, description="Reasoning share of monthly_output_cost") - # Pricing info - input_cost_per_token: float | None = None - output_cost_per_token: float | None = None + # Pricing info: the rates this request's usage bills at, after token tiers and regional multipliers + input_cost_per_token: float | None = Field(default=None, description="Rate billed per input token") + output_cost_per_token: float | None = Field(default=None, description="Rate billed per output token") cache_read_input_token_cost: float | None = Field(default=None, description="Rate billed per cache-read token") cache_creation_input_token_cost: float | None = Field(default=None, description="Rate billed per cache-write token") output_cost_per_reasoning_token: float | None = Field(default=None, description="Rate billed per reasoning token") diff --git a/litellm/proxy/management_endpoints/cost_tracking_settings.py b/litellm/proxy/management_endpoints/cost_tracking_settings.py index 51f31a757cd..17d82fd17e3 100644 --- a/litellm/proxy/management_endpoints/cost_tracking_settings.py +++ b/litellm/proxy/management_endpoints/cost_tracking_settings.py @@ -20,6 +20,7 @@ from pydantic import BaseModel import litellm from litellm._logging import verbose_proxy_logger from litellm.cost_calculator import completion_cost +from litellm.litellm_core_utils.llm_cost_calc.utils import get_billed_token_rates from litellm.proxy._types import ( CommonProxyErrors, CostEstimateRequest, @@ -170,44 +171,6 @@ def _cost_lines(cost_per_request: float, cost_breakdown: CostBreakdown | None) - ) -@dataclass(frozen=True, slots=True) -class EffectiveTokenRates: - input_cost_per_token: float | None - output_cost_per_token: float | None - cache_read_input_token_cost: float | None - cache_creation_input_token_cost: float | None - output_cost_per_reasoning_token: float | None - - -def _custom_token_rates(custom_cost_per_token: CostPerToken) -> EffectiveTokenRates: - input_rate: Final = custom_cost_per_token["input_cost_per_token"] - output_rate: Final = custom_cost_per_token["output_cost_per_token"] - return EffectiveTokenRates( - input_cost_per_token=input_rate, - output_cost_per_token=output_rate, - cache_read_input_token_cost=custom_cost_per_token.get("cache_read_input_token_cost", input_rate), - cache_creation_input_token_cost=custom_cost_per_token.get("cache_creation_input_token_cost", input_rate), - output_cost_per_reasoning_token=output_rate, - ) - - -def _cost_map_token_rates(model_info: ModelInfo | None) -> EffectiveTokenRates: - """Base rates the cost calculator bills flat usage at: a cost-map model without a cache price - bills cache tokens at zero, and one without a reasoning price bills reasoning at the output rate.""" - if model_info is None: - return EffectiveTokenRates(None, None, None, None, None) - sources: Final = (model_info,) - output_rate: Final = _configured_price("output_cost_per_token", sources) - reasoning_rate: Final = _configured_price("output_cost_per_reasoning_token", sources) - return EffectiveTokenRates( - input_cost_per_token=_configured_price("input_cost_per_token", sources), - output_cost_per_token=output_rate, - cache_read_input_token_cost=_configured_price("cache_read_input_token_cost", sources) or 0.0, - cache_creation_input_token_cost=_configured_price("cache_creation_input_token_cost", sources) or 0.0, - output_cost_per_reasoning_token=output_rate if reasoning_rate is None else reasoning_rate, - ) - - def _usage_for_estimate(request: CostEstimateRequest) -> Usage: cache_tokens: Final = request.cache_read_input_tokens + request.cache_creation_input_tokens return Usage( @@ -654,7 +617,8 @@ async def estimate_cost( verbose_proxy_logger.debug("Cost estimate: request.model='%s' resolved to '%s'", request.model, resolved_model) - mock_response: Final = ModelResponse(model=resolved_model, usage=_usage_for_estimate(request)) + usage: Final = _usage_for_estimate(request) + mock_response: Final = ModelResponse(model=resolved_model, usage=usage) # Create a logging object to capture cost breakdown litellm_logging_obj: Final = LiteLLMLoggingObj( @@ -688,12 +652,13 @@ async def estimate_cost( daily: Final = per_request.times(request.num_requests_per_day) monthly: Final = per_request.times(request.num_requests_per_month) - model_info: Final = _lookup_model_info(resolved_model) - rates: Final = ( - _custom_token_rates(resolved.custom_cost_per_token) - if resolved.custom_cost_per_token is not None - else _cost_map_token_rates(model_info) + rates: Final = get_billed_token_rates( + model=resolved_model, + custom_llm_provider=resolved_provider, + usage=usage, + custom_cost_per_token=resolved.custom_cost_per_token, ) + model_info: Final = _lookup_model_info(resolved_model) mapped_provider: Final = model_info.get("litellm_provider") if model_info is not None else None custom_llm_provider: Final = mapped_provider if mapped_provider is not None else resolved_provider @@ -727,10 +692,10 @@ async def estimate_cost( monthly_cache_read_cost=monthly.cache_read_cost if monthly is not None else None, monthly_cache_creation_cost=monthly.cache_creation_cost if monthly is not None else None, monthly_reasoning_cost=monthly.reasoning_cost if monthly is not None else None, - input_cost_per_token=rates.input_cost_per_token, - output_cost_per_token=rates.output_cost_per_token, - cache_read_input_token_cost=rates.cache_read_input_token_cost, - cache_creation_input_token_cost=rates.cache_creation_input_token_cost, - output_cost_per_reasoning_token=rates.output_cost_per_reasoning_token, + input_cost_per_token=rates.input_cost_per_token if rates is not None else None, + output_cost_per_token=rates.output_cost_per_token if rates is not None else None, + cache_read_input_token_cost=rates.cache_read_input_token_cost if rates is not None else None, + cache_creation_input_token_cost=rates.cache_creation_input_token_cost if rates is not None else None, + output_cost_per_reasoning_token=rates.output_cost_per_reasoning_token if rates is not None else None, provider=custom_llm_provider, ) diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py index 7065003839b..4de3c059e63 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py @@ -27,6 +27,7 @@ from litellm.types.utils import ( ) from litellm.litellm_core_utils.llm_cost_calc.utils import ( + BilledTokenRates, CostCalculatorUtils, PromptTokensDetailsResult, TokenRates, @@ -38,6 +39,7 @@ from litellm.litellm_core_utils.llm_cost_calc.utils import ( apply_off_peak_pricing, calculate_cache_writing_cost, generic_cost_per_token, + get_billed_token_rates, get_token_type_cost_breakdown, ) from litellm.types.utils import CacheCreationTokenDetails, Usage @@ -3938,6 +3940,53 @@ def test_token_type_cost_breakdown_reconciles_with_custom_pricing_totals(): assert 300 * 2e-6 + breakdown.reasoning_cost == pytest.approx(completion_cost) +def test_billed_token_rates_follow_the_token_tier_the_breakdown_bills_at(monkeypatch): + monkeypatch.setitem( + litellm.model_cost, + "tiered-cache-model", + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "cache_read_input_token_cost": 3e-7, + "cache_creation_input_token_cost": 3.75e-6, + "input_cost_per_token_above_200k_tokens": 6e-6, + "output_cost_per_token_above_200k_tokens": 3e-5, + "cache_read_input_token_cost_above_200k_tokens": 6e-7, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-6, + "litellm_provider": "openai", + "mode": "chat", + }, + ) + usage = Usage( + prompt_tokens=250_000, + completion_tokens=1_000, + total_tokens=251_000, + prompt_tokens_details=PromptTokensDetailsWrapper(cached_tokens=200_000, cache_creation_tokens=10_000), + completion_tokens_details=CompletionTokensDetailsWrapper(reasoning_tokens=200), + ) + + rates = get_billed_token_rates(model="tiered-cache-model", custom_llm_provider="openai", usage=usage) + breakdown = get_token_type_cost_breakdown(model="tiered-cache-model", custom_llm_provider="openai", usage=usage) + + assert rates == BilledTokenRates( + input_cost_per_token=6e-6, + output_cost_per_token=3e-5, + cache_read_input_token_cost=6e-7, + cache_creation_input_token_cost=7.5e-6, + cache_creation_input_token_cost_above_1hr=0.0, + output_cost_per_reasoning_token=3e-5, + ) + assert breakdown.cache_read_cost == pytest.approx(200_000 * rates.cache_read_input_token_cost) + assert breakdown.cache_creation_cost == pytest.approx(10_000 * rates.cache_creation_input_token_cost) + assert breakdown.reasoning_cost == pytest.approx(200 * rates.output_cost_per_reasoning_token) + + +def test_billed_token_rates_are_none_for_an_unpriced_model(): + usage = Usage(prompt_tokens=10, completion_tokens=5, total_tokens=15) + + assert get_billed_token_rates(model="no-such-model-anywhere", custom_llm_provider="openai", usage=usage) is None + + def test_token_type_cost_breakdown_zero_without_special_tokens(_local_model_cost_map): usage = Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150) diff --git a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py index 6d3ef2ffea9..9847c4092df 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py +++ b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py @@ -813,9 +813,7 @@ async def _estimate(mock_router: MagicMock | None, model: str = AN_ALIAS, **over request = CostEstimateRequest( model=model, - input_tokens=INPUT_TOKENS, - output_tokens=OUTPUT_TOKENS, - **overrides, + **{"input_tokens": INPUT_TOKENS, "output_tokens": OUTPUT_TOKENS, **overrides}, ) with patch( # test-quality-ok: proxy_server module global is the endpoint's only injection point "litellm.proxy.proxy_server.llm_router", mock_router @@ -1062,6 +1060,54 @@ class TestEstimateCostCacheAndReasoningTokens: assert response.cache_read_input_token_cost == pytest.approx(5e-7) assert response.cache_creation_input_token_cost == pytest.approx(6.25e-6) + @pytest.mark.asyncio + async def test_a_tiered_model_reports_the_rates_its_lines_were_billed_at(self, monkeypatch): + """Above a token tier the calculator bills every line at the tier's rate, so the reported + rates must be the tier's too: each line equals its token count times the rate next to it.""" + monkeypatch.setitem( + litellm.model_cost, + A_MAPPED_MODEL, + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "cache_read_input_token_cost": 3e-7, + "cache_creation_input_token_cost": 3.75e-6, + "input_cost_per_token_above_200k_tokens": 6e-6, + "output_cost_per_token_above_200k_tokens": 3e-5, + "cache_read_input_token_cost_above_200k_tokens": 6e-7, + "cache_creation_input_token_cost_above_200k_tokens": 7.5e-6, + "litellm_provider": "openai", + "mode": "chat", + }, + ) + + response = await _estimate( + None, + model=A_MAPPED_MODEL, + input_tokens=250_000, + cache_read_input_tokens=200_000, + cache_creation_input_tokens=10_000, + output_tokens=1_000, + reasoning_tokens=200, + ) + + assert response.input_cost_per_token == pytest.approx(6e-6) + assert response.output_cost_per_token == pytest.approx(3e-5) + assert response.cache_read_input_token_cost == pytest.approx(6e-7) + assert response.cache_creation_input_token_cost == pytest.approx(7.5e-6) + assert response.output_cost_per_reasoning_token == pytest.approx(3e-5) + assert response.cache_read_cost_per_request == pytest.approx(200_000 * response.cache_read_input_token_cost) + assert response.cache_creation_cost_per_request == pytest.approx( + 10_000 * response.cache_creation_input_token_cost + ) + assert response.reasoning_cost_per_request == pytest.approx(200 * response.output_cost_per_reasoning_token) + assert response.input_cost_per_request == pytest.approx( + 40_000 * response.input_cost_per_token + + response.cache_read_cost_per_request + + response.cache_creation_cost_per_request + ) + assert response.output_cost_per_request == pytest.approx(1_000 * response.output_cost_per_token) + class TestCostEstimateRequestTokenSubsets: def test_cache_tokens_beyond_the_input_tokens_are_rejected(self): diff --git a/ui/litellm-dashboard/src/lib/http/schema.d.ts b/ui/litellm-dashboard/src/lib/http/schema.d.ts index 6a9c86cc1b5..46cb5ad892a 100644 --- a/ui/litellm-dashboard/src/lib/http/schema.d.ts +++ b/ui/litellm-dashboard/src/lib/http/schema.d.ts @@ -26523,7 +26523,10 @@ export interface components { * @description Input token cost per request (before margin) */ input_cost_per_request: number; - /** Input Cost Per Token */ + /** + * Input Cost Per Token + * @description Rate billed per input token + */ input_cost_per_token?: number | null; /** Input Tokens */ input_tokens: number; @@ -26584,7 +26587,10 @@ export interface components { * @description Output token cost per request (before margin) */ output_cost_per_request: number; - /** Output Cost Per Token */ + /** + * Output Cost Per Token + * @description Rate billed per output token + */ output_cost_per_token?: number | null; /** Output Tokens */ output_tokens: number; From ba91588b15d308daf229f464b9aa89a0f7480afd Mon Sep 17 00:00:00 2001 From: ryan-crabbe-berri Date: Mon, 7 Sep 2026 18:38:12 -0700 Subject: [PATCH 04/26] fix(proxy): price one cost estimate at one moment The totals, the per-token-type lines and the reported rates each resolved off-peak pricing on their own clock read, so a quote taken as a window opened could bill on one side of the boundary and report rates from the other. /cost/estimate now pins a billing moment for the whole quote, and every rate lookup answers for the pinned moment instead of reading the clock again Claude-Session: https://claude.ai/code/session_011Tn3657NkV6ojLqewL64Kb --- litellm/_internal_context.py | 24 +++++++++++ .../litellm_core_utils/llm_cost_calc/utils.py | 9 ++-- .../cost_tracking_settings.py | 43 +++++++++++-------- .../llm_cost_calc/test_llm_cost_calc_utils.py | 42 ++++++++++++++++++ .../test_cost_tracking_settings.py | 30 +++++++++++++ 5 files changed, 126 insertions(+), 22 deletions(-) diff --git a/litellm/_internal_context.py b/litellm/_internal_context.py index f856fe0f2b3..8132008731f 100644 --- a/litellm/_internal_context.py +++ b/litellm/_internal_context.py @@ -6,9 +6,33 @@ be settable from user input. Context variables are scoped to the current asyncio task and cannot be injected via HTTP request bodies. """ +from collections.abc import Generator +from contextlib import contextmanager from contextvars import ContextVar +from datetime import datetime, timezone from typing import Final # When True, suppresses async logging and billing for internal sub-calls # (e.g., emulated file-search steps that make nested LLM calls). is_internal_call: Final[ContextVar[bool]] = ContextVar("is_internal_call", default=False) + +# One request prices its totals, its per-token-type lines and the rates it reports on +# separate code paths. Each reads the clock for off-peak pricing, so without a pinned +# moment they can land on either side of a window boundary and disagree with each other. +_billing_time: Final[ContextVar[datetime | None]] = ContextVar("billing_time", default=None) + + +@contextmanager +def pinned_billing_time(moment: datetime) -> Generator[None]: + """Price every rate lookup inside this block at ``moment`` rather than at each one's own clock read.""" + token: Final = _billing_time.set(moment) + try: + yield + finally: + _billing_time.reset(token) + + +def current_billing_time() -> datetime: + """The pinned billing moment, or now in UTC outside a pinned block.""" + pinned: Final = _billing_time.get() + return pinned if pinned is not None else datetime.now(timezone.utc) diff --git a/litellm/litellm_core_utils/llm_cost_calc/utils.py b/litellm/litellm_core_utils/llm_cost_calc/utils.py index 2bb15fb4c48..e2168528e6b 100644 --- a/litellm/litellm_core_utils/llm_cost_calc/utils.py +++ b/litellm/litellm_core_utils/llm_cost_calc/utils.py @@ -10,6 +10,7 @@ from typing import Any, Final, Literal, TypedDict, cast from zoneinfo import ZoneInfo, ZoneInfoNotFoundError import litellm +from litellm._internal_context import current_billing_time from litellm._logging import verbose_logger from litellm.litellm_core_utils.llm_cost_calc.tiered_pricing import ( select_tier_for_input, @@ -306,7 +307,7 @@ def _is_within_off_peak_window(off_peak_hours_utc: str | Sequence[str], current_ than being localised, so callers must pass datetime.now(timezone.utc), never datetime.now(), or every window shifts by the host's offset. """ - reference: Final = current_time if current_time is not None else datetime.now(timezone.utc) + reference: Final = current_time if current_time is not None else current_billing_time() now: Final = (reference.astimezone(timezone.utc) if reference.tzinfo is not None else reference).time() windows: Final = (off_peak_hours_utc,) if isinstance(off_peak_hours_utc, str) else off_peak_hours_utc for window in windows: @@ -393,7 +394,7 @@ def _is_off_peak(off_peak: Mapping[str, object], current_time: datetime | None = rules: the flat hours_utc windows, which apply every day, or any entry in windows, whose hours apply only on its weekdays. """ - reference: Final = current_time if current_time is not None else datetime.now(timezone.utc) + reference: Final = current_time if current_time is not None else current_billing_time() reference_utc: Final = ( reference.astimezone(timezone.utc) if reference.tzinfo is not None else reference.replace(tzinfo=timezone.utc) ) @@ -1187,7 +1188,7 @@ def generic_cost_per_token( usage.prompt_tokens - cache_hit - audio_tokens - cache_creation - image_tokens - video_tokens, 0 ) - billing_time: Final = current_time if current_time is not None else datetime.now(timezone.utc) + billing_time: Final = current_time if current_time is not None else current_billing_time() ( prompt_base_cost, completion_base_cost, @@ -1379,7 +1380,7 @@ def _cost_map_billed_rates( vertex_location: str | None, current_time: datetime | None, ) -> BilledTokenRates: - billing_time: Final = current_time if current_time is not None else datetime.now(timezone.utc) + billing_time: Final = current_time if current_time is not None else current_billing_time() ( prompt_base_cost, completion_base_cost, diff --git a/litellm/proxy/management_endpoints/cost_tracking_settings.py b/litellm/proxy/management_endpoints/cost_tracking_settings.py index 17d82fd17e3..1faa66584d5 100644 --- a/litellm/proxy/management_endpoints/cost_tracking_settings.py +++ b/litellm/proxy/management_endpoints/cost_tracking_settings.py @@ -18,6 +18,7 @@ from fastapi import APIRouter, Depends, HTTPException from pydantic import BaseModel import litellm +from litellm._internal_context import current_billing_time, pinned_billing_time from litellm._logging import verbose_proxy_logger from litellm.cost_calculator import completion_cost from litellm.litellm_core_utils.llm_cost_calc.utils import get_billed_token_rates @@ -631,33 +632,39 @@ async def estimate_cost( function_id="cost-estimate", ) - # Use completion_cost which handles all the logic including margins/discounts - try: - cost_per_request: Final = completion_cost( - completion_response=mock_response, + # The totals, the per-token-type lines and the reported rates each resolve pricing on their + # own path. Pinning one moment keeps an off-peak window that opens mid-quote from splitting them. + billed_at: Final = current_billing_time() + with pinned_billing_time(billed_at): + # Use completion_cost which handles all the logic including margins/discounts + try: + cost_per_request: Final = completion_cost( + completion_response=mock_response, + model=resolved_model, + custom_llm_provider=resolved_provider, + custom_cost_per_token=resolved.custom_cost_per_token, + litellm_logging_obj=litellm_logging_obj, + ) + except Exception as e: + raise HTTPException( + status_code=404, + detail={ + "error": f"Could not calculate cost for model '{request.model}' (resolved to '{resolved_model}'): {e}" + }, + ) + + rates: Final = get_billed_token_rates( model=resolved_model, custom_llm_provider=resolved_provider, + usage=usage, custom_cost_per_token=resolved.custom_cost_per_token, - litellm_logging_obj=litellm_logging_obj, - ) - except Exception as e: - raise HTTPException( - status_code=404, - detail={ - "error": f"Could not calculate cost for model '{request.model}' (resolved to '{resolved_model}'): {e}" - }, + current_time=billed_at, ) per_request: Final = _cost_lines(cost_per_request, litellm_logging_obj.cost_breakdown) daily: Final = per_request.times(request.num_requests_per_day) monthly: Final = per_request.times(request.num_requests_per_month) - rates: Final = get_billed_token_rates( - model=resolved_model, - custom_llm_provider=resolved_provider, - usage=usage, - custom_cost_per_token=resolved.custom_cost_per_token, - ) model_info: Final = _lookup_model_info(resolved_model) mapped_provider: Final = model_info.get("litellm_provider") if model_info is not None else None custom_llm_provider: Final = mapped_provider if mapped_provider is not None else resolved_provider diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py index 4de3c059e63..90178428018 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py @@ -1,9 +1,11 @@ import json +from datetime import datetime, timezone import pytest from fastapi.testclient import TestClient import litellm +from litellm._internal_context import pinned_billing_time from litellm.litellm_core_utils.llm_cost_calc.tool_call_cost_tracking import ( StandardBuiltInToolCostTracking, ) @@ -3981,6 +3983,46 @@ def test_billed_token_rates_follow_the_token_tier_the_breakdown_bills_at(monkeyp assert breakdown.reasoning_cost == pytest.approx(200 * rates.output_cost_per_reasoning_token) +def test_a_pinned_billing_time_prices_the_totals_and_the_reported_rates_at_one_moment(monkeypatch): + """Totals and reported rates resolve off-peak pricing on separate paths that each read the + clock, so a window opening between the two reads used to leave them describing one request + at two different prices. Pinned, both must answer for the pinned moment.""" + monkeypatch.setitem( + litellm.model_cost, + "off-peak-model", + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "off_peak_pricing": { + "hours_utc": "02:00-03:00", + "input_cost_per_token": 1e-6, + "output_cost_per_token": 5e-6, + }, + "litellm_provider": "openai", + "mode": "chat", + }, + ) + usage = Usage(prompt_tokens=1000, completion_tokens=500, total_tokens=1500) + + with pinned_billing_time(datetime(2026, 1, 1, 2, 30, tzinfo=timezone.utc)): + off_peak_prompt_cost, off_peak_completion_cost = generic_cost_per_token( + model="off-peak-model", usage=usage, custom_llm_provider="openai" + ) + off_peak_rates = get_billed_token_rates(model="off-peak-model", custom_llm_provider="openai", usage=usage) + with pinned_billing_time(datetime(2026, 1, 1, 12, 30, tzinfo=timezone.utc)): + peak_prompt_cost, peak_completion_cost = generic_cost_per_token( + model="off-peak-model", usage=usage, custom_llm_provider="openai" + ) + peak_rates = get_billed_token_rates(model="off-peak-model", custom_llm_provider="openai", usage=usage) + + assert off_peak_rates.input_cost_per_token == pytest.approx(1e-6) + assert peak_rates.input_cost_per_token == pytest.approx(3e-6) + assert off_peak_prompt_cost == pytest.approx(1000 * off_peak_rates.input_cost_per_token) + assert off_peak_completion_cost == pytest.approx(500 * off_peak_rates.output_cost_per_token) + assert peak_prompt_cost == pytest.approx(1000 * peak_rates.input_cost_per_token) + assert peak_completion_cost == pytest.approx(500 * peak_rates.output_cost_per_token) + + def test_billed_token_rates_are_none_for_an_unpriced_model(): usage = Usage(prompt_tokens=10, completion_tokens=5, total_tokens=15) diff --git a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py index 9847c4092df..e9485f3a044 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py +++ b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py @@ -4,6 +4,7 @@ Tests for cost tracking settings management endpoints. Tests the GET and PATCH endpoints for managing cost discount configuration. """ +from datetime import datetime, timezone from unittest.mock import AsyncMock, MagicMock, patch import pytest @@ -12,6 +13,7 @@ from pydantic import ValidationError import litellm +from litellm._internal_context import pinned_billing_time from litellm.proxy._types import CostEstimateRequest from litellm.proxy.management_endpoints.cost_tracking_settings import router from litellm.proxy.proxy_server import app @@ -1108,6 +1110,34 @@ class TestEstimateCostCacheAndReasoningTokens: ) assert response.output_cost_per_request == pytest.approx(1_000 * response.output_cost_per_token) + @pytest.mark.asyncio + async def test_a_quote_prices_its_totals_and_its_rates_at_the_same_moment(self, monkeypatch): + """The totals and the reported rates resolve off-peak pricing on separate paths. A quote + taken as a window opens must not bill on one side of it and report rates from the other.""" + monkeypatch.setitem( + litellm.model_cost, + A_MAPPED_MODEL, + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "off_peak_pricing": { + "hours_utc": "02:00-03:00", + "input_cost_per_token": 1e-6, + "output_cost_per_token": 5e-6, + }, + "litellm_provider": "openai", + "mode": "chat", + }, + ) + + with pinned_billing_time(datetime(2026, 1, 1, 2, 30, tzinfo=timezone.utc)): + response = await _estimate(None, model=A_MAPPED_MODEL) + + assert response.input_cost_per_token == pytest.approx(1e-6) + assert response.output_cost_per_token == pytest.approx(5e-6) + assert response.input_cost_per_request == pytest.approx(INPUT_TOKENS * response.input_cost_per_token) + assert response.output_cost_per_request == pytest.approx(OUTPUT_TOKENS * response.output_cost_per_token) + class TestCostEstimateRequestTokenSubsets: def test_cache_tokens_beyond_the_input_tokens_are_rejected(self): From d72eb4491ab2da2a56944c720a3200280cdd040d Mon Sep 17 00:00:00 2001 From: mubashir1osmani Date: Tue, 8 Sep 2026 17:22:30 -0400 Subject: [PATCH 05/26] fix(ui): repair pass-through delete confirm dialog and disable delete for config endpoints --- .../PassThroughEndpointsTable.test.tsx | 47 ++++++++++++++ .../PassThroughEndpointsTableColumns.tsx | 30 +++++++-- .../PassThroughSettings.tsx | 62 ++++++++----------- 3 files changed, 98 insertions(+), 41 deletions(-) diff --git a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx index 040fa463f9c..7a3fc621fc1 100644 --- a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx +++ b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx @@ -89,6 +89,53 @@ describe("PassThroughEndpointsTable", () => { expect(onDeleteClick).toHaveBeenCalledWith("ep-1"); }); + it("should disable edit and delete for config-defined endpoints", async () => { + const user = userEvent.setup(); + const onEndpointClick = vi.fn(); + const onDeleteClick = vi.fn(); + const configEndpoint: passThroughItem = { + id: "ep-config", + path: "/from-config", + target: "https://config.example.com", + headers: {}, + is_from_config: true, + }; + render( + , + ); + + await user.click(screen.getByTestId("endpoint-actions-ep-config")); + const editItem = await screen.findByTestId("endpoint-action-edit"); + const deleteItem = await screen.findByTestId("endpoint-action-delete"); + + expect(editItem).toHaveAttribute("data-disabled"); + expect(deleteItem).toHaveAttribute("data-disabled"); + + await user.click(editItem); + await user.click(deleteItem); + + expect(onEndpointClick).not.toHaveBeenCalled(); + expect(onDeleteClick).not.toHaveBeenCalled(); + }); + + it("should label endpoint source as Config or DB", () => { + const configEndpoint: passThroughItem = { + id: "ep-config", + path: "/from-config", + target: "https://config.example.com", + headers: {}, + is_from_config: true, + }; + render(); + expect(screen.getByText("Config")).toBeInTheDocument(); + expect(screen.getAllByText("DB")).toHaveLength(2); + }); + it("should disable edit and delete for endpoints without an id", async () => { const user = userEvent.setup(); const onEndpointClick = vi.fn(); diff --git a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx index d22b274861a..d63be628110 100644 --- a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx +++ b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx @@ -18,6 +18,10 @@ import { cn } from "@/lib/cva.config"; import type { passThroughItem } from "./PassThroughSettings"; +const CONFIG_EDIT_HINT = "Config pass-through endpoints cannot be edited on the dashboard. Please edit the config file."; +const CONFIG_DELETE_HINT = + "Config pass-through endpoints cannot be deleted on the dashboard. Please edit the config file."; + function HeaderWithTooltip({ title, tooltip }: { title: string; tooltip: string }) { return (
@@ -73,6 +77,7 @@ interface EndpointRowActionsProps { function EndpointRowActions({ endpoint, onEndpointClick, onDeleteClick }: EndpointRowActionsProps) { const endpointId = endpoint.id; + const isFromConfig = endpoint.is_from_config ?? false; return ( endpointId && onEndpointClick(endpointId)} + disabled={isFromConfig || !endpointId} + title={isFromConfig ? CONFIG_EDIT_HINT : undefined} + onClick={() => !isFromConfig && endpointId && onEndpointClick(endpointId)} > Edit @@ -95,8 +101,9 @@ function EndpointRowActions({ endpoint, onEndpointClick, onDeleteClick }: Endpoi endpointId && onDeleteClick(endpointId)} + disabled={isFromConfig || !endpointId} + title={isFromConfig ? CONFIG_DELETE_HINT : undefined} + onClick={() => !isFromConfig && endpointId && onDeleteClick(endpointId)} > Delete @@ -124,7 +131,9 @@ export const getPassThroughEndpointsTableColumns = ({ enableSorting: false, cell: ({ row }) => { const endpointId = row.original.id; - if (!endpointId) return —; + if (!endpointId || row.original.is_from_config) { + return —; + } return ( { + const isFromConfig = row.original.is_from_config ?? false; + return ; + }, + }, { id: "path", accessorKey: "path", diff --git a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughSettings.tsx b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughSettings.tsx index 2dc6fdbd32c..8ef2766d412 100644 --- a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughSettings.tsx +++ b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughSettings.tsx @@ -1,4 +1,13 @@ import React, { useState, useEffect } from "react"; +import { + AlertDialog, + AlertDialogCancel, + AlertDialogContent, + AlertDialogDescription, + AlertDialogFooter, + AlertDialogHeader, + AlertDialogTitle, +} from "@/components/ui/alert-dialog"; import { Button } from "@/components/ui/button"; import { deletePassThroughEndpointsCall, getPassThroughEndpointsCall } from "../networking"; import AddPassThroughEndpoint from "../add_pass_through"; @@ -25,6 +34,7 @@ export interface passThroughItem { methods?: string[]; guardrails?: Record; default_query_params?: Record; + is_from_config?: boolean; } const PassThroughSettings: React.FC = ({ accessToken, userRole, userID, premiumUser }) => { @@ -133,42 +143,22 @@ const PassThroughSettings: React.FC = ({ accessToken, onDeleteClick={handleDelete} /> - {isDeleteModalOpen && ( -
-
- - - - -
-
-
-
-

Delete Pass-Through Endpoint

-
-

- Are you sure you want to delete this pass-through endpoint? This action cannot be undone. -

-
-
-
-
-
- - -
-
-
-
- )} + !open && cancelDelete()}> + + + Delete Pass-Through Endpoint + + Are you sure you want to delete this pass-through endpoint? This action cannot be undone. + + + + Cancel + + + +
); }; From 093b473710313f22efd864e5262e5a376a0ee2ce Mon Sep 17 00:00:00 2001 From: mubashir1osmani Date: Tue, 8 Sep 2026 17:43:23 -0400 Subject: [PATCH 06/26] fix(ui): surface the config-endpoint hint as visible menu text --- .../PassThroughEndpointsTable.test.tsx | 12 ++++++++++++ .../PassThroughEndpointsTableColumns.tsx | 12 +++++++----- 2 files changed, 19 insertions(+), 5 deletions(-) diff --git a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx index 7a3fc621fc1..af3f98b746b 100644 --- a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx +++ b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTable.test.tsx @@ -115,6 +115,9 @@ describe("PassThroughEndpointsTable", () => { expect(editItem).toHaveAttribute("data-disabled"); expect(deleteItem).toHaveAttribute("data-disabled"); + expect(screen.getByTestId("endpoint-config-hint")).toHaveTextContent( + "This endpoint is defined in the config file and cannot be edited or deleted on the dashboard.", + ); await user.click(editItem); await user.click(deleteItem); @@ -123,6 +126,15 @@ describe("PassThroughEndpointsTable", () => { expect(onDeleteClick).not.toHaveBeenCalled(); }); + it("should not show the config hint for DB endpoints", async () => { + const user = userEvent.setup(); + render(); + + await user.click(screen.getByTestId("endpoint-actions-ep-1")); + await screen.findByTestId("endpoint-action-delete"); + expect(screen.queryByTestId("endpoint-config-hint")).not.toBeInTheDocument(); + }); + it("should label endpoint source as Config or DB", () => { const configEndpoint: passThroughItem = { id: "ep-config", diff --git a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx index d63be628110..5b18685a140 100644 --- a/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx +++ b/ui/litellm-dashboard/src/components/PassThroughSettings/PassThroughEndpointsTableColumns.tsx @@ -18,9 +18,8 @@ import { cn } from "@/lib/cva.config"; import type { passThroughItem } from "./PassThroughSettings"; -const CONFIG_EDIT_HINT = "Config pass-through endpoints cannot be edited on the dashboard. Please edit the config file."; -const CONFIG_DELETE_HINT = - "Config pass-through endpoints cannot be deleted on the dashboard. Please edit the config file."; +const CONFIG_ENDPOINT_HINT = + "This endpoint is defined in the config file and cannot be edited or deleted on the dashboard."; function HeaderWithTooltip({ title, tooltip }: { title: string; tooltip: string }) { return ( @@ -91,7 +90,6 @@ function EndpointRowActions({ endpoint, onEndpointClick, onDeleteClick }: Endpoi !isFromConfig && endpointId && onEndpointClick(endpointId)} > @@ -102,12 +100,16 @@ function EndpointRowActions({ endpoint, onEndpointClick, onDeleteClick }: Endpoi variant="destructive" data-testid="endpoint-action-delete" disabled={isFromConfig || !endpointId} - title={isFromConfig ? CONFIG_DELETE_HINT : undefined} onClick={() => !isFromConfig && endpointId && onDeleteClick(endpointId)} > Delete + {isFromConfig && ( +
+ {CONFIG_ENDPOINT_HINT} +
+ )} ); From 0086b62b4575fb5bb69653c2ad3d196822bd3f13 Mon Sep 17 00:00:00 2001 From: kerry Date: Wed, 9 Sep 2026 02:05:08 +0000 Subject: [PATCH 07/26] fix(cost-map): keep first fetch blocking, run retries in background Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/__init__.py | 4 +- .../litellm_core_utils/get_model_cost_map.py | 132 +++++++++- litellm/proxy/proxy_server.py | 23 +- .../test_get_model_cost_map.py | 235 +++++++++++++----- 4 files changed, 296 insertions(+), 98 deletions(-) diff --git a/litellm/__init__.py b/litellm/__init__.py index 42c0ea881fd..4e3754399da 100644 --- a/litellm/__init__.py +++ b/litellm/__init__.py @@ -541,7 +541,7 @@ _key_management_system: Optional["KeyManagementSystem"] = None #### PII MASKING #### output_parse_pii: bool = False ############################################# -from litellm.litellm_core_utils.get_model_cost_map import get_model_cost_map +from litellm.litellm_core_utils.get_model_cost_map import get_model_cost_map, mark_litellm_import_complete model_cost = get_model_cost_map(url=model_cost_map_url) cost_discount_config: Dict[str, float] = {} # Provider-specific cost discounts {"vertex_ai": 0.05} = 5% discount @@ -2397,3 +2397,5 @@ def __getattr__(name: str) -> Any: # ALL_LITELLM_RESPONSE_TYPES is lazy-loaded via __getattr__ to avoid loading utils at import time + +mark_litellm_import_complete() diff --git a/litellm/litellm_core_utils/get_model_cost_map.py b/litellm/litellm_core_utils/get_model_cost_map.py index 9cba5db8ab7..4118cd3420e 100644 --- a/litellm/litellm_core_utils/get_model_cost_map.py +++ b/litellm/litellm_core_utils/get_model_cost_map.py @@ -12,6 +12,7 @@ import asyncio import json import os import random +import threading import time from collections.abc import Awaitable, Callable from dataclasses import dataclass @@ -159,6 +160,15 @@ class GetModelCostMap: RETRYABLE_FETCH_STATUS_CODES: Final = frozenset({429, 500, 502, 503, 504}) MODEL_COST_MAP_FETCH_MAX_ATTEMPTS: Final = 3 MODEL_COST_MAP_FETCH_MAX_WAIT_SECONDS: Final = 30.0 +_litellm_import_complete = threading.Event() + + +def mark_litellm_import_complete() -> None: + _litellm_import_complete.set() + + +def _start_daemon_thread(fn: Callable[[], None]) -> None: + threading.Thread(target=fn, name="litellm-model-cost-map-retry", daemon=True).start() @dataclass(frozen=True, slots=True) @@ -297,9 +307,15 @@ def _fetch_remote_model_cost_map_with_retry_sync( sleep: Callable[[float], None], rng: random.Random, client: _SyncGetClient, + starting_attempt: int = 1, + initial_outcome: _FetchAttemptRetryable | None = None, ) -> ModelCostMapReloadResult: - for attempt in range(1, max_attempts + 1): - outcome = _attempt_fetch_sync(client=client, url=url, timeout=timeout) + for attempt in range(starting_attempt, max_attempts + 1): + outcome = ( + initial_outcome + if initial_outcome is not None and attempt == starting_attempt + else _attempt_fetch_sync(client=client, url=url, timeout=timeout) + ) if not isinstance(outcome, _FetchAttemptRetryable): return outcome wait_seconds = _next_retry_wait(outcome=outcome, attempt=attempt, max_attempts=max_attempts, rng=rng) @@ -464,6 +480,70 @@ def _finalize_model_cost_map(model_cost: dict) -> dict: return _expand_model_aliases(model_cost) +def adopt_model_cost_map( + new_model_cost_map: dict, # mutable-ok: public API preserves the mutable cost-map contract +) -> int: + import litellm + from litellm import utils + + litellm.model_cost = new_model_cost_map + utils._invalidate_model_cost_lowercase_map() # pyright: ignore[reportPrivateUsage] # required cache invalidation + litellm.add_known_models(model_cost_map=new_model_cost_map) + fetched_model_count: Final = len(new_model_cost_map) if new_model_cost_map else 0 + utils.reapply_runtime_model_cost_registrations() + return fetched_model_count + + +def _continue_remote_fetch_in_background( + url: str, + timeout: int, + max_attempts: int, + sleep: Callable[[float], None], + rng: random.Random, + client: _SyncGetClient, + first_outcome: _FetchAttemptRetryable, + apply: Callable[[dict], object], # mutable-ok: injected callback receives the mutable cost-map dict +) -> None: + try: + result: Final = _fetch_remote_model_cost_map_with_retry_sync( + url=url, + timeout=timeout, + max_attempts=max_attempts, + sleep=sleep, + rng=rng, + client=client, + initial_outcome=first_outcome, + ) + if isinstance(result, ModelCostMapReloadUnavailable): + verbose_logger.warning( + "LiteLLM: Failed to fetch remote model cost map from %s after %d attempts; keeping local backup", + url, + max_attempts, + ) + return + backup_model_count: Final = GetModelCostMap._get_backup_model_count() # pyright: ignore[reportPrivateUsage] # integrity cache + if not GetModelCostMap.validate_model_cost_map( + fetched_map=result.model_cost_map, + backup_model_count=backup_model_count, + ): + verbose_logger.warning( + "LiteLLM: Fetched model cost map failed integrity check. Keeping local backup. url=%s", + url, + ) + _cost_map_source_info.fallback_reason = "Remote data failed integrity validation" + return + finalized_map: Final = _finalize_model_cost_map(result.model_cost_map) + _litellm_import_complete.wait() + apply(finalized_map) + _cost_map_source_info.source = "remote" + _cost_map_source_info.url = url + _cost_map_source_info.is_env_forced = False + _cost_map_source_info.fallback_reason = None + _cost_map_source_info.loaded_at = datetime.now(timezone.utc) + except Exception as e: + verbose_logger.warning("LiteLLM: Background model cost map retry failed: %s", e) + + def get_model_cost_map( url: str, timeout: int = 5, @@ -471,14 +551,19 @@ def get_model_cost_map( sleep: Callable[[float], None] = time.sleep, rng: random.Random | None = None, client: "_SyncGetClient | None" = None, + start_background: Callable[[Callable[[], None]], None] = _start_daemon_thread, + apply: Callable[ # mutable-ok: injected callback receives the mutable cost-map dict + [dict], + object, + ] = adopt_model_cost_map, ) -> dict: """ Public entry point — returns the model cost map dict. 1. If ``LITELLM_LOCAL_MODEL_COST_MAP`` is set, uses the local backup only. - 2. Otherwise fetches from ``url``, retrying transient HTTP errors - (429/5xx/transport) with Retry-After-aware backoff, validates - integrity, and falls back to the local backup on any failure. + 2. Otherwise fetches from ``url``, validates the first response, and falls + back to the local backup while retrying transient HTTP errors in the + background. Only the backup model count is cached (a single int) for validation. The full backup dict is only parsed when it must be *returned* as a @@ -497,14 +582,35 @@ def get_model_cost_map( _cost_map_source_info.url = url _cost_map_source_info.is_env_forced = False - result: Final = _fetch_remote_model_cost_map_with_retry_sync( - url=url, - timeout=timeout, - max_attempts=max_attempts, - sleep=sleep, - rng=rng if rng is not None else random.Random(), - client=client if client is not None else httpx, - ) + fetch_client: Final = client if client is not None else httpx + fetch_rng: Final = rng if rng is not None else random.Random() + first_outcome: Final = _attempt_fetch_sync(client=fetch_client, url=url, timeout=timeout) + if isinstance(first_outcome, _FetchAttemptRetryable): + verbose_logger.warning( + "LiteLLM: model cost map fetch attempt 1/%d failed: %s; " + "using local backup while retrying in the background", + max_attempts, + first_outcome.reason, + ) + _cost_map_source_info.source = "local" + _cost_map_source_info.fallback_reason = f"Remote fetch failed: {first_outcome.reason}" + local_map: Final = _finalize_model_cost_map(GetModelCostMap.load_local_model_cost_map()) + if max_attempts > 1: + start_background( + lambda: _continue_remote_fetch_in_background( + url=url, + timeout=timeout, + max_attempts=max_attempts, + sleep=sleep, + rng=fetch_rng, + client=fetch_client, + first_outcome=first_outcome, + apply=apply, + ) + ) + return local_map + + result: Final = first_outcome if isinstance(result, ModelCostMapReloadUnavailable): verbose_logger.warning( "LiteLLM: Failed to fetch remote model cost map from %s: %s. Falling back to local backup.", diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index 1cd08fe27c0..efe38376353 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -132,11 +132,7 @@ from litellm.types.utils import ( TextCompletionResponse, TokenCountResponse, ) -from litellm.utils import ( - _invalidate_model_cost_lowercase_map, - load_credentials_from_list, - reapply_runtime_model_cost_registrations, -) +from litellm.utils import load_credentials_from_list if TYPE_CHECKING: from aiohttp import ClientSession @@ -4411,20 +4407,9 @@ def resolve_classifier_plugin( def _swap_in_model_cost_map(new_model_cost_map: dict) -> int: - """Adopt a freshly fetched cost map into this process's litellm state, return the model count""" - litellm.model_cost = new_model_cost_map - # Invalidate case-insensitive lookup map since model_cost was replaced - _invalidate_model_cost_lowercase_map() - # Repopulate provider model sets (e.g. litellm.anthropic_models) so that - # wildcard patterns like "anthropic/*" include any newly added models. - litellm.add_known_models(model_cost_map=new_model_cost_map) - # Counted before the re-apply below, which writes into this same dict, so the - # number reported describes the fetched price data alone. - fetched_model_count: Final = len(new_model_cost_map) if new_model_cost_map else 0 - # The swap discards everything registered at runtime (deployment model_info, - # register_model overrides), so put it back on top of the fresh catalog. - reapply_runtime_model_cost_registrations() - return fetched_model_count + from litellm.litellm_core_utils.get_model_cost_map import adopt_model_cost_map + + return adopt_model_cost_map(new_model_cost_map) def should_load_db_object(object_type: str | SupportedDBObjectType) -> bool: diff --git a/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py b/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py index 18185126775..dce1a431f13 100644 --- a/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py +++ b/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py @@ -20,21 +20,18 @@ from litellm.litellm_core_utils.get_model_cost_map import ( GetModelCostMap, _count_model_entries, _finalize_model_cost_map, + adopt_model_cost_map, ) def _load_root_cost_map() -> dict: - path = os.path.join( - os.path.dirname(__file__), "../../../model_prices_and_context_window.json" - ) + path = os.path.join(os.path.dirname(__file__), "../../../model_prices_and_context_window.json") with open(path) as f: return json.load(f) def _make_models(n: int) -> dict: - return { - f"model-{i}": {"litellm_provider": "openai", "mode": "chat"} for i in range(n) - } + return {f"model-{i}": {"litellm_provider": "openai", "mode": "chat"} for i in range(n)} def test_count_model_entries_excludes_reserved_keys(): @@ -117,9 +114,7 @@ def test_finalize_pops_key_and_installs_rules(): def test_finalize_with_no_block_clears_rules(): previous = list(get_fallback_generalization_rules()) try: - set_fallback_generalizations( - [{"name": "stale", "pattern": r"^x", "model_info": {"a": 1}}] - ) + set_fallback_generalizations([{"name": "stale", "pattern": r"^x", "model_info": {"a": 1}}]) _finalize_model_cost_map(_make_models(2)) assert match_capability_generalizations("x-1") is None finally: @@ -306,9 +301,7 @@ def test_get_model_cost_map_stamps_loaded_at(monkeypatch): from litellm.litellm_core_utils import get_model_cost_map as module monkeypatch.setattr(module._cost_map_source_info, "loaded_at", None) - client, _calls = _mock_client( - [httpx.Response(200, content=_real_map_bytes())], client_cls=httpx.Client - ) + client, _calls = _mock_client([httpx.Response(200, content=_real_map_bytes())], client_cls=httpx.Client) before = datetime.now(timezone.utc) module.get_model_cost_map(url="https://example.invalid/cost_map.json", client=client) @@ -317,6 +310,7 @@ def test_get_model_cost_map_stamps_loaded_at(monkeypatch): assert loaded_at is not None assert before <= loaded_at <= datetime.now(timezone.utc) + # --------------------------------------------------------------------------- # refetch_model_cost_map: retry/backoff behavior for runtime reloads # --------------------------------------------------------------------------- @@ -382,9 +376,7 @@ async def test_refetch_retries_429_honoring_retry_after(): ] ) sleeper = _SleepRecorder() - result = await refetch_model_cost_map( - url=_URL, sleep=sleeper, rng=random.Random(0), client=client - ) + result = await refetch_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) assert isinstance(result, ModelCostMapReloaded) assert len(result.model_cost_map) > 100 assert calls["count"] == 3 @@ -396,9 +388,7 @@ async def test_refetch_gives_up_after_max_attempts_with_exponential_backoff(): """All 429 without Retry-After: exponential backoff waits, then a failure value.""" client, calls = _mock_client([httpx.Response(429)]) sleeper = _SleepRecorder() - result = await refetch_model_cost_map( - url=_URL, sleep=sleeper, rng=random.Random(0), client=client - ) + result = await refetch_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) assert isinstance(result, ModelCostMapReloadUnavailable) assert "429" in result.reason assert "after 3 attempts" in result.reason @@ -418,9 +408,7 @@ async def test_refetch_caps_retry_after_wait(): ] ) sleeper = _SleepRecorder() - result = await refetch_model_cost_map( - url=_URL, sleep=sleeper, rng=random.Random(0), client=client - ) + result = await refetch_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) assert isinstance(result, ModelCostMapReloaded) assert sleeper.waits == [30.0] @@ -435,9 +423,7 @@ async def test_refetch_retries_transport_errors(): ] ) sleeper = _SleepRecorder() - result = await refetch_model_cost_map( - url=_URL, sleep=sleeper, rng=random.Random(0), client=client - ) + result = await refetch_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) assert isinstance(result, ModelCostMapReloaded) assert calls["count"] == 2 assert len(sleeper.waits) == 1 @@ -448,9 +434,7 @@ async def test_refetch_non_retryable_status_fails_immediately(): """A 404 is permanent: one attempt, no sleeps, failure value.""" client, calls = _mock_client([httpx.Response(404)]) sleeper = _SleepRecorder() - result = await refetch_model_cost_map( - url=_URL, sleep=sleeper, rng=random.Random(0), client=client - ) + result = await refetch_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) assert isinstance(result, ModelCostMapReloadUnavailable) assert "404" in result.reason assert calls["count"] == 1 @@ -461,9 +445,7 @@ async def test_refetch_non_retryable_status_fails_immediately(): async def test_refetch_invalid_json_fails_immediately(): client, calls = _mock_client([httpx.Response(200, content=b"not json")]) sleeper = _SleepRecorder() - result = await refetch_model_cost_map( - url=_URL, sleep=sleeper, rng=random.Random(0), client=client - ) + result = await refetch_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) assert isinstance(result, ModelCostMapReloadUnavailable) assert "invalid JSON" in result.reason assert calls["count"] == 1 @@ -475,9 +457,7 @@ async def test_refetch_shrunk_map_fails_integrity_not_swapped_in(): """A drastically shrunk upstream file is rejected instead of being adopted.""" tiny = json.dumps(_make_models(60)).encode() client, _calls = _mock_client([httpx.Response(200, content=tiny)]) - result = await refetch_model_cost_map( - url=_URL, sleep=_SleepRecorder(), rng=random.Random(0), client=client - ) + result = await refetch_model_cost_map(url=_URL, sleep=_SleepRecorder(), rng=random.Random(0), client=client) assert isinstance(result, ModelCostMapReloadUnavailable) assert "integrity validation" in result.reason @@ -520,62 +500,187 @@ class _SyncSleepRecorder: self.waits.append(seconds) -def test_boot_load_retries_transient_failures_instead_of_falling_back(): - """A refused connection then a 503 at pod boot used to pin the process to the bundled - backup for its lifetime; both are transient and must be retried before giving up.""" +class _BackgroundRecorder: + def __init__(self): + self.callbacks = [] + + def __call__(self, callback): + self.callbacks.append(callback) + + +def test_boot_load_returns_local_map_and_schedules_transient_retry(): + client, calls = _mock_client( + [ + httpx.ConnectError("connection refused"), + ], + client_cls=httpx.Client, + ) + sleeper = _SyncSleepRecorder() + background = _BackgroundRecorder() + + cost_map = get_model_cost_map( + url=_URL, + sleep=sleeper, + rng=random.Random(0), + client=client, + start_background=background, + ) + assert calls["count"] == 1 + assert sleeper.waits == [] + assert len(background.callbacks) == 1 + assert len(cost_map) > 100 + source = get_model_cost_map_source_info() + assert source["source"] == "local" + assert source["fallback_reason"] is not None + + +def test_background_retry_adopts_valid_remote_map(): client, calls = _mock_client( [ httpx.ConnectError("connection refused"), - httpx.Response(503), httpx.Response(200, content=_real_map_bytes()), ], client_cls=httpx.Client, ) sleeper = _SyncSleepRecorder() + background = _BackgroundRecorder() + applied = [] - cost_map = get_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) + get_model_cost_map( + url=_URL, + sleep=sleeper, + rng=random.Random(0), + client=client, + start_background=background, + apply=applied.append, + ) + assert calls["count"] == 1 + assert sleeper.waits == [] + assert len(background.callbacks) == 1 + + background.callbacks[0]() + + assert calls["count"] == 2 + assert len(sleeper.waits) == 1 + assert 2.0 <= sleeper.waits[0] < 3.0 + assert len(applied) == 1 + assert applied[0].keys() >= _load_root_cost_map().keys() - {"sample_spec", FALLBACK_GENERALIZATIONS_KEY} + source = get_model_cost_map_source_info() + assert source["source"] == "remote" + assert source["fallback_reason"] is None + + +def test_background_retry_keeps_local_map_after_remaining_failures(): + client, calls = _mock_client( + [httpx.ConnectError("connection refused"), httpx.Response(503)], + client_cls=httpx.Client, + ) + sleeper = _SyncSleepRecorder() + background = _BackgroundRecorder() + applied = [] + + get_model_cost_map( + url=_URL, + sleep=sleeper, + rng=random.Random(0), + client=client, + start_background=background, + apply=applied.append, + ) + background.callbacks[0]() assert calls["count"] == 3 assert len(sleeper.waits) == 2 assert 2.0 <= sleeper.waits[0] < 3.0 assert 4.0 <= sleeper.waits[1] < 5.0 + assert applied == [] + assert get_model_cost_map_source_info()["source"] == "local" + + +def test_boot_load_does_not_schedule_non_retryable_failure(): + client, calls = _mock_client([httpx.Response(404)], client_cls=httpx.Client) + sleeper = _SyncSleepRecorder() + background = _BackgroundRecorder() + + get_model_cost_map( + url=_URL, + sleep=sleeper, + rng=random.Random(0), + client=client, + start_background=background, + ) + + assert calls["count"] == 1 + assert sleeper.waits == [] + assert background.callbacks == [] source = get_model_cost_map_source_info() - assert source["source"] == "remote" - assert source["fallback_reason"] is None + assert source["source"] == "local" + assert source["fallback_reason"] is not None + + +def test_boot_load_success_does_not_schedule_background_retry(): + client, calls = _mock_client([httpx.Response(200, content=_real_map_bytes())], client_cls=httpx.Client) + sleeper = _SyncSleepRecorder() + background = _BackgroundRecorder() + + cost_map = get_model_cost_map( + url=_URL, + sleep=sleeper, + rng=random.Random(0), + client=client, + start_background=background, + ) + assert calls["count"] == 1 + assert sleeper.waits == [] + assert background.callbacks == [] + assert get_model_cost_map_source_info()["source"] == "remote" assert cost_map.keys() >= _load_root_cost_map().keys() - {"sample_spec", FALLBACK_GENERALIZATIONS_KEY} -def test_boot_load_honors_retry_after_then_falls_back_after_max_attempts(): - """An outage longer than the retry budget still ends on the bundled backup, and the - recorded fallback reason says how many attempts were spent so operators can tell.""" - client, calls = _mock_client( - [httpx.Response(429, headers={"Retry-After": "7"})], client_cls=httpx.Client +def test_boot_load_with_one_attempt_does_not_schedule_background_retry(): + client, calls = _mock_client([httpx.ConnectError("connection refused")], client_cls=httpx.Client) + sleeper = _SyncSleepRecorder() + background = _BackgroundRecorder() + + get_model_cost_map( + url=_URL, + max_attempts=1, + sleep=sleeper, + rng=random.Random(0), + client=client, + start_background=background, ) - sleeper = _SyncSleepRecorder() - cost_map = get_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) - - assert calls["count"] == 3 - assert sleeper.waits == [7.0, 7.0] - source = get_model_cost_map_source_info() - assert source["source"] == "local" - assert "after 3 attempts" in source["fallback_reason"] - assert len(cost_map) > 100 - - -def test_boot_load_does_not_retry_permanent_failures(): - """A 404 or a malformed URL cannot heal by waiting: one attempt, no sleeps, backup.""" - client, calls = _mock_client([httpx.Response(404)], client_cls=httpx.Client) - sleeper = _SyncSleepRecorder() - - get_model_cost_map(url=_URL, sleep=sleeper, rng=random.Random(0), client=client) assert calls["count"] == 1 assert sleeper.waits == [] + assert background.callbacks == [] assert get_model_cost_map_source_info()["source"] == "local" - get_model_cost_map(url="not a url", sleep=sleeper, rng=random.Random(0)) - assert sleeper.waits == [] - assert get_model_cost_map_source_info()["source"] == "local" + +def test_adopt_model_cost_map_replays_runtime_registration_and_provider_models(): + import litellm + from litellm import utils as litellm_utils + + original_model_cost = litellm.model_cost + original_registry = dict(litellm_utils._runtime_registered_model_cost) + original_anthropic_models = set(litellm.anthropic_models) + try: + litellm.register_model( + model_cost={"custom/deployment-model": {"litellm_provider": "custom", "max_input_tokens": 4321}} + ) + + models_count = adopt_model_cost_map({"anthropic/new-model": {"litellm_provider": "anthropic", "mode": "chat"}}) + + assert models_count == 1 + assert "anthropic/new-model" in litellm.anthropic_models + assert litellm.model_cost["custom/deployment-model"]["max_input_tokens"] == 4321 + finally: + litellm.model_cost = original_model_cost # test-quality-ok: restore the module state changed by adoption + litellm_utils._runtime_registered_model_cost.clear() + litellm_utils._runtime_registered_model_cost.update(original_registry) + litellm.anthropic_models.clear() + litellm.anthropic_models.update(original_anthropic_models) + litellm_utils._invalidate_model_cost_lowercase_map() def test_boot_load_respects_local_env_override(monkeypatch): From aa0a9ab3ea4dff53a22cbc60fbc0195c4ab6098c Mon Sep 17 00:00:00 2001 From: kerry Date: Wed, 9 Sep 2026 02:29:56 +0000 Subject: [PATCH 08/26] refactor(cost-map): drop initial_outcome flag from retry loop Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../litellm_core_utils/get_model_cost_map.py | 50 ++++++++++--------- 1 file changed, 26 insertions(+), 24 deletions(-) diff --git a/litellm/litellm_core_utils/get_model_cost_map.py b/litellm/litellm_core_utils/get_model_cost_map.py index 90ff696f926..d9b5a492539 100644 --- a/litellm/litellm_core_utils/get_model_cost_map.py +++ b/litellm/litellm_core_utils/get_model_cost_map.py @@ -324,19 +324,14 @@ async def _fetch_remote_model_cost_map_with_retry( def _fetch_remote_model_cost_map_with_retry_sync( url: str, timeout: int, - max_attempts: int, + attempts: range, sleep: Callable[[float], None], rng: random.Random, client: _SyncGetClient, - starting_attempt: int = 1, - initial_outcome: _FetchAttemptRetryable | None = None, ) -> ModelCostMapReloadResult: - for attempt in range(starting_attempt, max_attempts + 1): - outcome = ( - initial_outcome - if initial_outcome is not None and attempt == starting_attempt - else _attempt_fetch_sync(client=client, url=url, timeout=timeout) - ) + max_attempts: Final = attempts.stop - 1 + for attempt in attempts: + outcome = _attempt_fetch_sync(client=client, url=url, timeout=timeout) if not isinstance(outcome, _FetchAttemptRetryable): return outcome wait_seconds = _next_retry_wait(outcome=outcome, attempt=attempt, max_attempts=max_attempts, rng=rng) @@ -557,18 +552,18 @@ def _continue_remote_fetch_in_background( sleep: Callable[[float], None], rng: random.Random, client: _SyncGetClient, - first_outcome: _FetchAttemptRetryable, + first_wait: float, apply: Callable[[dict], object], # mutable-ok: injected callback receives the mutable cost-map dict ) -> None: try: + sleep(first_wait) result: Final = _fetch_remote_model_cost_map_with_retry_sync( url=url, timeout=timeout, - max_attempts=max_attempts, + attempts=range(2, max_attempts + 1), sleep=sleep, rng=rng, client=client, - initial_outcome=first_outcome, ) if isinstance(result, ModelCostMapReloadUnavailable): verbose_logger.warning( @@ -652,19 +647,26 @@ def get_model_cost_map( local_map: Final = _finalize_loaded_model_cost_map( GetModelCostMap.load_local_model_cost_map_with_revision() ).model_cost_map - if max_attempts > 1: - start_background( - lambda: _continue_remote_fetch_in_background( - url=url, - timeout=timeout, - max_attempts=max_attempts, - sleep=sleep, - rng=fetch_rng, - client=fetch_client, - first_outcome=first_outcome, - apply=apply, - ) + first_wait: Final = _next_retry_wait( + outcome=first_outcome, + attempt=1, + max_attempts=max_attempts, + rng=fetch_rng, + ) + if isinstance(first_wait, ModelCostMapReloadUnavailable): + return local_map + start_background( + lambda: _continue_remote_fetch_in_background( + url=url, + timeout=timeout, + max_attempts=max_attempts, + sleep=sleep, + rng=fetch_rng, + client=fetch_client, + first_wait=first_wait, + apply=apply, ) + ) return local_map result: Final = first_outcome From 536a85b42967fd0c9c6b49d2df9298232034427a Mon Sep 17 00:00:00 2001 From: kerry Date: Wed, 9 Sep 2026 02:48:17 +0000 Subject: [PATCH 09/26] fix(cost-map): keep register_model url fetch to a single attempt Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/utils.py | 2 +- tests/test_litellm/test_utils.py | 23 +++++++++++++++++++++++ 2 files changed, 24 insertions(+), 1 deletion(-) diff --git a/litellm/utils.py b/litellm/utils.py index 36b48d3b8d8..8df28870544 100644 --- a/litellm/utils.py +++ b/litellm/utils.py @@ -3079,7 +3079,7 @@ def register_model( # Convert stringified numbers to appropriate numeric types loaded_model_cost = model_cost elif isinstance(model_cost, str): - loaded_model_cost = litellm.get_model_cost_map(url=model_cost) + loaded_model_cost = litellm.get_model_cost_map(url=model_cost, max_attempts=1) if persist_across_reloads: _registrations: Final[Mapping[str, Mapping[str, object]]] = loaded_model_cost diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py index fee5e3a2e4c..e90372141bd 100644 --- a/tests/test_litellm/test_utils.py +++ b/tests/test_litellm/test_utils.py @@ -2,10 +2,12 @@ import asyncio import json import logging import os +import threading from datetime import datetime, timedelta, timezone from typing import Final from unittest.mock import AsyncMock, MagicMock, patch +import httpx import pytest import respx from jsonschema import validate @@ -2382,6 +2384,27 @@ def test_register_model_with_scientific_notation(): _invalidate_model_cost_lowercase_map() +@respx.mock +def test_register_model_url_fetch_uses_single_attempt(monkeypatch): + monkeypatch.setattr(litellm, "model_cost", dict(litellm.model_cost)) + before = dict(litellm.model_cost) + threads_before = {thread.name for thread in threading.enumerate()} + route = respx.get("https://example.invalid/custom_pricing.json").mock( + return_value=httpx.Response(503) + ) + + litellm.register_model(model_cost="https://example.invalid/custom_pricing.json") + + threads_after = {thread.name for thread in threading.enumerate()} + assert route.call_count == 1 + assert not (threads_after - threads_before) & {"litellm-model-cost-map-retry"} + assert not any( + thread.name == "litellm-model-cost-map-retry" and thread.is_alive() + for thread in threading.enumerate() + ) + assert litellm.model_cost.keys() >= before.keys() + + def test_register_model_openrouter_without_slash(): """ Test that register_model handles openrouter models without '/' in the name. From 4179f086e0196107dfc83da2a4e0bb1d15568edd Mon Sep 17 00:00:00 2001 From: kerry Date: Wed, 9 Sep 2026 02:58:51 +0000 Subject: [PATCH 10/26] refactor(cost-map): share local-fallback and remote-accept paths Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../litellm_core_utils/get_model_cost_map.py | 111 ++++++++---------- 1 file changed, 47 insertions(+), 64 deletions(-) diff --git a/litellm/litellm_core_utils/get_model_cost_map.py b/litellm/litellm_core_utils/get_model_cost_map.py index d9b5a492539..67d06a0758e 100644 --- a/litellm/litellm_core_utils/get_model_cost_map.py +++ b/litellm/litellm_core_utils/get_model_cost_map.py @@ -531,6 +531,32 @@ def _finalize_loaded_model_cost_map(loaded: ModelCostMapReloaded) -> ModelCostMa return replace(loaded, model_cost_map=_finalize_model_cost_map(loaded.model_cost_map)) +def _use_local_backup(reason: str | None) -> dict: # mutable-ok: returns the mutable model-cost map contract + _cost_map_source_info.source = "local" + _cost_map_source_info.fallback_reason = reason + return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map + + +def _accept_remote( + result: ModelCostMapReloaded, url: str +) -> dict | None: # mutable-ok: returns the mutable model-cost map contract + if not GetModelCostMap.validate_model_cost_map( + fetched_map=result.model_cost_map, + backup_model_count=GetModelCostMap._get_backup_model_count(), # pyright: ignore[reportPrivateUsage] # integrity cache + ): + verbose_logger.warning( + "LiteLLM: Fetched model cost map failed integrity check. Using local backup instead. url=%s", + url, + ) + return None + finalized: Final = _finalize_loaded_model_cost_map(result).model_cost_map + _cost_map_source_info.source = "remote" + _cost_map_source_info.url = url + _cost_map_source_info.is_env_forced = False + _cost_map_source_info.fallback_reason = None + return finalized + + def adopt_model_cost_map( new_model_cost_map: dict, # mutable-ok: public API preserves the mutable cost-map contract ) -> int: @@ -545,17 +571,20 @@ def adopt_model_cost_map( return fetched_model_count -def _continue_remote_fetch_in_background( +def _retry_remote_fetch_in_background( url: str, timeout: int, max_attempts: int, sleep: Callable[[float], None], rng: random.Random, client: _SyncGetClient, - first_wait: float, + first_outcome: _FetchAttemptRetryable, apply: Callable[[dict], object], # mutable-ok: injected callback receives the mutable cost-map dict ) -> None: try: + first_wait: Final = _next_retry_wait(outcome=first_outcome, attempt=1, max_attempts=max_attempts, rng=rng) + if isinstance(first_wait, ModelCostMapReloadUnavailable): + return sleep(first_wait) result: Final = _fetch_remote_model_cost_map_with_retry_sync( url=url, @@ -572,23 +601,12 @@ def _continue_remote_fetch_in_background( max_attempts, ) return - backup_model_count: Final = GetModelCostMap._get_backup_model_count() # pyright: ignore[reportPrivateUsage] # integrity cache - if not GetModelCostMap.validate_model_cost_map( - fetched_map=result.model_cost_map, - backup_model_count=backup_model_count, - ): - verbose_logger.warning( - "LiteLLM: Fetched model cost map failed integrity check. Keeping local backup. url=%s", - url, - ) + _litellm_import_complete.wait() + accepted: Final = _accept_remote(result, url) + if accepted is None: _cost_map_source_info.fallback_reason = "Remote data failed integrity validation" return - _litellm_import_complete.wait() - apply(_finalize_loaded_model_cost_map(result).model_cost_map) - _cost_map_source_info.source = "remote" - _cost_map_source_info.url = url - _cost_map_source_info.is_env_forced = False - _cost_map_source_info.fallback_reason = None + apply(accepted) _cost_map_source_info.loaded_at = datetime.now(timezone.utc) except Exception as e: verbose_logger.warning("LiteLLM: Background model cost map retry failed: %s", e) @@ -623,77 +641,42 @@ def get_model_cost_map( # Note: can't use get_secret_bool here — this runs during litellm.__init__ # before litellm._key_management_settings is set. if os.getenv("LITELLM_LOCAL_MODEL_COST_MAP", "").lower() == "true": - _cost_map_source_info.source = "local" _cost_map_source_info.url = None _cost_map_source_info.is_env_forced = True - _cost_map_source_info.fallback_reason = None - return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map + return _use_local_backup(None) _cost_map_source_info.url = url _cost_map_source_info.is_env_forced = False fetch_client: Final = client if client is not None else httpx fetch_rng: Final = rng if rng is not None else random.Random() - first_outcome: Final = _attempt_fetch_sync(client=fetch_client, url=url, timeout=timeout) - if isinstance(first_outcome, _FetchAttemptRetryable): + outcome: Final = _attempt_fetch_sync(client=fetch_client, url=url, timeout=timeout) + if isinstance(outcome, ModelCostMapReloaded): + accepted: Final = _accept_remote(outcome, url) + return accepted if accepted is not None else _use_local_backup("Remote data failed integrity validation") + if isinstance(outcome, _FetchAttemptRetryable) and max_attempts > 1: verbose_logger.warning( "LiteLLM: model cost map fetch attempt 1/%d failed: %s; " "using local backup while retrying in the background", max_attempts, - first_outcome.reason, + outcome.reason, ) - _cost_map_source_info.source = "local" - _cost_map_source_info.fallback_reason = f"Remote fetch failed: {first_outcome.reason}" - local_map: Final = _finalize_loaded_model_cost_map( - GetModelCostMap.load_local_model_cost_map_with_revision() - ).model_cost_map - first_wait: Final = _next_retry_wait( - outcome=first_outcome, - attempt=1, - max_attempts=max_attempts, - rng=fetch_rng, - ) - if isinstance(first_wait, ModelCostMapReloadUnavailable): - return local_map start_background( - lambda: _continue_remote_fetch_in_background( + lambda: _retry_remote_fetch_in_background( url=url, timeout=timeout, max_attempts=max_attempts, sleep=sleep, rng=fetch_rng, client=fetch_client, - first_wait=first_wait, + first_outcome=outcome, apply=apply, ) ) - return local_map - - result: Final = first_outcome - if isinstance(result, ModelCostMapReloadUnavailable): + else: verbose_logger.warning( "LiteLLM: Failed to fetch remote model cost map from %s: %s. Falling back to local backup.", url, - result.reason, + outcome.reason, ) - _cost_map_source_info.source = "local" - _cost_map_source_info.fallback_reason = f"Remote fetch failed: {result.reason}" - return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map - content: Final = result.model_cost_map - - # Validate using cached count (cheap int comparison, no file I/O) - if not GetModelCostMap.validate_model_cost_map( - fetched_map=content, - backup_model_count=GetModelCostMap._get_backup_model_count(), - ): - verbose_logger.warning( - "LiteLLM: Fetched model cost map failed integrity check. Using local backup instead. url=%s", - url, - ) - _cost_map_source_info.source = "local" - _cost_map_source_info.fallback_reason = "Remote data failed integrity validation" - return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map - - _cost_map_source_info.source = "remote" - _cost_map_source_info.fallback_reason = None - return _finalize_loaded_model_cost_map(result).model_cost_map + return _use_local_backup(f"Remote fetch failed: {outcome.reason}") From 9a721abf0d098f40caa6ba67d343e99de74f7fc2 Mon Sep 17 00:00:00 2001 From: kerry Date: Wed, 9 Sep 2026 03:50:49 +0000 Subject: [PATCH 11/26] test(cost-map): clear LITELLM_LOCAL_MODEL_COST_MAP in register_model url test Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- tests/test_litellm/test_utils.py | 1 + 1 file changed, 1 insertion(+) diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py index e90372141bd..e42608c9904 100644 --- a/tests/test_litellm/test_utils.py +++ b/tests/test_litellm/test_utils.py @@ -2386,6 +2386,7 @@ def test_register_model_with_scientific_notation(): @respx.mock def test_register_model_url_fetch_uses_single_attempt(monkeypatch): + monkeypatch.delenv("LITELLM_LOCAL_MODEL_COST_MAP", raising=False) monkeypatch.setattr(litellm, "model_cost", dict(litellm.model_cost)) before = dict(litellm.model_cost) threads_before = {thread.name for thread in threading.enumerate()} From 15392e7b3a298bec57f4aae31e863e39fb182cc6 Mon Sep 17 00:00:00 2001 From: Joshua Valluru <326636767+joshua-berri@users.noreply.github.com> Date: Tue, 8 Sep 2026 21:26:04 -0700 Subject: [PATCH 12/26] fix(mcp): surface connection test failures safely --- litellm/experimental_mcp_client/client.py | 21 +++ .../mcp_server/rest_endpoints.py | 59 ++++++- .../test_mcp_client.py | 124 ++++++++++++++- .../mcp_server/test_rest_endpoints.py | 148 +++++++++++++++++- 4 files changed, 334 insertions(+), 18 deletions(-) diff --git a/litellm/experimental_mcp_client/client.py b/litellm/experimental_mcp_client/client.py index 3503468c735..cabcc6c03ba 100644 --- a/litellm/experimental_mcp_client/client.py +++ b/litellm/experimental_mcp_client/client.py @@ -18,6 +18,7 @@ from mcp import ClientSession, McpError, ReadResourceResult, Resource, StdioServ from mcp.client.sse import sse_client from mcp.client.stdio import stdio_client from mcp.shared.message import SessionMessage +from mcp.shared.session import RequestResponder from typing_extensions import Unpack _TransportStreams: TypeAlias = tuple[ @@ -56,10 +57,13 @@ def missing_streamable_http_client_error() -> ImportError: from mcp.types import CallToolRequestParams as MCPCallToolRequestParams from mcp.types import CallToolResult as MCPCallToolResult from mcp.types import ( + ClientResult, GetPromptRequestParams, GetPromptResult, Prompt, ResourceTemplate, + ServerNotification, + ServerRequest, TextContent, ) from mcp.types import Tool as MCPTool @@ -442,6 +446,18 @@ class MCPClient: in_flight_error: BaseException | None = None try: read_stream, write_stream = transport[0], transport[1] + stream_error: Final[asyncio.Future[Exception]] = asyncio.get_running_loop().create_future() + + async def receive_message( + message: RequestResponder[ServerRequest, ClientResult] | ServerNotification | Exception, + ) -> None: + if not isinstance(message, ValueError): + return + if not stream_error.done(): + stream_error.set_result(message) + # The SDK closes pending requests when its message handler raises. + raise RuntimeError("MCP response stream failed") + # Build session kwargs with optional callbacks session_kwargs: Final[dict[str, Any]] = {} if self._sampling_callback is not None: @@ -456,6 +472,7 @@ class MCPClient: read_stream, write_stream, read_timeout_seconds=timedelta(seconds=self.timeout), + message_handler=receive_message if self.transport_type == MCPTransport.http else None, **session_kwargs, ) session: Final = await session_ctx.__aenter__() @@ -467,6 +484,10 @@ class MCPClient: if isinstance(ins, str) and ins.strip(): self._last_initialize_instructions = ins.strip() return await operation(session) + except McpError: + if stream_error.done(): + raise stream_error.result() + raise finally: try: await session_ctx.__aexit__(None, None, None) diff --git a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py index 329dddbdf05..c0ffeaeac62 100644 --- a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py +++ b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py @@ -3,12 +3,15 @@ import importlib from collections.abc import Awaitable, Callable, Mapping, Sequence from dataclasses import dataclass from datetime import datetime +from traceback import walk_tb from types import MappingProxyType from typing import TYPE_CHECKING, Any, Final, Literal +from uuid import uuid4 import anyio import httpx from fastapi import APIRouter, Depends, HTTPException, Query, Request, status +from pydantic import ValidationError from starlette.datastructures import Headers from litellm._logging import verbose_logger @@ -30,6 +33,8 @@ from litellm.proxy._experimental.mcp_server.faults.list_outcomes import ( list_fault_http_status, outcome_wire_value, ) +from litellm.proxy._experimental.mcp_server.faults.traversal import iter_exception_tree +from litellm.proxy._experimental.mcp_server.oauth_utils import _redact_mcp_resource_url from litellm.proxy._experimental.mcp_server.ui_session_utils import ( acting_user_auth, build_effective_auth_contexts, @@ -78,11 +83,38 @@ _MCP_GUARDRAIL_REJECTIONS: Final = ( def _connection_error_message(exc: BaseException, url: str | None, timeout_seconds: float) -> str: + reference: Final = uuid4().hex + verbose_logger.error( + "MCP connection test failed (reference=%s): %s", + reference, + tuple( + ( + type(cause).__name__, + tuple( + (frame.f_code.co_filename, lineno, frame.f_code.co_name) + for frame, lineno in walk_tb(cause.__traceback__) + ), + ) + for cause in iter_exception_tree(exc) + ), + ) + return next( + ( + message + for cause in iter_exception_tree(exc) + if (message := _known_connection_error_message(cause, url, timeout_seconds)) is not None + ), + "An unexpected error occurred while testing the MCP connection. " + f"Retry; if it persists, share reference {reference} with your gateway administrator.", + ) + + +def _known_connection_error_message(exc: BaseException, url: str | None, timeout_seconds: float) -> str | None: if isinstance(exc, MCPServerURLCredentialsError): return str(exc.detail) if isinstance(exc, TimeoutError): return ( - f"Failed to connect to MCP server: no response from {url or 'the server'} " + f"Failed to connect to MCP server: no response from {_redact_mcp_resource_url(url) or 'the server'} " f"within {timeout_seconds:.0f}s. Check that the LiteLLM proxy can reach this URL " "from its network (DNS, egress rules, firewalls) and that the server answers MCP requests." ) @@ -99,10 +131,32 @@ def _connection_error_message(exc: BaseException, url: str | None, timeout_secon return "Failed to connect to MCP server: the connection timed out." if isinstance(exc, httpx.HTTPStatusError): return f"Failed to connect to MCP server: it returned HTTP {exc.response.status_code}." - return "Failed to connect to MCP server. Check proxy logs for details." + if isinstance(exc, ValueError) and str(exc).startswith("Unexpected content type:"): + return ( + "Failed to connect to MCP server: the endpoint returned an unsupported content type. " + "Check that the URL is an MCP endpoint, not a web page, and matches the selected transport." + ) + if isinstance(exc, ValidationError) and exc.title in ("JSONRPCMessage", "InitializeResult", "ListToolsResult"): + return ( + "Failed to connect to MCP server: the endpoint returned invalid JSON or an invalid MCP response. " + "Check the MCP endpoint URL and the server's protocol implementation." + ) + if MCP_AVAILABLE and isinstance(exc, McpError): + if exc.error.code == 32600 and exc.error.message == "Session terminated": + return ( + "Failed to connect to MCP server: the MCP session was terminated. " + "Check that the URL points to an MCP endpoint and matches the selected transport, " + "then retry to start a new session." + ) + return ( + f"Failed to connect to MCP server: the MCP request failed (JSON-RPC code {exc.error.code}). " + "Check that the endpoint supports MCP initialization and tool listing, and check the upstream server logs." + ) + return None if MCP_AVAILABLE: + from mcp.shared.exceptions import McpError from mcp.types import Tool as MCPTool from litellm.experimental_mcp_client.client import MCPClient @@ -1342,7 +1396,6 @@ if MCP_AVAILABLE: except (KeyboardInterrupt, SystemExit, asyncio.CancelledError): raise except BaseException as e: - verbose_logger.error("Error in MCP operation: %s", e, exc_info=True) return { "status": "error", "error": True, diff --git a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py index 4db131da62c..49268b4d22d 100644 --- a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py +++ b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py @@ -1,9 +1,11 @@ import asyncio import base64 +import json import os import sys from importlib import metadata from pathlib import Path +from typing import Final from unittest.mock import AsyncMock, MagicMock, patch import anyio @@ -11,6 +13,8 @@ import httpx import pytest from litellm.proxy._experimental.mcp_server.outbound_credentials.httpx_auth import StaticHeaderAuth from mcp import McpError +from mcp.client.streamable_http import streamable_http_client +from pydantic import ValidationError from mcp.shared.message import SessionMessage from mcp.types import ( LATEST_PROTOCOL_VERSION, @@ -1224,14 +1228,14 @@ def test_without_a_configured_slot_the_existing_precedence_is_unchanged(): _REDIRECT_CASES = [ - ("https://upstream.example.com/mcp", "https://upstream.example.com/other"), # same origin + ("https://upstream.example.com/mcp", "https://upstream.example.com/other"), # same origin ("https://upstream.example.com/mcp", "https://upstream.example.com:443/other"), # explicit default port - ("https://upstream.example.com/mcp", "https://attacker.example.com/collect"), # different host - ("https://upstream.example.com/mcp", "http://upstream.example.com/collect"), # scheme downgrade - ("https://upstream.example.com/mcp", "https://upstream.example.com:8443/other"), # different port - ("https://upstream.example.com/mcp", "https://sub.upstream.example.com/x"), # different host - ("http://upstream.example.com/mcp", "https://upstream.example.com/other"), # http -> https upgrade - ("http://upstream.example.com/mcp", "http://upstream.example.com/other"), # same origin, plain http + ("https://upstream.example.com/mcp", "https://attacker.example.com/collect"), # different host + ("https://upstream.example.com/mcp", "http://upstream.example.com/collect"), # scheme downgrade + ("https://upstream.example.com/mcp", "https://upstream.example.com:8443/other"), # different port + ("https://upstream.example.com/mcp", "https://sub.upstream.example.com/x"), # different host + ("http://upstream.example.com/mcp", "https://upstream.example.com/other"), # http -> https upgrade + ("http://upstream.example.com/mcp", "http://upstream.example.com/other"), # same origin, plain http ] @@ -1283,3 +1287,109 @@ def test_a_differently_cased_injected_header_cannot_shadow_the_slot() -> None: headers = client._get_auth_headers() assert [v for k, v in headers.items() if k.lower() == "esb-oauth"] == ["Bearer minted-token"] assert headers["X-Trace"] == "keep" + + +@pytest.mark.asyncio +@pytest.mark.parametrize( + ("content_type", "body", "expected_type"), + [ + ("text/html", b"secret-page", ValueError), + ("application/json", b"secret-invalid-json", ValidationError), + ("application/json", b'{"secret":"invalid-rpc"}', ValidationError), + ("application/json", b'{"jsonrpc":"2.0","id":0,"result":{"secret":"invalid-schema"}}', ValidationError), + ], +) +async def test_invalid_http_response_surfaces_without_waiting_for_timeout( + content_type: str, body: bytes, expected_type: type[Exception] +) -> None: + from litellm.proxy._experimental.mcp_server.rest_endpoints import _connection_error_message + + def respond(request: httpx.Request) -> httpx.Response: + return httpx.Response(200, headers={"Content-Type": content_type}, content=body) + + async with httpx.AsyncClient(transport=httpx.MockTransport(respond)) as http_client: + client: Final = MCPClient(server_url="https://example.com/mcp", timeout=30) + with pytest.raises(expected_type) as caught: + await asyncio.wait_for( + client._execute_session_operation( + streamable_http_client(client.server_url, http_client=http_client), + lambda session: session.list_tools(), + ), + timeout=3, + ) + + message: Final = _connection_error_message(caught.value, client.server_url, 30) + assert "unsupported content type" in message or "invalid MCP response" in message + assert "secret" not in message + assert "timed out" not in message + + +@pytest.mark.asyncio +@pytest.mark.parametrize("status_code", [200, 401, 503]) +async def test_http_response_handler_preserves_success_and_http_errors(status_code: int) -> None: + def respond(request: httpx.Request) -> httpx.Response: + if request.method == "DELETE": + return httpx.Response(200) + payload: Final = json.loads(request.content) + if "id" not in payload: + return httpx.Response(202) + result: Final = ( + { + "protocolVersion": LATEST_PROTOCOL_VERSION, + "capabilities": {}, + "serverInfo": {"name": "test", "version": "1"}, + } + if payload["method"] == "initialize" + else {"tools": []} + ) + return httpx.Response(status_code, json={"jsonrpc": "2.0", "id": payload["id"], "result": result}) + + async with httpx.AsyncClient(transport=httpx.MockTransport(respond)) as http_client: + client: Final = MCPClient(server_url="https://example.com/mcp", timeout=30) + operation: Final = client._execute_session_operation( + streamable_http_client(client.server_url, http_client=http_client), lambda session: session.list_tools() + ) + if status_code == 200: + result: Final = await asyncio.wait_for(operation, timeout=3) + assert result.tools == [] + else: + with pytest.raises(httpx.HTTPStatusError) as caught: + await asyncio.wait_for(operation, timeout=3) + assert caught.value.response.status_code == status_code + + +@pytest.mark.asyncio +async def test_invalid_tool_list_schema_is_identified_as_an_upstream_response() -> None: + from litellm.proxy._experimental.mcp_server.rest_endpoints import _connection_error_message + + def respond(request: httpx.Request) -> httpx.Response: + if request.method == "DELETE": + return httpx.Response(200) + payload: Final = json.loads(request.content) + if "id" not in payload: + return httpx.Response(202) + result: Final = ( + { + "protocolVersion": LATEST_PROTOCOL_VERSION, + "capabilities": {}, + "serverInfo": {"name": "test", "version": "1"}, + } + if payload["method"] == "initialize" + else {"tools": "secret-invalid-tools"} + ) + return httpx.Response(200, json={"jsonrpc": "2.0", "id": payload["id"], "result": result}) + + async with httpx.AsyncClient(transport=httpx.MockTransport(respond)) as http_client: + client: Final = MCPClient(server_url="https://example.com/mcp", timeout=30) + with pytest.raises(ValidationError) as caught: + await asyncio.wait_for( + client._execute_session_operation( + streamable_http_client(client.server_url, http_client=http_client), + lambda session: session.list_tools(), + ), + timeout=3, + ) + + message: Final = _connection_error_message(caught.value, client.server_url, 30) + assert "invalid MCP response" in message + assert "secret" not in message diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py index bd692776c82..73734131130 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py @@ -3,7 +3,7 @@ import inspect import json import sys from datetime import datetime -from typing import Any, Dict, Optional +from typing import Any, Dict, Final, Optional from unittest.mock import AsyncMock, MagicMock if sys.version_info < (3, 11): # BaseExceptionGroup is a builtin only from 3.11 @@ -113,7 +113,7 @@ class TestExecuteWithMcpClient: assert "stack_trace" not in result @pytest.mark.asyncio - async def test_timeout_caps_hanging_operation_and_names_url(self, monkeypatch): + async def test_timeout_caps_hanging_operation_and_names_origin(self, monkeypatch): async def fake_create_client(*args, **kwargs): return object() @@ -138,7 +138,7 @@ class TestExecuteWithMcpClient: ) assert result["error"] is True - assert "https://mcp.example.com/mcp/" in result["message"] + assert "https://mcp.example.com" in result["message"] @pytest.mark.asyncio async def test_timeout_covers_client_creation(self, monkeypatch): @@ -166,15 +166,15 @@ class TestExecuteWithMcpClient: ) assert result["error"] is True - assert "https://mcp.example.com/mcp/" in result["message"] + assert "https://mcp.example.com" in result["message"] def test_timeout_defaults_to_tool_listing_timeout(self): default = inspect.signature(rest_endpoints._execute_with_mcp_client).parameters["timeout_seconds"].default assert default == MCP_TOOL_LISTING_TIMEOUT - def test_connection_error_message_timeout_names_url_and_budget(self): + def test_connection_error_message_timeout_names_origin_and_budget(self): message = rest_endpoints._connection_error_message(TimeoutError(), "https://api.example.com/mcp/", 30.0) - assert "https://api.example.com/mcp/" in message + assert "https://api.example.com" in message assert "30s" in message def test_connection_error_message_hides_arbitrary_http_exception_detail(self): @@ -592,7 +592,7 @@ class TestExecuteWithMcpClient: assert result["status"] == "error" assert result["error"] is True - assert "Failed to connect to MCP server" in result["message"] + assert "reference" in result["message"] # Error message must not leak raw exception details assert "cancel scope" not in result["message"] @@ -3430,7 +3430,139 @@ class TestConnectionErrorMessage: def test_unknown_error_falls_back_to_generic(self): message = rest_endpoints._connection_error_message(RuntimeError("weird"), "https://example.com", 30.0) assert "weird" not in message - assert "proxy logs" in message.lower() + assert "reference" in message.lower() + + def test_sdk_session_terminated_explains_endpoint_and_retry(self) -> None: + from mcp.shared.exceptions import McpError + from mcp.types import ErrorData + + message: Final = rest_endpoints._connection_error_message( + McpError(ErrorData(code=32600, message="Session terminated")), "https://example.com/mcp", 30.0 + ) + + assert "session was terminated" in message + assert "MCP endpoint" in message + assert "transport" in message + assert "retry" in message + assert "404" not in message + + @pytest.mark.parametrize("code", [-32700, -32601, -32602, -32603, -32000, 32600, 408]) + def test_rpc_errors_include_code_without_echoing_upstream_data(self, code: int) -> None: + from mcp.shared.exceptions import McpError + from mcp.types import ErrorData + + message: Final = rest_endpoints._connection_error_message( + McpError(ErrorData(code=code, message="secret-message", data={"token": "secret-data"})), + "https://example.com/secret-path?token=secret-query", + 30.0, + ) + + assert f"JSON-RPC code {code}" in message + assert "secret" not in message + assert "timed out" not in message + assert "session was terminated" not in message + + @pytest.mark.parametrize("status_code", [401, 403, 404, 405, 429, 503]) + def test_wrapped_http_failures_preserve_status(self, status_code: int) -> None: + response: Final = httpx.Response(status_code, text="secret-body") + upstream: Final = httpx.HTTPStatusError( + "secret-exception", + request=httpx.Request("POST", "https://example.com/?token=secret-query"), + response=response, + ) + wrapped: Final = BaseExceptionGroup( + "secret-group", [asyncio.CancelledError(), BaseExceptionGroup("nested", [upstream])] + ) + + message: Final = rest_endpoints._connection_error_message(wrapped, "https://example.com", 30.0) + + assert f"HTTP {status_code}" in message + assert "secret" not in message + + def test_explicit_cause_is_classified_before_incidental_context(self) -> None: + wrapped: Final = RuntimeError("secret-wrapper") + wrapped.__cause__ = httpx.ConnectError("secret-cause") + wrapped.__context__ = TimeoutError("secret-context") + + message: Final = rest_endpoints._connection_error_message(wrapped, "https://example.com", 30.0) + + assert "unreachable" in message + assert "secret" not in message + + def test_timeout_url_redacts_credentials_path_query_and_fragment(self) -> None: + message: Final = rest_endpoints._connection_error_message( + TimeoutError("secret-error"), + "https://secret-user:secret-pass@example.com:8443/secret-path?token=secret-query#secret-fragment", + 30.0, + ) + + assert "https://example.com:8443" in message + assert "30s" in message + assert "secret" not in message + + def test_unknown_failure_reference_matches_safe_diagnostics(self, caplog: pytest.LogCaptureFixture) -> None: + import re + + try: + raise RuntimeError("secret-exception-body") + except RuntimeError as exc: + message: Final = rest_endpoints._connection_error_message( + exc, "https://secret-user:secret-password@example.com/secret-path?token=secret-query", 30.0 + ) + + reference: Final = re.search(r"reference ([a-f0-9]{32})", message) + assert reference is not None + diagnostics: Final = tuple( + record for record in caplog.records if "MCP connection test failed" in record.message + ) + assert len(diagnostics) == 1 + assert reference.group(1) in diagnostics[0].message + assert "RuntimeError" in diagnostics[0].message + assert "test_unknown_failure_reference_matches_safe_diagnostics" in diagnostics[0].message + assert diagnostics[0].exc_info is None + assert "secret" not in message + diagnostics[0].message + + @pytest.mark.parametrize("exc", [ValueError("secret-config"), HTTPException(500, "secret-detail")]) + def test_unrelated_errors_are_not_misreported_as_invalid_mcp(self, exc: Exception) -> None: + message: Final = rest_endpoints._connection_error_message(exc, "https://example.com", 30.0) + + assert "reference" in message + assert "invalid MCP response" not in message + assert "secret" not in message + + def test_configuration_validation_error_uses_unknown_fallback(self) -> None: + from pydantic import ValidationError + + with pytest.raises(ValidationError) as caught: + NewMCPServerRequest.model_validate({"server_name": "example", "transport": "secret-invalid-transport"}) + + message: Final = rest_endpoints._connection_error_message(caught.value, "https://example.com", 30.0) + assert "reference" in message + assert "invalid MCP response" not in message + assert "secret" not in message + + @pytest.mark.asyncio + async def test_connection_test_preserves_cancellation(self) -> None: + async def cancelled_operation(client: rest_endpoints.MCPClient) -> dict[str, object]: + raise asyncio.CancelledError + + payload: Final = NewMCPServerRequest(server_name="cancelled", url="https://example.com", auth_type=MCPAuth.none) + with pytest.raises(asyncio.CancelledError): + await rest_endpoints._execute_with_mcp_client(payload, cancelled_operation) + + @pytest.mark.asyncio + async def test_unknown_failure_preserves_response_contract(self) -> None: + async def failing_operation(client: rest_endpoints.MCPClient) -> dict[str, object]: + raise RuntimeError("secret-operation") + + payload: Final = NewMCPServerRequest(server_name="unknown", url="https://example.com", auth_type=MCPAuth.none) + result: Final = await rest_endpoints._execute_with_mcp_client(payload, failing_operation) + + assert result["error"] is True + assert result["status"] == "error" + assert "reference" in result["message"] + assert "secret" not in result["message"] + assert "stack_trace" not in result class TestGetServerAuthHeaderGroupDefault: From 7b6ef9206e87c09adfc5844c8f8130da73d794fb Mon Sep 17 00:00:00 2001 From: Joshua Valluru <326636767+joshua-berri@users.noreply.github.com> Date: Tue, 8 Sep 2026 22:14:47 -0700 Subject: [PATCH 13/26] test(mcp): cover server notifications during tool listing --- .../test_mcp_client.py | 53 +++++++++++++++++++ 1 file changed, 53 insertions(+) diff --git a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py index 49268b4d22d..90292b9b162 100644 --- a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py +++ b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py @@ -24,6 +24,7 @@ from mcp.types import ( JSONRPCError, JSONRPCMessage, JSONRPCResponse, + LoggingMessageNotificationParams, ServerCapabilities, ) @@ -1358,6 +1359,58 @@ async def test_http_response_handler_preserves_success_and_http_errors(status_co assert caught.value.response.status_code == status_code +@pytest.mark.asyncio +async def test_http_response_handler_preserves_notifications_and_tool_listing() -> None: + notification: Final = { + "jsonrpc": "2.0", + "method": "notifications/message", + "params": {"level": "info", "data": "Listing tools"}, + } + logging_callback: Final = AsyncMock() + + def respond(request: httpx.Request) -> httpx.Response: + if request.method == "DELETE": + return httpx.Response(200) + payload: Final = json.loads(request.content) + if "id" not in payload: + return httpx.Response(202) + if payload["method"] == "initialize": + return httpx.Response( + 200, + json={ + "jsonrpc": "2.0", + "id": payload["id"], + "result": { + "protocolVersion": LATEST_PROTOCOL_VERSION, + "capabilities": {"logging": {}, "tools": {}}, + "serverInfo": {"name": "test", "version": "1"}, + }, + }, + ) + response: Final = { + "jsonrpc": "2.0", + "id": payload["id"], + "result": {"tools": [{"name": "search", "inputSchema": {"type": "object"}}]}, + } + return httpx.Response( + 200, + headers={"Content-Type": "text/event-stream"}, + content="".join(f"event: message\ndata: {json.dumps(message)}\n\n" for message in (notification, response)), + ) + + async with httpx.AsyncClient(transport=httpx.MockTransport(respond)) as http_client: + client: Final = MCPClient(server_url="https://example.com/mcp", timeout=30, logging_callback=logging_callback) + result: Final = await asyncio.wait_for( + client._execute_session_operation( + streamable_http_client(client.server_url, http_client=http_client), lambda session: session.list_tools() + ), + timeout=3, + ) + + assert [tool.name for tool in result.tools] == ["search"] + logging_callback.assert_awaited_once_with(LoggingMessageNotificationParams(level="info", data="Listing tools")) + + @pytest.mark.asyncio async def test_invalid_tool_list_schema_is_identified_as_an_upstream_response() -> None: from litellm.proxy._experimental.mcp_server.rest_endpoints import _connection_error_message From 884f90c72715b0cfe0d5e2ea09f3744cb8376a78 Mon Sep 17 00:00:00 2001 From: kerry Date: Wed, 9 Sep 2026 05:15:53 +0000 Subject: [PATCH 14/26] refactor(cost-map): inline background retry, trim tests Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../litellm_core_utils/get_model_cost_map.py | 119 ++++----- .../test_get_model_cost_map.py | 243 +++++++----------- 2 files changed, 139 insertions(+), 223 deletions(-) diff --git a/litellm/litellm_core_utils/get_model_cost_map.py b/litellm/litellm_core_utils/get_model_cost_map.py index 67d06a0758e..f81ddbfee2e 100644 --- a/litellm/litellm_core_utils/get_model_cost_map.py +++ b/litellm/litellm_core_utils/get_model_cost_map.py @@ -184,10 +184,6 @@ def mark_litellm_import_complete() -> None: _litellm_import_complete.set() -def _start_daemon_thread(fn: Callable[[], None]) -> None: - threading.Thread(target=fn, name="litellm-model-cost-map-retry", daemon=True).start() - - @dataclass(frozen=True, slots=True) class ModelCostMapReloaded: model_cost_map: dict # mutable-ok: adopted as litellm.model_cost, whose consumer contract is a plain mutable dict @@ -531,32 +527,6 @@ def _finalize_loaded_model_cost_map(loaded: ModelCostMapReloaded) -> ModelCostMa return replace(loaded, model_cost_map=_finalize_model_cost_map(loaded.model_cost_map)) -def _use_local_backup(reason: str | None) -> dict: # mutable-ok: returns the mutable model-cost map contract - _cost_map_source_info.source = "local" - _cost_map_source_info.fallback_reason = reason - return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map - - -def _accept_remote( - result: ModelCostMapReloaded, url: str -) -> dict | None: # mutable-ok: returns the mutable model-cost map contract - if not GetModelCostMap.validate_model_cost_map( - fetched_map=result.model_cost_map, - backup_model_count=GetModelCostMap._get_backup_model_count(), # pyright: ignore[reportPrivateUsage] # integrity cache - ): - verbose_logger.warning( - "LiteLLM: Fetched model cost map failed integrity check. Using local backup instead. url=%s", - url, - ) - return None - finalized: Final = _finalize_loaded_model_cost_map(result).model_cost_map - _cost_map_source_info.source = "remote" - _cost_map_source_info.url = url - _cost_map_source_info.is_env_forced = False - _cost_map_source_info.fallback_reason = None - return finalized - - def adopt_model_cost_map( new_model_cost_map: dict, # mutable-ok: public API preserves the mutable cost-map contract ) -> int: @@ -579,7 +549,6 @@ def _retry_remote_fetch_in_background( rng: random.Random, client: _SyncGetClient, first_outcome: _FetchAttemptRetryable, - apply: Callable[[dict], object], # mutable-ok: injected callback receives the mutable cost-map dict ) -> None: try: first_wait: Final = _next_retry_wait(outcome=first_outcome, attempt=1, max_attempts=max_attempts, rng=rng) @@ -602,12 +571,20 @@ def _retry_remote_fetch_in_background( ) return _litellm_import_complete.wait() - accepted: Final = _accept_remote(result, url) - if accepted is None: - _cost_map_source_info.fallback_reason = "Remote data failed integrity validation" + if not GetModelCostMap.validate_model_cost_map( + fetched_map=result.model_cost_map, + backup_model_count=GetModelCostMap._get_backup_model_count(), # pyright: ignore[reportPrivateUsage] # integrity cache + ): + verbose_logger.warning( + "LiteLLM: Fetched model cost map failed integrity check. Using local backup instead. url=%s", + url, + ) return - apply(accepted) + finalized: Final = _finalize_loaded_model_cost_map(result).model_cost_map + _cost_map_source_info.source = "remote" + _cost_map_source_info.fallback_reason = None _cost_map_source_info.loaded_at = datetime.now(timezone.utc) + adopt_model_cost_map(finalized) except Exception as e: verbose_logger.warning("LiteLLM: Background model cost map retry failed: %s", e) @@ -619,19 +596,12 @@ def get_model_cost_map( sleep: Callable[[float], None] = time.sleep, rng: random.Random | None = None, client: "_SyncGetClient | None" = None, - start_background: Callable[[Callable[[], None]], None] = _start_daemon_thread, - apply: Callable[ # mutable-ok: injected callback receives the mutable cost-map dict - [dict], - object, - ] = adopt_model_cost_map, ) -> dict: """ Public entry point — returns the model cost map dict. 1. If ``LITELLM_LOCAL_MODEL_COST_MAP`` is set, uses the local backup only. - 2. Otherwise fetches from ``url``, validates the first response, and falls - back to the local backup while retrying transient HTTP errors in the - background. + 2. Otherwise fetches from ``url``, retrying transient errors in a background thread. Only the backup model count is cached (a single int) for validation. The full backup dict is only parsed when it must be *returned* as a @@ -641,9 +611,11 @@ def get_model_cost_map( # Note: can't use get_secret_bool here — this runs during litellm.__init__ # before litellm._key_management_settings is set. if os.getenv("LITELLM_LOCAL_MODEL_COST_MAP", "").lower() == "true": + _cost_map_source_info.source = "local" _cost_map_source_info.url = None _cost_map_source_info.is_env_forced = True - return _use_local_backup(None) + _cost_map_source_info.fallback_reason = None + return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map _cost_map_source_info.url = url _cost_map_source_info.is_env_forced = False @@ -651,32 +623,45 @@ def get_model_cost_map( fetch_client: Final = client if client is not None else httpx fetch_rng: Final = rng if rng is not None else random.Random() outcome: Final = _attempt_fetch_sync(client=fetch_client, url=url, timeout=timeout) - if isinstance(outcome, ModelCostMapReloaded): - accepted: Final = _accept_remote(outcome, url) - return accepted if accepted is not None else _use_local_backup("Remote data failed integrity validation") if isinstance(outcome, _FetchAttemptRetryable) and max_attempts > 1: - verbose_logger.warning( - "LiteLLM: model cost map fetch attempt 1/%d failed: %s; " - "using local backup while retrying in the background", - max_attempts, - outcome.reason, - ) - start_background( - lambda: _retry_remote_fetch_in_background( - url=url, - timeout=timeout, - max_attempts=max_attempts, - sleep=sleep, - rng=fetch_rng, - client=fetch_client, - first_outcome=outcome, - apply=apply, - ) - ) - else: + threading.Thread( + target=_retry_remote_fetch_in_background, + kwargs={ # mutable-ok: threading requires a mutable keyword-arguments mapping + "url": url, + "timeout": timeout, + "max_attempts": max_attempts, + "sleep": sleep, + "rng": fetch_rng, + "client": fetch_client, + "first_outcome": outcome, + }, + name="litellm-model-cost-map-retry", + daemon=True, + ).start() + if not isinstance(outcome, ModelCostMapReloaded): verbose_logger.warning( "LiteLLM: Failed to fetch remote model cost map from %s: %s. Falling back to local backup.", url, outcome.reason, ) - return _use_local_backup(f"Remote fetch failed: {outcome.reason}") + _cost_map_source_info.source = "local" + _cost_map_source_info.fallback_reason = f"Remote fetch failed: {outcome.reason}" + return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map + content: Final = outcome.model_cost_map + + # Validate using cached count (cheap int comparison, no file I/O) + if not GetModelCostMap.validate_model_cost_map( + fetched_map=content, + backup_model_count=GetModelCostMap._get_backup_model_count(), + ): + verbose_logger.warning( + "LiteLLM: Fetched model cost map failed integrity check. Using local backup instead. url=%s", + url, + ) + _cost_map_source_info.source = "local" + _cost_map_source_info.fallback_reason = "Remote data failed integrity validation" + return _finalize_loaded_model_cost_map(GetModelCostMap.load_local_model_cost_map_with_revision()).model_cost_map + + _cost_map_source_info.source = "remote" + _cost_map_source_info.fallback_reason = None + return _finalize_loaded_model_cost_map(outcome).model_cost_map diff --git a/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py b/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py index e39658a8f37..266c2ca1465 100644 --- a/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py +++ b/tests/test_litellm/litellm_core_utils/test_get_model_cost_map.py @@ -6,6 +6,7 @@ count actual model entries, not reserved meta keys) and the extraction of the import json import os +import threading import pytest @@ -20,7 +21,6 @@ from litellm.litellm_core_utils.get_model_cost_map import ( GetModelCostMap, _count_model_entries, _finalize_model_cost_map, - adopt_model_cost_map, get_model_cost_map_provenance, git_blob_id, ) @@ -566,194 +566,125 @@ from litellm.litellm_core_utils.get_model_cost_map import ( class _SyncSleepRecorder: """Injected in place of time.sleep so the boot path's waits are asserted without delay.""" - def __init__(self): + def __init__(self, block=False): self.waits = [] + self.block = block + self.started = threading.Event() + self.release = threading.Event() def __call__(self, seconds: float) -> None: + if self.block: + self.started.set() + self.release.wait(timeout=10) self.waits.append(seconds) -class _BackgroundRecorder: - def __init__(self): - self.callbacks = [] - - def __call__(self, callback): - self.callbacks.append(callback) +def _retry_threads(): + return [thread for thread in threading.enumerate() if thread.name == "litellm-model-cost-map-retry"] -def test_boot_load_returns_local_map_and_schedules_transient_retry(): - client, calls = _mock_client( - [ - httpx.ConnectError("connection refused"), - ], - client_cls=httpx.Client, - ) - sleeper = _SyncSleepRecorder() - background = _BackgroundRecorder() - - cost_map = get_model_cost_map( - url=_URL, - sleep=sleeper, - rng=random.Random(0), - client=client, - start_background=background, - ) - assert calls["count"] == 1 - assert sleeper.waits == [] - assert len(background.callbacks) == 1 - assert len(cost_map) > 100 - source = get_model_cost_map_source_info() - assert source["source"] == "local" - assert source["fallback_reason"] is not None - - -def test_background_retry_adopts_valid_remote_map(): - client, calls = _mock_client( - [ - httpx.ConnectError("connection refused"), - httpx.Response(200, content=_real_map_bytes()), - ], - client_cls=httpx.Client, - ) - sleeper = _SyncSleepRecorder() - background = _BackgroundRecorder() - applied = [] - - get_model_cost_map( - url=_URL, - sleep=sleeper, - rng=random.Random(0), - client=client, - start_background=background, - apply=applied.append, - ) - assert calls["count"] == 1 - assert sleeper.waits == [] - assert len(background.callbacks) == 1 - - background.callbacks[0]() - - assert calls["count"] == 2 - assert len(sleeper.waits) == 1 - assert 2.0 <= sleeper.waits[0] < 3.0 - assert len(applied) == 1 - assert applied[0].keys() >= _load_root_cost_map().keys() - {"sample_spec", FALLBACK_GENERALIZATIONS_KEY} - source = get_model_cost_map_source_info() - assert source["source"] == "remote" - assert source["fallback_reason"] is None - - -def test_background_retry_keeps_local_map_after_remaining_failures(): - client, calls = _mock_client( - [httpx.ConnectError("connection refused"), httpx.Response(503)], - client_cls=httpx.Client, - ) - sleeper = _SyncSleepRecorder() - background = _BackgroundRecorder() - applied = [] - - get_model_cost_map( - url=_URL, - sleep=sleeper, - rng=random.Random(0), - client=client, - start_background=background, - apply=applied.append, - ) - background.callbacks[0]() - - assert calls["count"] == 3 - assert len(sleeper.waits) == 2 - assert 2.0 <= sleeper.waits[0] < 3.0 - assert 4.0 <= sleeper.waits[1] < 5.0 - assert applied == [] - assert get_model_cost_map_source_info()["source"] == "local" - - -def test_boot_load_does_not_schedule_non_retryable_failure(): - client, calls = _mock_client([httpx.Response(404)], client_cls=httpx.Client) - sleeper = _SyncSleepRecorder() - background = _BackgroundRecorder() - - get_model_cost_map( - url=_URL, - sleep=sleeper, - rng=random.Random(0), - client=client, - start_background=background, - ) - - assert calls["count"] == 1 - assert sleeper.waits == [] - assert background.callbacks == [] - source = get_model_cost_map_source_info() - assert source["source"] == "local" - assert source["fallback_reason"] is not None - - -def test_boot_load_success_does_not_schedule_background_retry(): +def test_boot_load_success_does_not_start_background_retry(): client, calls = _mock_client([httpx.Response(200, content=_real_map_bytes())], client_cls=httpx.Client) sleeper = _SyncSleepRecorder() - background = _BackgroundRecorder() cost_map = get_model_cost_map( url=_URL, sleep=sleeper, rng=random.Random(0), client=client, - start_background=background, ) assert calls["count"] == 1 assert sleeper.waits == [] - assert background.callbacks == [] - assert get_model_cost_map_source_info()["source"] == "remote" + assert _retry_threads() == [] assert cost_map.keys() >= _load_root_cost_map().keys() - {"sample_spec", FALLBACK_GENERALIZATIONS_KEY} + assert get_model_cost_map_source_info()["source"] == "remote" -def test_boot_load_with_one_attempt_does_not_schedule_background_retry(): - client, calls = _mock_client([httpx.ConnectError("connection refused")], client_cls=httpx.Client) - sleeper = _SyncSleepRecorder() - background = _BackgroundRecorder() +def test_boot_load_transient_failure_returns_local_then_background_retry_adopts_remote(monkeypatch): + import litellm + from litellm import utils as litellm_utils + from litellm.litellm_core_utils import get_model_cost_map as module - get_model_cost_map( + original_model_cost = litellm.model_cost + monkeypatch.setattr(litellm, "model_cost", dict(original_model_cost)) + for name, provider_models in tuple(vars(litellm).items()): + if name.endswith("_models") and isinstance(provider_models, set): + monkeypatch.setattr(litellm, name, set(provider_models)) + monkeypatch.setattr(litellm, "models_by_provider", dict(litellm.models_by_provider)) + monkeypatch.setattr( + litellm_utils, + "_runtime_registered_model_cost", + dict(litellm_utils._runtime_registered_model_cost), + ) + source_info = module._cost_map_source_info + for name in ("source", "url", "is_env_forced", "fallback_reason", "loaded_at", "source_revision", "etag"): + monkeypatch.setattr(source_info, name, getattr(source_info, name)) + + remote_map = _load_root_cost_map() + remote_map["claude-remote-only-test"] = {"litellm_provider": "anthropic", "mode": "chat"} + client, calls = _mock_client( + [ + httpx.ConnectError("connection refused"), + httpx.Response(200, content=json.dumps(remote_map).encode()), + ], + client_cls=httpx.Client, + ) + sleeper = _SyncSleepRecorder(block=True) + litellm.register_model({"my-runtime-model": {"litellm_provider": "custom", "max_input_tokens": 4321}}) + + cost_map = get_model_cost_map( url=_URL, - max_attempts=1, + max_attempts=3, sleep=sleeper, rng=random.Random(0), client=client, - start_background=background, ) assert calls["count"] == 1 assert sleeper.waits == [] - assert background.callbacks == [] - assert get_model_cost_map_source_info()["source"] == "local" - - -def test_adopt_model_cost_map_replays_runtime_registration_and_provider_models(): - import litellm - from litellm import utils as litellm_utils - - original_model_cost = litellm.model_cost - original_registry = dict(litellm_utils._runtime_registered_model_cost) - original_anthropic_models = set(litellm.anthropic_models) + assert sleeper.started.wait(timeout=10) + threads = _retry_threads() try: - litellm.register_model( - model_cost={"custom/deployment-model": {"litellm_provider": "custom", "max_input_tokens": 4321}} - ) - - models_count = adopt_model_cost_map({"anthropic/new-model": {"litellm_provider": "anthropic", "mode": "chat"}}) - - assert models_count == 1 - assert "anthropic/new-model" in litellm.anthropic_models - assert litellm.model_cost["custom/deployment-model"]["max_input_tokens"] == 4321 + assert len(threads) == 1 + assert "claude-remote-only-test" not in cost_map + assert cost_map.keys() == _finalize_model_cost_map(GetModelCostMap.load_local_model_cost_map()).keys() + source = get_model_cost_map_source_info() + assert source["source"] == "local" + assert source["fallback_reason"].startswith("Remote fetch failed:") + sleeper.release.set() + for thread in threads: + thread.join(timeout=10) + assert all(not thread.is_alive() for thread in threads) + assert sleeper.waits and 2.0 <= sleeper.waits[0] < 3.0 + assert calls["count"] == 2 + assert "claude-remote-only-test" in litellm.model_cost + assert "claude-remote-only-test" in litellm.anthropic_models + assert "my-runtime-model" in litellm.model_cost + source = get_model_cost_map_source_info() + assert source["source"] == "remote" + assert source["fallback_reason"] is None finally: - litellm.model_cost = original_model_cost # test-quality-ok: restore the module state changed by adoption - litellm_utils._runtime_registered_model_cost.clear() - litellm_utils._runtime_registered_model_cost.update(original_registry) - litellm.anthropic_models.clear() - litellm.anthropic_models.update(original_anthropic_models) - litellm_utils._invalidate_model_cost_lowercase_map() + sleeper.release.set() + for thread in _retry_threads(): + thread.join(timeout=10) + + +def test_boot_load_does_not_retry_non_retryable_failure(): + client, calls = _mock_client([httpx.Response(404)], client_cls=httpx.Client) + sleeper = _SyncSleepRecorder() + + get_model_cost_map( + url=_URL, + sleep=sleeper, + rng=random.Random(0), + client=client, + ) + assert calls["count"] == 1 + assert sleeper.waits == [] + assert _retry_threads() == [] + source = get_model_cost_map_source_info() + assert source["source"] == "local" + assert source["fallback_reason"] is not None def test_boot_load_respects_local_env_override(monkeypatch): From a1588c260232dd6ed22605fdd4db24983bb1fcee Mon Sep 17 00:00:00 2001 From: Joshua Valluru <326636767+joshua-berri@users.noreply.github.com> Date: Wed, 9 Sep 2026 08:05:15 -0700 Subject: [PATCH 15/26] fix(mcp): preserve stream failures across transports --- litellm/experimental_mcp_client/client.py | 7 +- .../mcp_server/rest_endpoints.py | 25 +- .../test_mcp_client.py | 240 ++++++++++++++++++ .../mcp_server/test_rest_endpoints.py | 48 ++++ 4 files changed, 313 insertions(+), 7 deletions(-) diff --git a/litellm/experimental_mcp_client/client.py b/litellm/experimental_mcp_client/client.py index cabcc6c03ba..5cdc4efdfef 100644 --- a/litellm/experimental_mcp_client/client.py +++ b/litellm/experimental_mcp_client/client.py @@ -451,7 +451,7 @@ class MCPClient: async def receive_message( message: RequestResponder[ServerRequest, ClientResult] | ServerNotification | Exception, ) -> None: - if not isinstance(message, ValueError): + if not isinstance(message, (ValueError, httpx.RequestError, OSError)): return if not stream_error.done(): stream_error.set_result(message) @@ -472,7 +472,7 @@ class MCPClient: read_stream, write_stream, read_timeout_seconds=timedelta(seconds=self.timeout), - message_handler=receive_message if self.transport_type == MCPTransport.http else None, + message_handler=receive_message, **session_kwargs, ) session: Final = await session_ctx.__aenter__() @@ -525,8 +525,7 @@ class MCPClient: read_timeout: Final = _as_read_timeout(e) if read_timeout is not None: verbose_logger.warning( - "MCP client timed out after %ss waiting for %s to answer; the server accepted the " - "request and ended its response stream without a JSON-RPC reply", + "MCP client timed out after %ss waiting for a valid MCP response from %s", self.timeout, self.server_url or "stdio", ) diff --git a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py index c0ffeaeac62..63badc6f703 100644 --- a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py +++ b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py @@ -114,7 +114,8 @@ def _known_connection_error_message(exc: BaseException, url: str | None, timeout return str(exc.detail) if isinstance(exc, TimeoutError): return ( - f"Failed to connect to MCP server: no response from {_redact_mcp_resource_url(url) or 'the server'} " + "Failed to connect to MCP server: no valid MCP response received from " + f"{_redact_mcp_resource_url(url) or 'the server'} " f"within {timeout_seconds:.0f}s. Check that the LiteLLM proxy can reach this URL " "from its network (DNS, egress rules, firewalls) and that the server answers MCP requests." ) @@ -131,6 +132,11 @@ def _known_connection_error_message(exc: BaseException, url: str | None, timeout return "Failed to connect to MCP server: the connection timed out." if isinstance(exc, httpx.HTTPStatusError): return f"Failed to connect to MCP server: it returned HTTP {exc.response.status_code}." + if isinstance(exc, (httpx.NetworkError, httpx.RemoteProtocolError, ConnectionError)): + return ( + "Failed to connect to MCP server: the connection was interrupted. " + "Check the server and network connection, then retry." + ) if isinstance(exc, ValueError) and str(exc).startswith("Unexpected content type:"): return ( "Failed to connect to MCP server: the endpoint returned an unsupported content type. " @@ -142,6 +148,11 @@ def _known_connection_error_message(exc: BaseException, url: str | None, timeout "Check the MCP endpoint URL and the server's protocol implementation." ) if MCP_AVAILABLE and isinstance(exc, McpError): + if exc.error.code == -32000 and exc.error.message == "Connection closed": + return ( + "Failed to connect to MCP server: the connection was closed before the request completed. " + "Check that the server stays running and returns a complete MCP response, then retry." + ) if exc.error.code == 32600 and exc.error.message == "Session terminated": return ( "Failed to connect to MCP server: the MCP session was terminated. " @@ -159,7 +170,7 @@ if MCP_AVAILABLE: from mcp.shared.exceptions import McpError from mcp.types import Tool as MCPTool - from litellm.experimental_mcp_client.client import MCPClient + from litellm.experimental_mcp_client.client import MCPClient, _as_read_timeout from litellm.llms.litellm_proxy.skills.skill_search import ( DEFAULT_SKILL_SEARCH_TOP_K, ) @@ -1396,10 +1407,18 @@ if MCP_AVAILABLE: except (KeyboardInterrupt, SystemExit, asyncio.CancelledError): raise except BaseException as e: + effective_timeout: Final = ( + min(request.timeout or MCP_CLIENT_TIMEOUT, timeout_seconds) + if any( + isinstance(cause, McpError) and _as_read_timeout(cause) is not None + for cause in iter_exception_tree(e) + ) + else timeout_seconds + ) return { "status": "error", "error": True, - "message": _connection_error_message(e, request.url, timeout_seconds), + "message": _connection_error_message(e, request.url, effective_timeout), } async def _preview_openapi_tools(spec_path: str) -> dict: diff --git a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py index 90292b9b162..a1e781d17b7 100644 --- a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py +++ b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py @@ -3,6 +3,7 @@ import base64 import json import os import sys +from collections.abc import AsyncIterator from importlib import metadata from pathlib import Path from typing import Final @@ -18,6 +19,7 @@ from pydantic import ValidationError from mcp.shared.message import SessionMessage from mcp.types import ( LATEST_PROTOCOL_VERSION, + CallToolResult, ErrorData, Implementation, InitializeResult, @@ -36,6 +38,7 @@ from litellm.experimental_mcp_client.client import ( MCPClient, _as_read_timeout, _first_non_cancelled_cause, + _TransportContext, missing_streamable_http_client_error, strip_auth_scheme, ) @@ -1296,6 +1299,7 @@ def test_a_differently_cased_injected_header_cannot_shadow_the_slot() -> None: [ ("text/html", b"secret-page", ValueError), ("application/json", b"secret-invalid-json", ValidationError), + ("application/json", b"", ValidationError), ("application/json", b'{"secret":"invalid-rpc"}', ValidationError), ("application/json", b'{"jsonrpc":"2.0","id":0,"result":{"secret":"invalid-schema"}}', ValidationError), ], @@ -1446,3 +1450,239 @@ async def test_invalid_tool_list_schema_is_identified_as_an_upstream_response() message: Final = _connection_error_message(caught.value, client.server_url, 30) assert "invalid MCP response" in message assert "secret" not in message + + +class _DiagnosticSSEStream(httpx.AsyncByteStream): + def __init__(self, messages: asyncio.Queue[bytes | Exception | None]) -> None: + self.messages = messages + + async def __aiter__(self) -> AsyncIterator[bytes]: + yield b"event: endpoint\ndata: /messages\n\n" + while True: + message: Final = await self.messages.get() + if message is None: + return + if isinstance(message, Exception): + raise message + yield b"event: message\ndata: " + message + b"\n\n" + + +_DIAGNOSTIC_STDIO_SERVER: Final = """ +import json, sys +mode, failure_method = sys.argv[1:] +for line in sys.stdin: + request = json.loads(line) + if "method" not in request or "id" not in request: + continue + if request["method"] == failure_method: + if mode == "bad-json": + print("secret-invalid-json", flush=True) + continue + if mode == "closed": + sys.exit(0) + if mode == "silent": + print(json.dumps({"jsonrpc": "2.0", "method": "notifications/message", "params": {"level": "info", "data": "Waiting"}}), flush=True) + continue + if request["method"] == "initialize": + result = {"protocolVersion": request["params"]["protocolVersion"], "capabilities": {"tools": {}, "logging": {}}, "serverInfo": {"name": "diagnostic", "version": "1"}} + elif request["method"] == "tools/list": + print(json.dumps({"jsonrpc": "2.0", "method": "notifications/message", "params": {"level": "info", "data": "Listing tools"}}), flush=True) + print(json.dumps({"jsonrpc": "2.0", "id": "unmatched", "result": {}}), flush=True) + print(json.dumps({"jsonrpc": "2.0", "id": "server-ping", "method": "ping"}), flush=True) + result = {"tools": [{"name": "ping", "inputSchema": {"type": "object"}}]} + else: + result = {"content": [{"type": "text", "text": "pong"}], "isError": False} + print(json.dumps({"jsonrpc": "2.0", "id": request["id"], "result": result}), flush=True) +""" + + +def _diagnostic_transport(transport: MCPTransport, mode: str, failure_method: str) -> _TransportContext: + from mcp import StdioServerParameters + from mcp.client.sse import sse_client + from mcp.client.stdio import stdio_client + + if transport == MCPTransport.stdio: + return stdio_client( + StdioServerParameters( + command=sys.executable, args=["-u", "-c", _DIAGNOSTIC_STDIO_SERVER, mode, failure_method] + ) + ) + messages: Final[asyncio.Queue[bytes | Exception | None]] = asyncio.Queue() + + async def respond(request: httpx.Request) -> httpx.Response: + if request.method == "GET": + return httpx.Response( + 200, headers={"Content-Type": "text/event-stream"}, stream=_DiagnosticSSEStream(messages) + ) + payload: Final = json.loads(request.content) + if "method" not in payload or "id" not in payload: + return httpx.Response(202) + if payload["method"] == failure_method and mode != "ok": + if mode == "bad-json": + await messages.put(b"secret-invalid-json") + elif mode == "io-error": + await messages.put(httpx.ReadError("secret-read-error")) + elif mode == "closed": + await messages.put(None) + elif mode == "silent": + await messages.put( + b'{"jsonrpc":"2.0","method":"notifications/message","params":{"level":"info","data":"Waiting"}}' + ) + return httpx.Response(202) + if payload["method"] == "tools/list": + for message in ( + { + "jsonrpc": "2.0", + "method": "notifications/message", + "params": {"level": "info", "data": "Listing tools"}, + }, + {"jsonrpc": "2.0", "id": "unmatched", "result": {}}, + {"jsonrpc": "2.0", "id": "server-ping", "method": "ping"}, + ): + await messages.put(json.dumps(message).encode()) + result: Final = ( + { + "protocolVersion": LATEST_PROTOCOL_VERSION, + "capabilities": {"tools": {}, "logging": {}}, + "serverInfo": {"name": "diagnostic", "version": "1"}, + } + if payload["method"] == "initialize" + else {"tools": [{"name": "ping", "inputSchema": {"type": "object"}}]} + if payload["method"] == "tools/list" + else {"content": [{"type": "text", "text": "pong"}], "isError": False} + ) + await messages.put(json.dumps({"jsonrpc": "2.0", "id": payload["id"], "result": result}).encode()) + return httpx.Response(202) + + def factory( + headers: dict[str, str] | None = None, + timeout: httpx.Timeout | None = None, + auth: httpx.Auth | None = None, + ) -> httpx.AsyncClient: + return httpx.AsyncClient(transport=httpx.MockTransport(respond), headers=headers, timeout=timeout, auth=auth) + + return sse_client("https://example.com/sse", httpx_client_factory=factory) + + +@pytest.mark.asyncio +@pytest.mark.parametrize("transport", [MCPTransport.sse, MCPTransport.stdio]) +@pytest.mark.parametrize("failure_method", ["initialize", "tools/list"]) +async def test_transport_parsing_failure_is_preserved(transport: MCPTransport, failure_method: str) -> None: + client: Final = MCPClient(server_url="https://example.com/sse", transport_type=transport, timeout=0.2) + with pytest.raises(ValidationError): + await asyncio.wait_for( + client._execute_session_operation( + _diagnostic_transport(transport, "bad-json", failure_method), lambda session: session.list_tools() + ), + timeout=3, + ) + + +@pytest.mark.asyncio +async def test_sse_read_failure_is_preserved() -> None: + client: Final = MCPClient(server_url="https://example.com/sse", transport_type=MCPTransport.sse, timeout=0.2) + with pytest.raises(httpx.ReadError, match="secret-read-error"): + await asyncio.wait_for( + client._execute_session_operation( + _diagnostic_transport(MCPTransport.sse, "io-error", "tools/list"), lambda session: session.list_tools() + ), + timeout=3, + ) + + +@pytest.mark.asyncio +@pytest.mark.parametrize("transport", [MCPTransport.sse, MCPTransport.stdio]) +@pytest.mark.parametrize("mode", ["ok", "closed", "silent"]) +async def test_transport_completion_and_normal_messages(transport: MCPTransport, mode: str) -> None: + from mcp import ClientSession + from litellm.proxy._experimental.mcp_server.rest_endpoints import _connection_error_message + + logging_callback: Final = AsyncMock() + client: Final = MCPClient( + server_url="https://example.com/sse", transport_type=transport, timeout=0.2, logging_callback=logging_callback + ) + + async def operation(session: ClientSession) -> CallToolResult: + tools: Final = await session.list_tools() + assert [tool.name for tool in tools.tools] == ["ping"] + return await session.call_tool("ping", {}) + + pending: Final = client._execute_session_operation(_diagnostic_transport(transport, mode, "tools/list"), operation) + if mode == "ok": + result: Final = await asyncio.wait_for(pending, timeout=3) + assert result.isError is False + assert result.content[0].text == "pong" + logging_callback.assert_awaited_once_with(LoggingMessageNotificationParams(level="info", data="Listing tools")) + else: + with pytest.raises(McpError) as caught: + await asyncio.wait_for(pending, timeout=3) + if mode == "closed": + assert "connection was closed" in _connection_error_message(caught.value, client.server_url, 0.2) + else: + assert isinstance(_as_read_timeout(caught.value), TimeoutError) + + +@pytest.mark.asyncio +@pytest.mark.parametrize("transport", [MCPTransport.sse, MCPTransport.stdio]) +async def test_transport_cancellation_cleans_up_a_pending_request(transport: MCPTransport) -> None: + ready: Final = asyncio.Event() + + async def on_log(message: LoggingMessageNotificationParams) -> None: + if message.data == "Waiting": + ready.set() + + client: Final = MCPClient( + server_url="https://example.com/sse", transport_type=transport, timeout=30, logging_callback=on_log + ) + task: Final = asyncio.create_task( + client._execute_session_operation( + _diagnostic_transport(transport, "silent", "tools/list"), lambda session: session.list_tools() + ) + ) + try: + await asyncio.wait_for(ready.wait(), timeout=3) + finally: + task.cancel() + with pytest.raises(asyncio.CancelledError): + await asyncio.wait_for(task, timeout=3) + + +class _InterruptedHTTPBody(httpx.AsyncByteStream): + async def __aiter__(self) -> AsyncIterator[bytes]: + yield b'{"jsonrpc":' + raise httpx.RemoteProtocolError("secret-incomplete-response") + + +@pytest.mark.asyncio +async def test_interrupted_http_response_preserves_the_transport_failure() -> None: + def respond(request: httpx.Request) -> httpx.Response: + return httpx.Response(200, headers={"Content-Type": "application/json"}, stream=_InterruptedHTTPBody()) + + async with httpx.AsyncClient(transport=httpx.MockTransport(respond)) as http_client: + client: Final = MCPClient(server_url="https://example.com/mcp", timeout=30) + with pytest.raises(httpx.RemoteProtocolError, match="secret-incomplete-response"): + await asyncio.wait_for( + client._execute_session_operation( + streamable_http_client(client.server_url, http_client=http_client), + lambda session: session.list_tools(), + ), + timeout=3, + ) + + +@pytest.mark.asyncio +async def test_empty_http_event_stream_uses_the_existing_request_deadline() -> None: + def respond(request: httpx.Request) -> httpx.Response: + return httpx.Response(200, headers={"Content-Type": "text/event-stream"}, content=b"") + + async with httpx.AsyncClient(transport=httpx.MockTransport(respond)) as http_client: + client: Final = MCPClient(server_url="https://example.com/mcp", timeout=0.2) + with pytest.raises(McpError) as caught: + await asyncio.wait_for( + client._execute_session_operation( + streamable_http_client(client.server_url, http_client=http_client), + lambda session: session.list_tools(), + ), + timeout=3, + ) + assert isinstance(_as_read_timeout(caught.value), TimeoutError) diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py index 73734131130..101b02e6ffd 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py @@ -3427,6 +3427,54 @@ class TestConnectionErrorMessage: message = rest_endpoints._connection_error_message(exc, "https://example.com", 30.0) assert "503" in message + @pytest.mark.parametrize( + "error_type", [httpx.ReadError, httpx.WriteError, httpx.RemoteProtocolError, ConnectionResetError] + ) + def test_interrupted_connection_message_is_safe(self, error_type: type[Exception]) -> None: + message: Final = rest_endpoints._connection_error_message( + error_type("secret-transport-detail"), "https://example.com/?token=secret-query", 30 + ) + assert "connection was interrupted" in message + assert "secret" not in message + + def test_closed_connection_explains_incomplete_request(self) -> None: + from mcp import McpError + from mcp.types import ErrorData + + message: Final = rest_endpoints._connection_error_message( + McpError(ErrorData(code=-32000, message="Connection closed", data="secret-data")), None, 30 + ) + assert "connection was closed before the request completed" in message + assert "secret" not in message + + def test_timeout_does_not_claim_the_server_sent_nothing(self) -> None: + message: Final = rest_endpoints._connection_error_message(TimeoutError(), None, 30) + assert "no valid MCP response received" in message + + @pytest.mark.asyncio + @pytest.mark.parametrize("sdk_timeout", [True, False]) + async def test_timeout_message_uses_the_deadline_that_expired(self, sdk_timeout: bool) -> None: + from mcp import McpError + from mcp.types import ErrorData + + async def operation(client: rest_endpoints.MCPClient) -> dict[str, object]: + try: + raise TimeoutError("secret-timeout") + except TimeoutError as elapsed: + if not sdk_timeout: + raise + try: + raise McpError(ErrorData(code=408, message="secret-sdk-timeout")) from elapsed + except McpError as sdk_error: + raise TimeoutError() from sdk_error + + payload: Final = NewMCPServerRequest( + server_name="timeout", url="https://example.com", auth_type=MCPAuth.none, timeout=1 + ) + result: Final = await rest_endpoints._execute_with_mcp_client(payload, operation, timeout_seconds=30) + assert ("within 1s" if sdk_timeout else "within 30s") in result["message"] + assert "secret" not in result["message"] + def test_unknown_error_falls_back_to_generic(self): message = rest_endpoints._connection_error_message(RuntimeError("weird"), "https://example.com", 30.0) assert "weird" not in message From 9b6c7c8bf0e2e0e3218caa8a18b29b17bb8bf566 Mon Sep 17 00:00:00 2001 From: Mateo Wang <277851410+mateo-berri@users.noreply.github.com> Date: Wed, 9 Sep 2026 09:17:14 -0700 Subject: [PATCH 16/26] Apply suggestion from @greptile-apps[bot] Co-authored-by: greptile-apps[bot] <165735046+greptile-apps[bot]@users.noreply.github.com> --- litellm/proxy/_experimental/mcp_server/rest_endpoints.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py index 63badc6f703..31252479445 100644 --- a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py +++ b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py @@ -1408,7 +1408,7 @@ if MCP_AVAILABLE: raise except BaseException as e: effective_timeout: Final = ( - min(request.timeout or MCP_CLIENT_TIMEOUT, timeout_seconds) + min(request.timeout if request.timeout is not None else MCP_CLIENT_TIMEOUT, timeout_seconds) if any( isinstance(cause, McpError) and _as_read_timeout(cause) is not None for cause in iter_exception_tree(e) From dbf94902294cf5fa4fe5af6681fcd83ffd743231 Mon Sep 17 00:00:00 2001 From: Joshua Valluru <326636767+joshua-berri@users.noreply.github.com> Date: Wed, 9 Sep 2026 09:44:14 -0700 Subject: [PATCH 17/26] fix(mcp): expose shared SDK timeout normalization --- litellm/experimental_mcp_client/client.py | 6 +++--- .../_experimental/mcp_server/rest_endpoints.py | 4 ++-- .../experimental_mcp_client/test_mcp_client.py | 18 +++++++++--------- .../mcp_server/test_rest_endpoints.py | 7 ++++--- 4 files changed, 18 insertions(+), 17 deletions(-) diff --git a/litellm/experimental_mcp_client/client.py b/litellm/experimental_mcp_client/client.py index 5cdc4efdfef..8c7f4557992 100644 --- a/litellm/experimental_mcp_client/client.py +++ b/litellm/experimental_mcp_client/client.py @@ -150,8 +150,8 @@ _SDK_READ_TIMEOUT_CODE: Final = int(httpx.codes.REQUEST_TIMEOUT) otherwise carries JSON-RPC error codes.""" -def _as_read_timeout(exc: BaseException) -> TimeoutError | None: - """The session read timeout elapsing, re-expressed as a ``TimeoutError``, or ``None``. +def as_mcp_read_timeout(exc: BaseException) -> TimeoutError | None: + """Normalize an MCP SDK read timeout for client and gateway diagnostics, or return ``None``. The SDK reports its own elapsed read timeout as ``McpError`` carrying an HTTP status code in a field that otherwise holds JSON-RPC error codes, and it relays an upstream's JSON-RPC error @@ -522,7 +522,7 @@ class MCPClient: transport_ctx, http_client = self._create_transport_context() return await self._execute_session_operation(transport_ctx, operation) except Exception as e: - read_timeout: Final = _as_read_timeout(e) + read_timeout: Final = as_mcp_read_timeout(e) if read_timeout is not None: verbose_logger.warning( "MCP client timed out after %ss waiting for a valid MCP response from %s", diff --git a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py index 31252479445..5b74caee3a2 100644 --- a/litellm/proxy/_experimental/mcp_server/rest_endpoints.py +++ b/litellm/proxy/_experimental/mcp_server/rest_endpoints.py @@ -170,7 +170,7 @@ if MCP_AVAILABLE: from mcp.shared.exceptions import McpError from mcp.types import Tool as MCPTool - from litellm.experimental_mcp_client.client import MCPClient, _as_read_timeout + from litellm.experimental_mcp_client.client import MCPClient, as_mcp_read_timeout from litellm.llms.litellm_proxy.skills.skill_search import ( DEFAULT_SKILL_SEARCH_TOP_K, ) @@ -1410,7 +1410,7 @@ if MCP_AVAILABLE: effective_timeout: Final = ( min(request.timeout if request.timeout is not None else MCP_CLIENT_TIMEOUT, timeout_seconds) if any( - isinstance(cause, McpError) and _as_read_timeout(cause) is not None + isinstance(cause, McpError) and as_mcp_read_timeout(cause) is not None for cause in iter_exception_tree(e) ) else timeout_seconds diff --git a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py index a1e781d17b7..b07e5876e8b 100644 --- a/tests/test_litellm/experimental_mcp_client/test_mcp_client.py +++ b/tests/test_litellm/experimental_mcp_client/test_mcp_client.py @@ -36,9 +36,9 @@ import litellm.experimental_mcp_client.client as mcp_client_module from litellm.experimental_mcp_client.client import ( MCP_STREAMABLE_HTTP_REQUIREMENT, MCPClient, - _as_read_timeout, _first_non_cancelled_cause, _TransportContext, + as_mcp_read_timeout, missing_streamable_http_client_error, strip_auth_scheme, ) @@ -867,25 +867,25 @@ def _raise_mcp_error_while_handling_a_timeout(code: int, message: str) -> McpErr return raised -def test_as_read_timeout_separates_the_sdk_timeout_from_a_relayed_upstream_error(): +def test_as_mcp_read_timeout_separates_the_sdk_timeout_from_a_relayed_upstream_error(): """Neither signal alone is enough. The code alone cannot separate the SDK's own timeout from an upstream JSON-RPC error that happens to use 408, and the context chain alone cannot separate it from any other relayed error that surfaces while a timeout is being handled, so both must hold. """ timeout_code = int(httpx.codes.REQUEST_TIMEOUT) - translated = _as_read_timeout(_raise_mcp_error_while_handling_a_timeout(timeout_code, "Timed out while waiting")) + translated = as_mcp_read_timeout(_raise_mcp_error_while_handling_a_timeout(timeout_code, "Timed out while waiting")) assert isinstance(translated, TimeoutError) assert str(translated) == "Timed out while waiting" relayed_408 = McpError(ErrorData(code=timeout_code, message="upstream said 408")) - assert _as_read_timeout(relayed_408) is None, "an upstream 408 with no elapsed timeout is not our timeout" + assert as_mcp_read_timeout(relayed_408) is None, "an upstream 408 with no elapsed timeout is not our timeout" relayed_other = _raise_mcp_error_while_handling_a_timeout(-32603, "upstream internal error") - assert _as_read_timeout(relayed_other) is None, "a non-timeout code is not our timeout, whatever the chain" + assert as_mcp_read_timeout(relayed_other) is None, "a non-timeout code is not our timeout, whatever the chain" - assert _as_read_timeout(McpError(ErrorData(code=-32603, message="boom"))) is None - assert _as_read_timeout(RuntimeError("not an McpError")) is None + assert as_mcp_read_timeout(McpError(ErrorData(code=-32603, message="boom"))) is None + assert as_mcp_read_timeout(RuntimeError("not an McpError")) is None @pytest.mark.asyncio @@ -1619,7 +1619,7 @@ async def test_transport_completion_and_normal_messages(transport: MCPTransport, if mode == "closed": assert "connection was closed" in _connection_error_message(caught.value, client.server_url, 0.2) else: - assert isinstance(_as_read_timeout(caught.value), TimeoutError) + assert isinstance(as_mcp_read_timeout(caught.value), TimeoutError) @pytest.mark.asyncio @@ -1685,4 +1685,4 @@ async def test_empty_http_event_stream_uses_the_existing_request_deadline() -> N ), timeout=3, ) - assert isinstance(_as_read_timeout(caught.value), TimeoutError) + assert isinstance(as_mcp_read_timeout(caught.value), TimeoutError) diff --git a/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py b/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py index 101b02e6ffd..3487f634251 100644 --- a/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py +++ b/tests/test_litellm/proxy/_experimental/mcp_server/test_rest_endpoints.py @@ -3453,7 +3453,8 @@ class TestConnectionErrorMessage: @pytest.mark.asyncio @pytest.mark.parametrize("sdk_timeout", [True, False]) - async def test_timeout_message_uses_the_deadline_that_expired(self, sdk_timeout: bool) -> None: + @pytest.mark.parametrize("read_timeout", [0, 1]) + async def test_timeout_message_uses_the_deadline_that_expired(self, sdk_timeout: bool, read_timeout: int) -> None: from mcp import McpError from mcp.types import ErrorData @@ -3469,10 +3470,10 @@ class TestConnectionErrorMessage: raise TimeoutError() from sdk_error payload: Final = NewMCPServerRequest( - server_name="timeout", url="https://example.com", auth_type=MCPAuth.none, timeout=1 + server_name="timeout", url="https://example.com", auth_type=MCPAuth.none, timeout=read_timeout ) result: Final = await rest_endpoints._execute_with_mcp_client(payload, operation, timeout_seconds=30) - assert ("within 1s" if sdk_timeout else "within 30s") in result["message"] + assert (f"within {read_timeout}s" if sdk_timeout else "within 30s") in result["message"] assert "secret" not in result["message"] def test_unknown_error_falls_back_to_generic(self): From 1183b2abc6645a92f9a852daa7843662b6fa6f20 Mon Sep 17 00:00:00 2001 From: yujonglee Date: Wed, 9 Sep 2026 09:46:37 -0700 Subject: [PATCH 18/26] fix(integrations): pass original request object to post-call guardrail hooks (#40414) --- litellm/integrations/custom_guardrail.py | 30 ++++--- .../integrations/test_custom_guardrail.py | 90 ++++++++++++++++++- 2 files changed, 106 insertions(+), 14 deletions(-) diff --git a/litellm/integrations/custom_guardrail.py b/litellm/integrations/custom_guardrail.py index 37d6a7e793d..77bf4820a1a 100644 --- a/litellm/integrations/custom_guardrail.py +++ b/litellm/integrations/custom_guardrail.py @@ -850,20 +850,24 @@ class CustomGuardrail(CustomLogger): if self.should_run_guardrail(data=request_data, event_type=GuardrailEventHooks.post_call) is not True: return None - # CHECK IF GUARDRAIL REJECTS THE REQUEST target: Final = self._deployment_hook_target() - hook_request_data: Final = {**request_data, "guardrail_to_apply": self} if target is not self else request_data - result: Final = await target.async_post_call_success_hook( - user_api_key_dict=UserAPIKeyAuth( - user_id=request_data.get("user_api_key_user_id"), - team_id=request_data.get("user_api_key_team_id"), - end_user_id=request_data.get("user_api_key_end_user_id"), - api_key=request_data.get("user_api_key_hash"), - request_route=request_data.get("user_api_key_request_route"), - ), - data=hook_request_data, - response=response, - ) + try: + if target is not self: + request_data["guardrail_to_apply"] = self # rebind-ok: dispatch consumes this key + result: Final = await target.async_post_call_success_hook( + user_api_key_dict=UserAPIKeyAuth( + user_id=request_data.get("user_api_key_user_id"), + team_id=request_data.get("user_api_key_team_id"), + end_user_id=request_data.get("user_api_key_end_user_id"), + api_key=request_data.get("user_api_key_hash"), + request_route=request_data.get("user_api_key_request_route"), + ), + data=request_data, + response=response, + ) + finally: + if target is not self: + request_data.pop("guardrail_to_apply", None) if not self._is_valid_response_type(result): return None diff --git a/tests/test_litellm/integrations/test_custom_guardrail.py b/tests/test_litellm/integrations/test_custom_guardrail.py index cd8d609cf71..ddc8439a83a 100644 --- a/tests/test_litellm/integrations/test_custom_guardrail.py +++ b/tests/test_litellm/integrations/test_custom_guardrail.py @@ -1842,12 +1842,14 @@ class _ApplyStyleGuardrail(CustomGuardrail): self.block = block self.apply_called = False self.seen_texts = None + self.seen_request_data = None async def apply_guardrail(self, inputs, request_data, input_type, logging_obj=None): from fastapi import HTTPException self.apply_called = True self.seen_texts = inputs.get("texts") + self.seen_request_data = request_data if self.block: raise HTTPException(status_code=400, detail={"error": "Violated moderation policy"}) return inputs @@ -2646,6 +2648,91 @@ class TestCustomGuardrailPostCallSuccessDeploymentHook: which starved every later callback in litellm.callbacks (notably the lazily-appended VectorStorePreCallHook that attaches provider_specific_fields["search_results"]).""" + @pytest.mark.asyncio + async def test_apply_guardrail_retains_request_identity(self) -> None: + from litellm.types.guardrails import GuardrailEventHooks + from litellm.types.utils import Choices, Message, ModelResponse + + guardrail: Final = _ApplyStyleGuardrail(block=False) + guardrail.event_hook = GuardrailEventHooks.post_call + request_data: Final = {"guardrails": ["apply-style-guardrail"]} + response: Final = ModelResponse(choices=[Choices(message=Message(content="review me"))]) + + await guardrail.async_post_call_success_deployment_hook( + request_data=request_data, response=response, call_type=CallTypes.acompletion + ) + + assert guardrail.seen_request_data is request_data + assert guardrail.seen_texts == ["review me"] + assert "guardrail_to_apply" not in request_data + + @pytest.mark.asyncio + @pytest.mark.parametrize("call_type", (None, CallTypes.acompletion)) + async def test_apply_guardrail_masks_response_and_records_metadata(self, call_type: CallTypes | None) -> None: + from litellm.proxy.guardrails.guardrail_hooks.litellm_content_filter.content_filter import ( + ContentFilterGuardrail, + ) + from litellm.types.guardrails import BlockedWord, ContentFilterAction, GuardrailEventHooks + from litellm.types.utils import Choices, Message, ModelResponse + + guardrail: Final = ContentFilterGuardrail( + guardrail_name="response-filter", + event_hook=GuardrailEventHooks.post_call, + blocked_words=[BlockedWord(keyword="secret", action=ContentFilterAction.MASK)], + ) + request_data: Final = {"guardrails": ["response-filter"]} + response: Final = ModelResponse(choices=[Choices(message=Message(content="a secret"))]) + + result: Final = await guardrail.async_post_call_success_deployment_hook( + request_data=request_data, response=response, call_type=call_type + ) + + assert isinstance(result, ModelResponse) + assert result.choices[0].message.content == f"a {guardrail.keyword_redaction_tag}" + entries: Final = _guardrail_entries(request_data) + assert len(entries) == 1 + assert entries[0]["guardrail_name"] == "response-filter" + assert entries[0]["guardrail_mode"] == "post_call" + assert "guardrail_to_apply" not in request_data + + @pytest.mark.asyncio + @pytest.mark.parametrize("error_type", (None, RuntimeError, asyncio.CancelledError)) + async def test_dispatch_cleans_up_request_on_every_exit(self, error_type: type[BaseException] | None) -> None: + from contextlib import nullcontext + + from litellm.integrations.custom_logger import CustomLogger + from litellm.types.utils import LLMResponseTypes, ModelResponse + + error: Final = error_type("dispatch interrupted") if error_type is not None else None + + class Dispatch(CustomLogger): + request_data: dict[str, object] | None = None + + async def async_post_call_success_hook( + self, data: dict[str, object], user_api_key_dict: UserAPIKeyAuth, response: LLMResponseTypes + ) -> LLMResponseTypes: + self.request_data = data + if error is not None: + raise error + return response + + dispatch: Final = Dispatch() + + class Guardrail(_ApplyStyleGuardrail): + def _deployment_hook_target(self) -> CustomLogger: + return dispatch + + guardrail: Final = Guardrail(block=False) + guardrail.event_hook = GuardrailEventHooks.post_call + request_data: Final = {"guardrails": ["apply-style-guardrail"]} + with pytest.raises(error_type) if error_type is not None else nullcontext(): + await guardrail.async_post_call_success_deployment_hook( + request_data=request_data, response=ModelResponse(), call_type=CallTypes.acompletion + ) + + assert dispatch.request_data is request_data + assert "guardrail_to_apply" not in request_data + @pytest.mark.asyncio async def test_returns_none_when_request_has_no_guardrails(self): from litellm.types.utils import ModelResponse @@ -2740,4 +2827,5 @@ class TestCustomGuardrailPostCallSuccessDeploymentHook: assert result is response assert response.choices[0].message.content == "filtered response" - assert request_data == {"guardrails": ["test-guardrail"]} + assert "guardrail_to_apply" not in request_data + assert len(_guardrail_entries(request_data)) == 1 From 36f3ca95d8bdb6b03ab4632982cdc7c0cef99c6f Mon Sep 17 00:00:00 2001 From: ryan-crabbe-berri Date: Wed, 9 Sep 2026 09:57:38 -0700 Subject: [PATCH 19/26] fix(proxy): report /cost/estimate rates from the call that billed them The estimate looked the reported per-token rates up a second time, with the provider this endpoint resolved rather than the one completion_cost infers. The provider decides whether a token tier threshold is inclusive, so an unrouted xai model sitting exactly on 200k billed at the tier rate and reported the base rate, half of it. completion_cost now hands back the rates its own lines were billed at, and the endpoint reports those. Claude-Session: https://claude.ai/code/session_01RLKy5DMi3XCBUJ37WzfNi1 --- litellm/cost_calculator.py | 6 ++ litellm/litellm_core_utils/litellm_logging.py | 5 ++ .../litellm_core_utils/llm_cost_calc/utils.py | 60 ++++++++++--------- .../cost_tracking_settings.py | 29 ++++----- .../llm_cost_calc/test_llm_cost_calc_utils.py | 52 +++++++++++++--- .../test_cost_tracking_settings.py | 39 ++++++++++++ tests/test_litellm/test_cost_calculator.py | 56 +++++++++++++++++ 7 files changed, 195 insertions(+), 52 deletions(-) diff --git a/litellm/cost_calculator.py b/litellm/cost_calculator.py index a1181ee4124..b4be023540f 100644 --- a/litellm/cost_calculator.py +++ b/litellm/cost_calculator.py @@ -25,6 +25,7 @@ from litellm.litellm_core_utils.llm_cost_calc.usage_object_transformation import TranscriptionUsageObjectTransformation, ) from litellm.litellm_core_utils.llm_cost_calc.utils import ( + BilledTokenRates, CostCalculatorUtils, _generic_cost_per_character, _get_regional_uplift_multiplier, @@ -1122,6 +1123,7 @@ def _store_cost_breakdown_in_logging_obj( service_tier: str | None = None, data_residency: str | None = None, vertex_location: str | None = None, + billed_token_rates: BilledTokenRates | None = None, ) -> None: """ Helper function to store cost breakdown in the logging object. @@ -1166,6 +1168,7 @@ def _store_cost_breakdown_in_logging_obj( service_tier=service_tier, data_residency=data_residency, vertex_location=vertex_location, + billed_token_rates=billed_token_rates, ) except Exception as breakdown_error: @@ -1735,6 +1738,7 @@ def completion_cost( _reasoning_cost: float | None = None _cache_read_cost: float | None = None _cache_creation_cost: float | None = None + _billed_token_rates: BilledTokenRates | None = None if cost_per_token_usage_object is not None and model: _breakdown_provider: str | None = ( custom_llm_provider if isinstance(custom_llm_provider, str) else None @@ -1751,6 +1755,7 @@ def completion_cost( _reasoning_cost = _token_type_breakdown.reasoning_cost _cache_read_cost = _token_type_breakdown.cache_read_cost _cache_creation_cost = _token_type_breakdown.cache_creation_cost + _billed_token_rates = _token_type_breakdown.rates _store_cost_breakdown_in_logging_obj( litellm_logging_obj=litellm_logging_obj, prompt_tokens_cost_usd_dollar=prompt_tokens_cost_usd_dollar, @@ -1770,6 +1775,7 @@ def completion_cost( service_tier=service_tier, data_residency=data_residency, vertex_location=vertex_location, + billed_token_rates=_billed_token_rates, ) return _final_cost diff --git a/litellm/litellm_core_utils/litellm_logging.py b/litellm/litellm_core_utils/litellm_logging.py index ca2cca5360f..a9f3c0e9325 100644 --- a/litellm/litellm_core_utils/litellm_logging.py +++ b/litellm/litellm_core_utils/litellm_logging.py @@ -201,6 +201,7 @@ if TYPE_CHECKING: from mcp.types import EmbeddedResource, ImageContent, TextContent from litellm.integrations.otel.logger import OpenTelemetryV2 + from litellm.litellm_core_utils.llm_cost_calc.utils import BilledTokenRates from litellm.llms.base_llm.passthrough.transformation import BasePassthroughConfig try: from litellm_enterprise.enterprise_callbacks.callback_controls import ( @@ -581,6 +582,7 @@ class Logging(LiteLLMLoggingBaseClass): # Initialize cost breakdown field self.cost_breakdown: CostBreakdown | None = None + self.billed_token_rates: BilledTokenRates | None = None # Init Caching related details self.caching_details: CachingDetails | None = None @@ -1585,6 +1587,7 @@ class Logging(LiteLLMLoggingBaseClass): service_tier: str | None = None, data_residency: str | None = None, vertex_location: str | None = None, + billed_token_rates: "BilledTokenRates | None" = None, ) -> None: """ Helper method to store cost breakdown in the logging object. @@ -1604,8 +1607,10 @@ class Logging(LiteLLMLoggingBaseClass): service_tier: Tier the costs above were priced on, already resolved data_residency: Region uplift the costs above were priced on, already resolved vertex_location: Vertex AI location the costs above were priced on, already resolved + billed_token_rates: Per-token rates the costs above were billed at, already resolved """ + self.billed_token_rates = billed_token_rates self.cost_breakdown = CostBreakdown( input_cost=input_cost, output_cost=output_cost, diff --git a/litellm/litellm_core_utils/llm_cost_calc/utils.py b/litellm/litellm_core_utils/llm_cost_calc/utils.py index e2168528e6b..42f88bbb425 100644 --- a/litellm/litellm_core_utils/llm_cost_calc/utils.py +++ b/litellm/litellm_core_utils/llm_cost_calc/utils.py @@ -1302,34 +1302,6 @@ def _coerce_token_count(value: object) -> int: return value if isinstance(value, int) and value > 0 else 0 -@dataclass(frozen=True, slots=True) -class TokenTypeCostBreakdown: - reasoning_cost: float - cache_read_cost: float - cache_creation_cost: float - - -def _reasoning_token_count(usage: Usage) -> int: - parsed: Final = ( - parse_completion_tokens_details(usage)["reasoning_tokens"] if usage.completion_tokens_details is not None else 0 - ) - return parsed or _coerce_token_count(getattr(usage, "reasoning_tokens", 0)) - - -def _cache_token_counts(usage: Usage) -> tuple[int, int, CacheCreationTokenDetails | None]: - """(cache read tokens, cache creation tokens, cache creation details): read from prompt_tokens_details - first, then the private top-level counters the Usage constructor mirrors cache tokens onto for - providers/callers that bypass the details.""" - parsed: Final = parse_prompt_tokens_details(usage) if usage.prompt_tokens_details is not None else None - parsed_read: Final = parsed["cache_hit_tokens"] if parsed is not None else 0 - parsed_creation: Final = parsed["cache_creation_tokens"] if parsed is not None else 0 - return ( - parsed_read or _coerce_token_count(getattr(usage, "_cache_read_input_tokens", 0)), - parsed_creation or _coerce_token_count(getattr(usage, "_cache_creation_input_tokens", 0)), - parsed["cache_creation_token_details"] if parsed is not None else None, - ) - - @dataclass(frozen=True, slots=True) class BilledTokenRates: """Per-token rates one request's usage bills at, after token tiers, off-peak windows and the @@ -1355,6 +1327,37 @@ class BilledTokenRates: ) +@dataclass(frozen=True, slots=True) +class TokenTypeCostBreakdown: + reasoning_cost: float + cache_read_cost: float + cache_creation_cost: float + rates: BilledTokenRates | None = None + """Rates these lines were billed at, so a caller reporting both cannot resolve them a second, + differently-argued way. None when the model's pricing could not be resolved.""" + + +def _reasoning_token_count(usage: Usage) -> int: + parsed: Final = ( + parse_completion_tokens_details(usage)["reasoning_tokens"] if usage.completion_tokens_details is not None else 0 + ) + return parsed or _coerce_token_count(getattr(usage, "reasoning_tokens", 0)) + + +def _cache_token_counts(usage: Usage) -> tuple[int, int, CacheCreationTokenDetails | None]: + """(cache read tokens, cache creation tokens, cache creation details): read from prompt_tokens_details + first, then the private top-level counters the Usage constructor mirrors cache tokens onto for + providers/callers that bypass the details.""" + parsed: Final = parse_prompt_tokens_details(usage) if usage.prompt_tokens_details is not None else None + parsed_read: Final = parsed["cache_hit_tokens"] if parsed is not None else 0 + parsed_creation: Final = parsed["cache_creation_tokens"] if parsed is not None else 0 + return ( + parsed_read or _coerce_token_count(getattr(usage, "_cache_read_input_tokens", 0)), + parsed_creation or _coerce_token_count(getattr(usage, "_cache_creation_input_tokens", 0)), + parsed["cache_creation_token_details"] if parsed is not None else None, + ) + + def _custom_pricing_rates(custom_cost_per_token: CostPerToken) -> BilledTokenRates: """Flat custom pricing has no tiers, uplifts or reasoning rate: cache tokens bill at the configured cache rates (else the input rate) and reasoning at the output rate, as _cost_per_token_custom_pricing_helper does.""" @@ -1497,6 +1500,7 @@ def get_token_type_cost_breakdown( reasoning_cost=float(_reasoning_token_count(usage)) * rates.output_cost_per_reasoning_token, cache_read_cost=float(cache_read_tokens) * rates.cache_read_input_token_cost, cache_creation_cost=cache_creation_cost, + rates=rates, ) diff --git a/litellm/proxy/management_endpoints/cost_tracking_settings.py b/litellm/proxy/management_endpoints/cost_tracking_settings.py index 1faa66584d5..493f75008c3 100644 --- a/litellm/proxy/management_endpoints/cost_tracking_settings.py +++ b/litellm/proxy/management_endpoints/cost_tracking_settings.py @@ -21,7 +21,6 @@ import litellm from litellm._internal_context import current_billing_time, pinned_billing_time from litellm._logging import verbose_proxy_logger from litellm.cost_calculator import completion_cost -from litellm.litellm_core_utils.llm_cost_calc.utils import get_billed_token_rates from litellm.proxy._types import ( CommonProxyErrors, CostEstimateRequest, @@ -85,9 +84,9 @@ def _extract_custom_pricing( ) -def _lookup_model_info(model: str) -> ModelInfo | None: +def _lookup_model_info(model: str, custom_llm_provider: str | None = None) -> ModelInfo | None: try: - return litellm.get_model_info(model=model) + return litellm.get_model_info(model=model, custom_llm_provider=custom_llm_provider) except Exception: return None @@ -122,7 +121,7 @@ def _resolve_model_for_cost_lookup(model: str) -> ResolvedCostModel: if resolved_model: verbose_proxy_logger.debug("Resolved model '%s' to '%s' from router", model, resolved_model) custom_cost_per_token: Final = _extract_custom_pricing( - litellm_params, model_info, _lookup_model_info(str(resolved_model)) + litellm_params, model_info, _lookup_model_info(str(resolved_model), provider) ) return ResolvedCostModel(str(resolved_model), provider, custom_cost_per_token) except Exception as e: @@ -632,10 +631,9 @@ async def estimate_cost( function_id="cost-estimate", ) - # The totals, the per-token-type lines and the reported rates each resolve pricing on their - # own path. Pinning one moment keeps an off-peak window that opens mid-quote from splitting them. - billed_at: Final = current_billing_time() - with pinned_billing_time(billed_at): + # Pinning one moment keeps an off-peak window that opens mid-quote from pricing the totals on + # one side of it and the reported rates on the other. + with pinned_billing_time(current_billing_time()): # Use completion_cost which handles all the logic including margins/discounts try: cost_per_request: Final = completion_cost( @@ -653,19 +651,16 @@ async def estimate_cost( }, ) - rates: Final = get_billed_token_rates( - model=resolved_model, - custom_llm_provider=resolved_provider, - usage=usage, - custom_cost_per_token=resolved.custom_cost_per_token, - current_time=billed_at, - ) - + # The rates come back from the pricing call itself rather than a second lookup, so they are the + # ones the cost lines above billed at even when completion_cost infers a provider this endpoint + # never resolved (an unrouted "xai/grok-4" prices on xai's inclusive tier thresholds; a lookup + # here without that provider would report the sub-200k rate for a line billed above it). + rates: Final = litellm_logging_obj.billed_token_rates per_request: Final = _cost_lines(cost_per_request, litellm_logging_obj.cost_breakdown) daily: Final = per_request.times(request.num_requests_per_day) monthly: Final = per_request.times(request.num_requests_per_month) - model_info: Final = _lookup_model_info(resolved_model) + model_info: Final = _lookup_model_info(resolved_model, resolved_provider) mapped_provider: Final = model_info.get("litellm_provider") if model_info is not None else None custom_llm_provider: Final = mapped_provider if mapped_provider is not None else resolved_provider diff --git a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py index 90178428018..5a19ed0277a 100644 --- a/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py +++ b/tests/test_litellm/litellm_core_utils/llm_cost_calc/test_llm_cost_calc_utils.py @@ -33,7 +33,6 @@ from litellm.litellm_core_utils.llm_cost_calc.utils import ( CostCalculatorUtils, PromptTokensDetailsResult, TokenRates, - TokenTypeCostBreakdown, _calculate_input_cost, _get_token_base_cost, _is_off_peak, @@ -4023,6 +4022,49 @@ def test_a_pinned_billing_time_prices_the_totals_and_the_reported_rates_at_one_m assert peak_completion_cost == pytest.approx(500 * peak_rates.output_cost_per_token) +def test_the_token_type_breakdown_carries_the_rates_it_billed_at(monkeypatch): + """Callers that report both the lines and the rates read the rates off the breakdown rather than + resolving them a second time, so the breakdown has to hand back exactly what it billed at.""" + monkeypatch.setitem( + litellm.model_cost, + "xai/tiered-model", + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "cache_read_input_token_cost": 3e-7, + "input_cost_per_token_above_200k_tokens": 6e-6, + "output_cost_per_token_above_200k_tokens": 3e-5, + "cache_read_input_token_cost_above_200k_tokens": 6e-7, + "litellm_provider": "xai", + "mode": "chat", + }, + ) + usage = Usage( + prompt_tokens=200_000, + completion_tokens=1_000, + total_tokens=201_000, + prompt_tokens_details=PromptTokensDetailsWrapper(cached_tokens=100_000), + ) + + breakdown = get_token_type_cost_breakdown(model="xai/tiered-model", custom_llm_provider="xai", usage=usage) + + assert breakdown.rates == get_billed_token_rates( + model="xai/tiered-model", custom_llm_provider="xai", usage=usage + ) + assert breakdown.rates.cache_read_input_token_cost == pytest.approx(6e-7) + assert breakdown.cache_read_cost == pytest.approx(100_000 * breakdown.rates.cache_read_input_token_cost) + + +def test_the_token_type_breakdown_reports_no_rates_for_an_unpriced_model(): + usage = Usage(prompt_tokens=10, completion_tokens=5, total_tokens=15) + + breakdown = get_token_type_cost_breakdown( + model="no-such-model-anywhere", custom_llm_provider="openai", usage=usage + ) + + assert breakdown.rates is None + + def test_billed_token_rates_are_none_for_an_unpriced_model(): usage = Usage(prompt_tokens=10, completion_tokens=5, total_tokens=15) @@ -4036,9 +4078,7 @@ def test_token_type_cost_breakdown_zero_without_special_tokens(_local_model_cost model="gpt-4o", custom_llm_provider="openai", usage=usage ) - assert breakdown == TokenTypeCostBreakdown( - reasoning_cost=0.0, cache_read_cost=0.0, cache_creation_cost=0.0 - ) + assert (breakdown.reasoning_cost, breakdown.cache_read_cost, breakdown.cache_creation_cost) == (0.0, 0.0, 0.0) @pytest.mark.parametrize( @@ -4110,9 +4150,7 @@ def test_token_type_cost_breakdown_handles_unknown_model_gracefully(): completion_tokens_details=CompletionTokensDetailsWrapper(reasoning_tokens=5), ), ) - assert breakdown == TokenTypeCostBreakdown( - reasoning_cost=0.0, cache_read_cost=0.0, cache_creation_cost=0.0 - ) + assert (breakdown.reasoning_cost, breakdown.cache_read_cost, breakdown.cache_creation_cost) == (0.0, 0.0, 0.0) def test_token_type_cost_breakdown_applies_regional_uplift(_local_model_cost_map): diff --git a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py index e9485f3a044..7ece35ceedf 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py +++ b/tests/test_litellm/proxy/management_endpoints/test_cost_tracking_settings.py @@ -1139,6 +1139,45 @@ class TestEstimateCostCacheAndReasoningTokens: assert response.output_cost_per_request == pytest.approx(OUTPUT_TOKENS * response.output_cost_per_token) + @pytest.mark.asyncio + async def test_an_unrouted_model_reports_the_rates_of_the_provider_the_calculator_inferred(self, monkeypatch): + """The cost calculator infers a provider this endpoint never resolved, and the provider decides + whether a tier threshold is inclusive. xai bills a request sitting exactly on the 200k threshold + at the tier rate, so the reported rates have to be the tier's rather than the sub-tier base.""" + an_xai_model = "xai/tiered-model" + monkeypatch.setitem( + litellm.model_cost, + an_xai_model, + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "cache_read_input_token_cost": 3e-7, + "input_cost_per_token_above_200k_tokens": 6e-6, + "output_cost_per_token_above_200k_tokens": 3e-5, + "cache_read_input_token_cost_above_200k_tokens": 6e-7, + "litellm_provider": "xai", + "mode": "chat", + }, + ) + + response = await _estimate( + None, + model=an_xai_model, + input_tokens=200_000, + cache_read_input_tokens=100_000, + output_tokens=1_000, + ) + + assert response.input_cost_per_token == pytest.approx(6e-6) + assert response.output_cost_per_token == pytest.approx(3e-5) + assert response.cache_read_input_token_cost == pytest.approx(6e-7) + assert response.cache_read_cost_per_request == pytest.approx(100_000 * response.cache_read_input_token_cost) + assert response.input_cost_per_request == pytest.approx( + 100_000 * response.input_cost_per_token + response.cache_read_cost_per_request + ) + assert response.output_cost_per_request == pytest.approx(1_000 * response.output_cost_per_token) + + class TestCostEstimateRequestTokenSubsets: def test_cache_tokens_beyond_the_input_tokens_are_rejected(self): with pytest.raises(ValidationError, match="cannot exceed input_tokens"): diff --git a/tests/test_litellm/test_cost_calculator.py b/tests/test_litellm/test_cost_calculator.py index 910587d3dc5..8f8a7640c08 100644 --- a/tests/test_litellm/test_cost_calculator.py +++ b/tests/test_litellm/test_cost_calculator.py @@ -3736,6 +3736,62 @@ def test_completion_cost_logs_reasoning_and_cache_breakdown(_local_model_cost_ma assert logging_obj.cost_breakdown["cache_read_cost"] == pytest.approx(100 * 3e-08) +def test_completion_cost_logs_the_rates_it_billed_at(monkeypatch): + """A caller reporting the cost lines beside their per-token rates reads both off this one call. + completion_cost infers the provider, and xai's inclusive tier thresholds put a request sitting + exactly on 200k at the tier rate, which a lookup made without that inferred provider would miss. + """ + from datetime import datetime + + from litellm.litellm_core_utils.litellm_logging import Logging + + monkeypatch.setitem( + litellm.model_cost, + "xai/tiered-model", + { + "input_cost_per_token": 3e-6, + "output_cost_per_token": 15e-6, + "cache_read_input_token_cost": 3e-7, + "input_cost_per_token_above_200k_tokens": 6e-6, + "output_cost_per_token_above_200k_tokens": 3e-5, + "cache_read_input_token_cost_above_200k_tokens": 6e-7, + "litellm_provider": "xai", + "mode": "chat", + }, + ) + logging_obj = Logging( + model="xai/tiered-model", + messages=[{"role": "user", "content": "Hello"}], + stream=False, + call_type="completion", + start_time=datetime.now(), + litellm_call_id="billed-rates", + function_id="f", + ) + usage = Usage( + prompt_tokens=200_000, + completion_tokens=1_000, + total_tokens=201_000, + prompt_tokens_details=PromptTokensDetailsWrapper(cached_tokens=100_000), + ) + + litellm.completion_cost( + completion_response=ModelResponse(model="xai/tiered-model", usage=usage), + model="xai/tiered-model", + custom_llm_provider=None, + litellm_logging_obj=logging_obj, + ) + + rates = logging_obj.billed_token_rates + assert rates is not None + assert rates.input_cost_per_token == pytest.approx(6e-6) + assert rates.cache_read_input_token_cost == pytest.approx(6e-7) + assert logging_obj.cost_breakdown["cache_read_cost"] == pytest.approx( + 100_000 * rates.cache_read_input_token_cost + ) + assert logging_obj.cost_breakdown["output_cost"] == pytest.approx(1_000 * rates.output_cost_per_token) + + def test_completion_cost_logs_cache_and_reasoning_breakdown_for_custom_pricing(): """ A custom-priced deployment bills cache tokens at its custom cache rates, but the From 996ee5635ab79854992661279c2e8db793b3dda4 Mon Sep 17 00:00:00 2001 From: "devin-ai-integration[bot]" <158243242+devin-ai-integration[bot]@users.noreply.github.com> Date: Wed, 9 Sep 2026 10:35:50 -0700 Subject: [PATCH 20/26] perf(proxy): pipeline spend counter increments into one Redis call per request (#40371) * 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 Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/proxy/proxy_server.py | 331 +++++++---- .../proxy/proxy_server/test_spend_counters.py | 538 ++++++++++-------- .../test_budget_reservation_redis_failure.py | 9 + .../proxy/test_budget_reservation.py | 9 + tests/test_litellm/proxy/test_proxy_server.py | 94 ++- 5 files changed, 609 insertions(+), 372 deletions(-) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index 7aed0553b4b..c241e66049b 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -17,6 +17,7 @@ import time import traceback import warnings from collections.abc import AsyncGenerator, AsyncIterator, Callable, Collection, Mapping, MutableMapping, Sequence +from dataclasses import dataclass from datetime import datetime, timedelta, timezone from types import MappingProxyType, UnionType from typing import ( @@ -131,6 +132,7 @@ from litellm.router_utils.auto_router_tuning_baseline import ( snapshot_tuning_baselines, tuning_limit_violation, ) +from litellm.types.caching import RedisPipelineIncrementOperation from litellm.types.utils import ( ModelResponse, ModelResponseStream, @@ -2703,6 +2705,12 @@ async def _read_spend_counter_estimate(counter_key: str, fallback_spend: float) return fallback_spend, False +@dataclass(frozen=True, slots=True) +class _PendingSpendIncrement: + counter_key: str + increment: float + + async def increment_spend_counters( token: str | None, team_id: str | None, @@ -2737,7 +2745,7 @@ async def increment_spend_counters( cost: Final[float] = response_cost - async def _key_scope(key_token: str) -> None: + async def _key_scope(key_token: str) -> tuple[_PendingSpendIncrement | BaseException, ...]: # key_token arrives pre-hashed from metadata["user_api_key"] (auth flow # hashes raw "sk-..." keys before they reach the callback). The # startswith("sk-") check is a safety net matching update_cache — @@ -2748,30 +2756,29 @@ async def increment_spend_counters( hash_token(token=key_token) if isinstance(key_token, str) and key_token.startswith("sk-") else key_token ) key_counter_key: Final = f"spend:key:{hashed_token}" - if key_counter_key not in reserved_counter_keys: - await _init_and_increment_spend_counter( - counter_key=key_counter_key, - source_cache_key=hashed_token, - increment=cost, + key_pending: Final[tuple[_PendingSpendIncrement, ...]] = ( + () + if key_counter_key in reserved_counter_keys + else ( + await _prepare_spend_counter_increment( + counter_key=key_counter_key, + source_cache_key=hashed_token, + increment=cost, + ), ) - - key_obj: Final[object] = await user_api_key_cache.async_get_cache(key=hashed_token) - if key_obj is None: - return - key_budget_limits = getattr(key_obj, "budget_limits", None) or ( - key_obj.get("budget_limits") if isinstance(key_obj, dict) else None ) - if isinstance(key_budget_limits, str): - key_budget_limits = json.loads(key_budget_limits) - if not isinstance(key_budget_limits, list): - return - for window in key_budget_limits: - duration = window["budget_duration"] if isinstance(window, dict) else window.budget_duration - key_window_reset_at = window.get("reset_at") if isinstance(window, dict) else window.reset_at - key_window_counter = f"spend:key:{hashed_token}:window:{duration}" + + async def _key_window_increment(window: object) -> _PendingSpendIncrement | None: + duration = ( + window["budget_duration"] if isinstance(window, dict) else getattr(window, "budget_duration", None) + ) + key_window_reset_at = ( + window.get("reset_at") if isinstance(window, dict) else getattr(window, "reset_at", None) + ) + key_window_counter: Final = f"spend:key:{hashed_token}:window:{duration}" key_window_start = get_budget_window_start(window) - if key_window_counter not in reserved_counter_keys: - await _init_and_increment_window_spend_counter( + pending_window: Final = ( + await _prepare_window_spend_counter_increment( counter_key=key_window_counter, entity_type="Key", entity_id=hashed_token, @@ -2779,6 +2786,9 @@ async def increment_spend_counters( window_start=key_window_start, increment=cost, ) + if key_window_counter not in reserved_counter_keys + else None + ) await _enqueue_window_spend_row_update( entity_type=Litellm_EntityType.KEY, entity_id=hashed_token, @@ -2788,33 +2798,48 @@ async def increment_spend_counters( increment=cost, request_started_at=request_started_at, ) + return pending_window - async def _team_scope(scope_team_id: str) -> None: - team_counter_key: Final = f"spend:team:{scope_team_id}" - if team_counter_key not in reserved_counter_keys: - await _init_and_increment_spend_counter( - counter_key=team_counter_key, - source_cache_key=f"team_id:{scope_team_id}", - increment=cost, - ) - - team_obj: Final[object] = await user_api_key_cache.async_get_cache(key=f"team_id:{scope_team_id}") - if team_obj is None: - return - team_budget_limits = getattr(team_obj, "budget_limits", None) or ( - team_obj.get("budget_limits") if isinstance(team_obj, dict) else None + key_obj: Final[object] = await user_api_key_cache.async_get_cache(key=hashed_token) + if key_obj is None: + return key_pending + key_budget_limits = getattr(key_obj, "budget_limits", None) or ( + key_obj.get("budget_limits") if isinstance(key_obj, dict) else None ) - if isinstance(team_budget_limits, str): - team_budget_limits = json.loads(team_budget_limits) - if not isinstance(team_budget_limits, list): - return - for window in team_budget_limits: - duration = window["budget_duration"] if isinstance(window, dict) else window.budget_duration - team_window_reset_at = window.get("reset_at") if isinstance(window, dict) else window.reset_at - team_window_counter = f"spend:team:{scope_team_id}:window:{duration}" + if isinstance(key_budget_limits, str): + key_budget_limits = json.loads(key_budget_limits) + if not isinstance(key_budget_limits, list): + return key_pending + window_pending: Final = await asyncio.gather( + *(_key_window_increment(window) for window in key_budget_limits), return_exceptions=True + ) + return key_pending + tuple(item for item in window_pending if item is not None) + + async def _team_scope(scope_team_id: str) -> tuple[_PendingSpendIncrement | BaseException, ...]: + team_counter_key: Final = f"spend:team:{scope_team_id}" + team_pending: Final[tuple[_PendingSpendIncrement, ...]] = ( + () + if team_counter_key in reserved_counter_keys + else ( + await _prepare_spend_counter_increment( + counter_key=team_counter_key, + source_cache_key=f"team_id:{scope_team_id}", + increment=cost, + ), + ) + ) + + async def _team_window_increment(window: object) -> _PendingSpendIncrement | None: + duration = ( + window["budget_duration"] if isinstance(window, dict) else getattr(window, "budget_duration", None) + ) + team_window_reset_at = ( + window.get("reset_at") if isinstance(window, dict) else getattr(window, "reset_at", None) + ) + team_window_counter: Final = f"spend:team:{scope_team_id}:window:{duration}" team_window_start = get_budget_window_start(window) - if team_window_counter not in reserved_counter_keys: - await _init_and_increment_window_spend_counter( + pending_window: Final = ( + await _prepare_window_spend_counter_increment( counter_key=team_window_counter, entity_type="Team", entity_id=scope_team_id, @@ -2822,6 +2847,9 @@ async def increment_spend_counters( window_start=team_window_start, increment=cost, ) + if team_window_counter not in reserved_counter_keys + else None + ) await _enqueue_window_spend_row_update( entity_type=Litellm_EntityType.TEAM, entity_id=scope_team_id, @@ -2831,25 +2859,47 @@ async def increment_spend_counters( increment=cost, request_started_at=request_started_at, ) + return pending_window - async def _team_member_scope(scope_user_id: str, scope_team_id: str) -> None: + team_obj: Final[object] = await user_api_key_cache.async_get_cache(key=f"team_id:{scope_team_id}") + if team_obj is None: + return team_pending + team_budget_limits = getattr(team_obj, "budget_limits", None) or ( + team_obj.get("budget_limits") if isinstance(team_obj, dict) else None + ) + if isinstance(team_budget_limits, str): + team_budget_limits = json.loads(team_budget_limits) + if not isinstance(team_budget_limits, list): + return team_pending + window_pending: Final = await asyncio.gather( + *(_team_window_increment(window) for window in team_budget_limits), return_exceptions=True + ) + return team_pending + tuple(item for item in window_pending if item is not None) + + async def _team_member_scope( + scope_user_id: str, scope_team_id: str + ) -> tuple[_PendingSpendIncrement | BaseException, ...]: team_member_counter_key: Final = f"spend:team_member:{scope_user_id}:{scope_team_id}" if team_member_counter_key in reserved_counter_keys: - return - await _init_and_increment_spend_counter( - counter_key=team_member_counter_key, - source_cache_key=f"team_membership:{scope_user_id}:{scope_team_id}", - increment=cost, + return () + return ( + await _prepare_spend_counter_increment( + counter_key=team_member_counter_key, + source_cache_key=f"team_membership:{scope_user_id}:{scope_team_id}", + increment=cost, + ), ) - async def _user_scope(scope_user_id: str) -> None: + async def _user_scope(scope_user_id: str) -> tuple[_PendingSpendIncrement | BaseException, ...]: user_counter_key: Final = f"spend:user:{scope_user_id}" if user_counter_key in reserved_counter_keys: - return - await _init_and_increment_spend_counter( - counter_key=user_counter_key, - source_cache_key=scope_user_id, - increment=cost, + return () + return ( + await _prepare_spend_counter_increment( + counter_key=user_counter_key, + source_cache_key=scope_user_id, + increment=cost, + ), ) scope_coros: Final = tuple( @@ -2859,7 +2909,7 @@ async def increment_spend_counters( _team_scope(team_id) if team_id is not None else None, _team_member_scope(user_id, team_id) if user_id is not None and team_id is not None else None, _user_scope(user_id) if user_id is not None else None, - _increment_end_user_and_tag_spend_counters( + _prepare_end_user_and_tag_spend_increments( end_user_id=end_user_id, tags=tags, response_cost=cost, @@ -2867,14 +2917,14 @@ async def increment_spend_counters( ) if end_user_id is not None or tags is not None else None, - _increment_model_access_group_spend_counters( + _prepare_model_access_group_spend_increments( model_access_groups=model_access_groups, response_cost=cost, reserved_counter_keys=reserved_counter_keys, ) if model_access_groups else None, - _increment_org_spend_counter( + _prepare_org_spend_increment( org_id=org_id, response_cost=cost, reserved_counter_keys=reserved_counter_keys, @@ -2889,7 +2939,20 @@ async def increment_spend_counters( # as orphaned tasks that race the caller's reservation-counter invalidation; # all scopes settle, then the first error propagates as before. scope_results: Final = await asyncio.gather(*scope_coros, return_exceptions=True) - scope_errors: Final = [r for r in scope_results if isinstance(r, BaseException)] + scope_errors: Final = tuple( + item + for scope in scope_results + for item in (scope if isinstance(scope, tuple) else (scope,)) + if isinstance(item, BaseException) + ) + pending: Final = tuple( + item + for scope in scope_results + if not isinstance(scope, BaseException) + for item in scope + if not isinstance(item, BaseException) + ) + await _apply_spend_counter_increments(pending=pending) if scope_errors: raise scope_errors[0] @@ -2932,41 +2995,49 @@ async def _reconcile_budget_reservation_for_counter_update( return reserved_counter_keys -async def _increment_end_user_and_tag_spend_counters( +async def _prepare_end_user_and_tag_spend_increments( end_user_id: str | None, tags: list[str] | None, response_cost: float, reserved_counter_keys: set[str], -) -> None: - if end_user_id is not None: - await _init_and_increment_unreserved_spend_counter( - counter_key=f"spend:end_user:{end_user_id}", - source_cache_key=end_user_cache_key(end_user_id), - increment=response_cost, - reserved_counter_keys=reserved_counter_keys, - ) - - if tags is None: - return - - seen_tags: Final[set[str]] = set() - for tag_name in tags: - if not tag_name or not isinstance(tag_name, str) or tag_name in seen_tags: - continue - seen_tags.add(tag_name) - await _init_and_increment_unreserved_spend_counter( - counter_key=f"spend:tag:{tag_name}", - source_cache_key=tag_cache_key(tag_name), - increment=response_cost, - reserved_counter_keys=reserved_counter_keys, - ) +) -> tuple[_PendingSpendIncrement | BaseException, ...]: + unique_tags: Final = ( + tuple(dict.fromkeys(tag for tag in tags if tag and isinstance(tag, str))) if tags is not None else () + ) + results: Final = await asyncio.gather( + *( + coro + for coro in ( + _prepare_unreserved_spend_counter_increment( + counter_key=f"spend:end_user:{end_user_id}", + source_cache_key=end_user_cache_key(end_user_id), + increment=response_cost, + reserved_counter_keys=reserved_counter_keys, + ) + if end_user_id is not None + else None, + *( + _prepare_unreserved_spend_counter_increment( + counter_key=f"spend:tag:{tag_name}", + source_cache_key=tag_cache_key(tag_name), + increment=response_cost, + reserved_counter_keys=reserved_counter_keys, + ) + for tag_name in unique_tags + ), + ) + if coro is not None + ), + return_exceptions=True, + ) + return tuple(item for item in results if item is not None) -async def _increment_model_access_group_spend_counters( +async def _prepare_model_access_group_spend_increments( model_access_groups: Sequence[object], response_cost: float, reserved_counter_keys: set[str], -) -> None: +) -> tuple[_PendingSpendIncrement | BaseException, ...]: """Charge the model access groups that authorized this request. Without this the counter auth reads is written only by the reservation path, so @@ -2980,55 +3051,63 @@ async def _increment_model_access_group_spend_counters( unique_groups: Final = tuple( dict.fromkeys(group for group in model_access_groups if group and isinstance(group, str)) ) - for group in unique_groups: - await _init_and_increment_unreserved_spend_counter( - counter_key=model_access_group_spend_counter_key(group), - source_cache_key=model_access_group_cache_key(group), - increment=response_cost, - reserved_counter_keys=reserved_counter_keys, - ) + results: Final = await asyncio.gather( + *( + _prepare_unreserved_spend_counter_increment( + counter_key=model_access_group_spend_counter_key(group), + source_cache_key=model_access_group_cache_key(group), + increment=response_cost, + reserved_counter_keys=reserved_counter_keys, + ) + for group in unique_groups + ), + return_exceptions=True, + ) + return tuple(item for item in results if item is not None) -async def _increment_org_spend_counter( +async def _prepare_org_spend_increment( org_id: str | None, response_cost: float, reserved_counter_keys: set[str], -) -> None: +) -> tuple[_PendingSpendIncrement, ...]: if org_id is None: - return + return () - await _init_and_increment_unreserved_spend_counter( + pending: Final = await _prepare_unreserved_spend_counter_increment( counter_key=f"spend:org:{org_id}", source_cache_key=[f"org_id:{org_id}:with_budget", f"org_id:{org_id}"], increment=response_cost, reserved_counter_keys=reserved_counter_keys, ) + return (pending,) if pending is not None else () -async def _init_and_increment_unreserved_spend_counter( +async def _prepare_unreserved_spend_counter_increment( counter_key: str, source_cache_key: str | list[str], increment: float, reserved_counter_keys: set[str], -) -> None: +) -> _PendingSpendIncrement | None: if counter_key in reserved_counter_keys: - return + return None - await _init_and_increment_spend_counter( + return await _prepare_spend_counter_increment( counter_key=counter_key, source_cache_key=source_cache_key, increment=increment, ) -async def _init_and_increment_spend_counter( +async def _prepare_spend_counter_increment( counter_key: str, source_cache_key: str | list[str], increment: float, -): +) -> _PendingSpendIncrement: """ Initialize counter from the authoritative DB spend value if not yet - set, then atomically increment in both in-memory and Redis. + set, then return the pending increment for the caller to apply in one + pipelined Redis call. On first access per pod: 1. Check spend_counter_cache (in-memory -> Redis via DualCache) @@ -3040,13 +3119,13 @@ async def _init_and_increment_spend_counter( the counter as absent and seed it. Using increment means the worst case is over-counting (conservative, blocks slightly early) rather than under-counting (would allow overspend). - 4. Increment atomically (both in-memory + Redis) + 4. Increment is returned for the caller to apply via pipeline """ await _ensure_spend_counter_initialized( counter_key=counter_key, source_cache_key=source_cache_key, ) - await _increment_spend_counter_cache(counter_key=counter_key, increment=increment) + return _PendingSpendIncrement(counter_key=counter_key, increment=increment) async def _enqueue_window_spend_row_update( @@ -3098,20 +3177,20 @@ async def _enqueue_window_spend_row_update( ) -async def _init_and_increment_window_spend_counter( +async def _prepare_window_spend_counter_increment( counter_key: str, entity_type: str, entity_id: str, window_duration: str | None, window_start: datetime | None, increment: float, -): +) -> _PendingSpendIncrement | None: if window_start is None: verbose_proxy_logger.warning( "Skipping spend counter increment for invalid budget window %s", counter_key, ) - return + return None initialized: Final = await _ensure_window_spend_counter_initialized( counter_key=counter_key, @@ -3121,8 +3200,8 @@ async def _init_and_increment_window_spend_counter( window_start=window_start, ) if initialized is False: - return - await _increment_spend_counter_cache(counter_key=counter_key, increment=increment) + return None + return _PendingSpendIncrement(counter_key=counter_key, increment=increment) async def _ensure_spend_counter_initialized( @@ -3255,6 +3334,32 @@ async def _invalidate_spend_counter(counter_key: str): ) +async def _apply_spend_counter_increments(pending: Sequence[_PendingSpendIncrement]) -> None: + if not pending: + return + redis_cache: Final = spend_counter_cache.redis_cache + if redis_cache is None: + for item in pending: + await spend_counter_cache.async_increment_cache( + key=item.counter_key, + value=item.increment, + refresh_ttl=True, + ) + return + ttl: Final = redis_cache.get_ttl() + increment_list: Final = [ # mutable-ok: async_increment_pipeline signature requires list[RedisPipelineIncrementOperation] + RedisPipelineIncrementOperation(key=item.counter_key, increment_value=item.increment, ttl=ttl) + for item in pending + ] + try: + results: Final = await redis_cache.async_increment_pipeline(increment_list=increment_list) + except Exception: + await asyncio.gather(*(_invalidate_spend_counter(counter_key=item.counter_key) for item in pending)) + raise + for item, current_value in zip(pending, results or ()): + spend_counter_cache.in_memory_cache.set_cache(key=item.counter_key, value=current_value) + + async def update_cache( token: str | None, user_id: str | None, diff --git a/tests/test_litellm/proxy/proxy_server/test_spend_counters.py b/tests/test_litellm/proxy/proxy_server/test_spend_counters.py index 86e97a334df..c343652efd9 100644 --- a/tests/test_litellm/proxy/proxy_server/test_spend_counters.py +++ b/tests/test_litellm/proxy/proxy_server/test_spend_counters.py @@ -4,11 +4,12 @@ Pins covered: - ``get_current_spend`` - ``increment_spend_counters`` - ``_reconcile_budget_reservation_for_counter_update`` -- ``_increment_end_user_and_tag_spend_counters`` -- ``_increment_org_spend_counter`` -- ``_init_and_increment_unreserved_spend_counter`` -- ``_init_and_increment_spend_counter`` -- ``_init_and_increment_window_spend_counter`` +- ``_prepare_end_user_and_tag_spend_increments`` +- ``_prepare_org_spend_increment`` +- ``_prepare_unreserved_spend_counter_increment`` +- ``_prepare_spend_counter_increment`` +- ``_prepare_window_spend_counter_increment`` +- ``_apply_spend_counter_increments`` - ``_ensure_spend_counter_initialized`` - ``_get_source_cache_base_spend`` - ``_ensure_window_spend_counter_initialized`` @@ -48,9 +49,7 @@ def _make_spend_counter_cache( cache.in_memory_cache.delete_cache = MagicMock() if with_redis: cache.redis_cache = MagicMock() - cache.redis_cache.async_get_cache = AsyncMock( - return_value=redis_get_value, side_effect=redis_get_side_effect - ) + cache.redis_cache.async_get_cache = AsyncMock(return_value=redis_get_value, side_effect=redis_get_side_effect) cache.redis_cache.async_increment = AsyncMock( return_value=redis_increment_value, side_effect=redis_increment_side_effect, @@ -58,6 +57,8 @@ def _make_spend_counter_cache( cache.redis_cache.async_delete_cache = AsyncMock() cache.redis_cache.async_set_cache = AsyncMock() cache.redis_cache.async_set_max = AsyncMock() + cache.redis_cache.async_increment_pipeline = AsyncMock(return_value=None) + cache.redis_cache.get_ttl = MagicMock(return_value=None) else: cache.redis_cache = None cache.async_increment_cache = AsyncMock(return_value=redis_increment_value) @@ -70,9 +71,7 @@ def _make_spend_counter_cache( def _make_user_api_key_cache(get_value=None, get_side_effect=None): cache = MagicMock() - cache.async_get_cache = AsyncMock( - return_value=get_value, side_effect=get_side_effect - ) + cache.async_get_cache = AsyncMock(return_value=get_value, side_effect=get_side_effect) cache.async_set_cache_pipeline = AsyncMock() return cache @@ -109,9 +108,7 @@ async def test_get_current_spend_redis_error_falls_back_to_in_memory(monkeypatch ) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - result = await ps.get_current_spend( - counter_key="spend:key:abc", fallback_spend=99.0 - ) + result = await ps.get_current_spend(counter_key="spend:key:abc", fallback_spend=99.0) assert result == 17.0 @@ -136,9 +133,7 @@ async def test_get_current_spend_floors_stale_low_counter_against_db(monkeypatch # the stale counter is repaired up to the authoritative DB value via a # monotonic set-max so other workers read the corrected total, and a # concurrent increment cannot be clobbered - fake_cache.redis_cache.async_set_max.assert_awaited_once_with( - key="spend:key:abc", value=12.0 - ) + fake_cache.redis_cache.async_set_max.assert_awaited_once_with(key="spend:key:abc", value=12.0) @pytest.mark.asyncio @@ -169,9 +164,7 @@ async def test_get_current_spend_no_floor_without_max_budget(monkeypatch): from_db = AsyncMock(return_value=12.0) monkeypatch.setattr(ps.SpendCounterReseed, "from_db", from_db) - result = await ps.get_current_spend( - counter_key="spend:key:abc", fallback_spend=12.0 - ) + result = await ps.get_current_spend(counter_key="spend:key:abc", fallback_spend=12.0) assert result == 2.0 assert from_db.await_count == 0 @@ -210,12 +203,8 @@ async def test_get_current_spend_floor_caches_db_read(monkeypatch): from_db = AsyncMock(return_value=12.0) monkeypatch.setattr(ps.SpendCounterReseed, "from_db", from_db) - first = await ps.get_current_spend( - counter_key="spend:key:abc", fallback_spend=12.0, max_budget=10.0 - ) - second = await ps.get_current_spend( - counter_key="spend:key:abc", fallback_spend=12.0, max_budget=10.0 - ) + first = await ps.get_current_spend(counter_key="spend:key:abc", fallback_spend=12.0, max_budget=10.0) + second = await ps.get_current_spend(counter_key="spend:key:abc", fallback_spend=12.0, max_budget=10.0) assert first == 12.0 assert second == 12.0 @@ -336,9 +325,7 @@ async def test_get_current_spend_floors_window_against_spend_logs(monkeypatch): assert result == 15.0 assert wfsl.await_count == 1 - fake_cache.redis_cache.async_set_max.assert_awaited_once_with( - key=counter_key, value=15.0 - ) + fake_cache.redis_cache.async_set_max.assert_awaited_once_with(key=counter_key, value=15.0) def _make_window_spend_prisma(row=None, spend_logs_total=0.0): @@ -379,9 +366,7 @@ async def test_get_current_spend_floors_window_against_maintained_row(monkeypatc assert result == 15.0 fake_prisma.db.litellm_spendlogs.group_by.assert_not_awaited() - fake_cache.redis_cache.async_set_max.assert_awaited_once_with( - key=counter_key, value=15.0 - ) + fake_cache.redis_cache.async_set_max.assert_awaited_once_with(key=counter_key, value=15.0) @pytest.mark.asyncio @@ -393,9 +378,7 @@ async def test_get_current_spend_floors_window_against_logs_when_row_stale(monke window_start = datetime(2026, 1, 8, tzinfo=timezone.utc) fake_prisma = _make_window_spend_prisma( - row=SimpleNamespace( - window_start=window_start - timedelta(days=7), spend=999.0 - ), + row=SimpleNamespace(window_start=window_start - timedelta(days=7), spend=999.0), spend_logs_total=15.0, ) fake_cache = _make_spend_counter_cache(redis_get_value=2.0) @@ -423,21 +406,13 @@ async def test_get_current_spend_fail_closed_rejects_when_unverifiable(monkeypat rather than admitted on an unverifiable budget.""" from fastapi import HTTPException - fake_cache = _make_spend_counter_cache( - redis_get_side_effect=RuntimeError("redis down") - ) + fake_cache = _make_spend_counter_cache(redis_get_side_effect=RuntimeError("redis down")) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - monkeypatch.setattr( - ps, "general_settings", {"fail_closed_budget_enforcement": True} - ) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps, "general_settings", {"fail_closed_budget_enforcement": True}) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) with pytest.raises(HTTPException) as exc: - await ps.get_current_spend( - counter_key="spend:key:abc", fallback_spend=1.0, max_budget=10.0 - ) + await ps.get_current_spend(counter_key="spend:key:abc", fallback_spend=1.0, max_budget=10.0) assert exc.value.status_code == 503 @@ -445,18 +420,12 @@ async def test_get_current_spend_fail_closed_rejects_when_unverifiable(monkeypat async def test_get_current_spend_fail_closed_off_admits_when_unverifiable(monkeypatch): """Default (flag off): an unverifiable read keeps the existing behavior and admits using the cached fallback — no new rejection.""" - fake_cache = _make_spend_counter_cache( - redis_get_side_effect=RuntimeError("redis down") - ) + fake_cache = _make_spend_counter_cache(redis_get_side_effect=RuntimeError("redis down")) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "general_settings", {}) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) - result = await ps.get_current_spend( - counter_key="spend:key:abc", fallback_spend=1.0, max_budget=10.0 - ) + result = await ps.get_current_spend(counter_key="spend:key:abc", fallback_spend=1.0, max_budget=10.0) assert result == 1.0 @@ -466,13 +435,9 @@ async def test_get_current_spend_fail_closed_admits_when_redis_verified(monkeypa authoritative, so an under-budget request is admitted normally.""" fake_cache = _make_spend_counter_cache(redis_get_value=1.0) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - monkeypatch.setattr( - ps, "general_settings", {"fail_closed_budget_enforcement": True} - ) + monkeypatch.setattr(ps, "general_settings", {"fail_closed_budget_enforcement": True}) - result = await ps.get_current_spend( - counter_key="spend:key:abc", fallback_spend=1.0, max_budget=10.0 - ) + result = await ps.get_current_spend(counter_key="spend:key:abc", fallback_spend=1.0, max_budget=10.0) assert result == 1.0 @@ -481,16 +446,10 @@ async def test_get_current_spend_fail_closed_allows_authoritative_fallback(monke """End-user/tag callers pass fallback_authoritative=True (their spend is loaded fresh from the DB in auth), so fail-closed does not reject them even when the counter path is unreadable.""" - fake_cache = _make_spend_counter_cache( - redis_get_side_effect=RuntimeError("redis down") - ) + fake_cache = _make_spend_counter_cache(redis_get_side_effect=RuntimeError("redis down")) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - monkeypatch.setattr( - ps, "general_settings", {"fail_closed_budget_enforcement": True} - ) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps, "general_settings", {"fail_closed_budget_enforcement": True}) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) result = await ps.get_current_spend( counter_key="spend:end_user:e1", @@ -508,9 +467,7 @@ async def test_get_current_spend_strict_floors_when_fallback_also_stale(monkeypa re-checks the authoritative DB and enforces against it.""" fake_cache = _make_spend_counter_cache(redis_get_value=0.00001) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - monkeypatch.setattr( - ps, "general_settings", {"fail_closed_budget_enforcement": True} - ) + monkeypatch.setattr(ps, "general_settings", {"fail_closed_budget_enforcement": True}) from_db = AsyncMock(return_value=0.5) monkeypatch.setattr(ps.SpendCounterReseed, "from_db", from_db) @@ -532,9 +489,7 @@ async def test_get_current_spend_strict_floors_when_fallback_also_stale(monkeypa @pytest.mark.asyncio async def test_increment_spend_counters_increments_all_buckets(monkeypatch): - fake_cache = _make_spend_counter_cache( - redis_get_value=None, redis_increment_value=5.0 - ) + fake_cache = _make_spend_counter_cache(redis_get_value=None, redis_increment_value=5.0) fake_user_cache = _make_user_api_key_cache(get_value=None) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) @@ -543,9 +498,7 @@ async def test_increment_spend_counters_increments_all_buckets(monkeypatch): async def _fake_coalesced(**kwargs): return None - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(side_effect=_fake_coalesced) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(side_effect=_fake_coalesced)) await ps.increment_spend_counters( token="hashed-tok", @@ -554,25 +507,36 @@ async def test_increment_spend_counters_increments_all_buckets(monkeypatch): response_cost=5.0, ) + pipeline = fake_cache.redis_cache.async_increment_pipeline + pipeline.assert_awaited_once() + increment_list = pipeline.await_args.kwargs["increment_list"] + assert {op["key"] for op in increment_list} == { + "spend:key:hashed-tok", + "spend:team:t1", + "spend:team_member:u1:t1", + "spend:user:u1", + } + assert all(op["increment_value"] == 5.0 for op in increment_list) observed = { "redis_increment_called": fake_cache.redis_cache.async_increment.called, - "increment_calls": fake_cache.redis_cache.async_increment.call_count, + "pipeline_calls": pipeline.await_count, "user_cache_used": fake_user_cache.async_get_cache.called, } assert normalize(observed) == { - "redis_increment_called": True, - "increment_calls": 4, + "redis_increment_called": False, + "pipeline_calls": 1, "user_cache_used": True, } class _ConcurrencyProbe: - """Stand-in for redis_cache.async_increment that pins concurrency. + """Stand-in for redis_cache.async_get_cache that pins concurrency. - Each call registers itself as in-flight and blocks on ``release`` until the - test lets it proceed. ``all_arrived`` fires once ``expected`` distinct scope - increments are simultaneously suspended here, which can only happen if the - per-scope increments are gathered rather than awaited one after another. + Each warm-check read registers itself as in-flight and blocks on ``release`` + until the test lets it proceed. ``all_arrived`` fires once ``expected`` + distinct scope warm-checks are simultaneously suspended here, which can only + happen if the per-scope prepares are gathered rather than awaited one after + another. """ def __init__(self, expected_concurrency: int): @@ -581,36 +545,45 @@ class _ConcurrencyProbe: self.max_in_flight = 0 self.all_arrived = asyncio.Event() self.release = asyncio.Event() - self.values: dict[str, float] = {} + self.keys: list[str] = [] - async def async_increment(self, *, key, value, refresh_ttl=True): + async def async_get_cache(self, *, key, **kwargs): self.in_flight += 1 self.max_in_flight = max(self.max_in_flight, self.in_flight) + self.keys.append(key) if self.in_flight >= self.expected: self.all_arrived.set() if not self.release.is_set(): await self.release.wait() self.in_flight -= 1 - self.values[key] = self.values.get(key, 0.0) + value - return self.values[key] + return 1.0 @pytest.mark.asyncio async def test_increment_spend_counters_runs_scopes_concurrently(monkeypatch): """The six independent scopes (key, team, team_member, user, end_user+tags, - org) must be incremented concurrently. The probe only fires once all six are - suspended in async_increment at the same time, which is impossible if the + org) must prepare their increments concurrently. The probe only fires once + all eight warm-check reads (one per counter: 6 scopes + 2 tags) are + suspended in async_get_cache at the same time, which is impossible if the awaits are chained sequentially.""" - probe = _ConcurrencyProbe(expected_concurrency=6) - fake_cache = _make_spend_counter_cache(redis_get_value=None) - fake_cache.redis_cache.async_increment = probe.async_increment + probe = _ConcurrencyProbe(expected_concurrency=8) + fake_cache = _make_spend_counter_cache() + fake_cache.redis_cache.async_get_cache = probe.async_get_cache + recorded: dict[str, float] = {} + + async def _record_pipeline(increment_list, **_): + results = [] + for op in increment_list: + recorded[op["key"]] = recorded.get(op["key"], 0.0) + op["increment_value"] + results.append(recorded[op["key"]]) + return results + + fake_cache.redis_cache.async_increment_pipeline = AsyncMock(side_effect=_record_pipeline) fake_user_cache = _make_user_api_key_cache(get_value=None) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) monkeypatch.setattr(ps, "prisma_client", None) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) task = asyncio.create_task( ps.increment_spend_counters( @@ -630,16 +603,16 @@ async def test_increment_spend_counters_runs_scopes_concurrently(monkeypatch): probe.release.set() await task pytest.fail( - "scope increments did not run concurrently; sequential awaits " - f"detected (peak in-flight was {probe.max_in_flight}, expected 6)" + "scope prepares did not run concurrently; sequential awaits " + f"detected (peak in-flight was {probe.max_in_flight}, expected 8)" ) - assert probe.in_flight == 6 - assert probe.max_in_flight == 6 + assert probe.in_flight == 8 + assert probe.max_in_flight == 8 probe.release.set() await task - assert probe.values == { + assert recorded == { "spend:key:hashed-tok": 5.0, "spend:team:t1": 5.0, "spend:team_member:u1:t1": 5.0, @@ -659,26 +632,25 @@ async def test_increment_spend_counters_skips_reserved_counter_keys(monkeypatch) import litellm.proxy.spend_tracking.budget_reservation as br reserved = {"spend:key:hashed-tok", "spend:org:org1"} - monkeypatch.setattr( - br, "get_reserved_counter_keys", MagicMock(return_value=set(reserved)) - ) + monkeypatch.setattr(br, "get_reserved_counter_keys", MagicMock(return_value=set(reserved))) monkeypatch.setattr(br, "reconcile_budget_reservation", AsyncMock()) recorded: dict[str, float] = {} - async def _record_increment(*, key, value, refresh_ttl=True): - recorded[key] = recorded.get(key, 0.0) + value - return recorded[key] + async def _record_pipeline(increment_list, **_): + results = [] + for op in increment_list: + recorded[op["key"]] = recorded.get(op["key"], 0.0) + op["increment_value"] + results.append(recorded[op["key"]]) + return results fake_cache = _make_spend_counter_cache(redis_get_value=None) - fake_cache.redis_cache.async_increment = _record_increment + fake_cache.redis_cache.async_increment_pipeline = AsyncMock(side_effect=_record_pipeline) fake_user_cache = _make_user_api_key_cache(get_value=None) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) monkeypatch.setattr(ps, "prisma_client", None) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) reservation = {"finalized": False} await ps.increment_spend_counters( @@ -708,27 +680,46 @@ async def test_increment_spend_counters_failing_scope_propagates_after_siblings_ ): """A failure in one scope must propagate to the caller (so it can invalidate reserved counters) while every other scope still settles rather than being - left as an orphaned background task, and the reservation is not finalized.""" - recorded: dict[str, float] = {} + left as an orphaned background task, and the reservation is not finalized. + The surviving scopes' increments are still applied in the single pipeline: + dropping them would under-count spend, the unsafe direction for budget + enforcement.""" + warmed_keys: list[str] = [] - async def _increment(*, key, value, refresh_ttl=True): + async def _warm_check(*, key, **kwargs): + warmed_keys.append(key) if key == "spend:team:t1": - raise RuntimeError("redis increment failed") - recorded[key] = recorded.get(key, 0.0) + value - return recorded[key] + raise RuntimeError("redis get failed") + return 1.0 - fake_cache = _make_spend_counter_cache(redis_get_value=None) - fake_cache.redis_cache.async_increment = _increment + async def _reseed_fails(*, counter_key, **kwargs): + if counter_key == "spend:team:t1": + raise RuntimeError("reseed failed") + + applied: dict[str, float] = {} + + async def _record_pipeline(increment_list, **_): + results = [] + for op in increment_list: + applied[op["key"]] = op["increment_value"] + results.append(op["increment_value"]) + return results + + fake_cache = _make_spend_counter_cache() + fake_cache.redis_cache.async_get_cache = AsyncMock(side_effect=_warm_check) + fake_cache.redis_cache.async_increment_pipeline = AsyncMock(side_effect=_record_pipeline) fake_user_cache = _make_user_api_key_cache(get_value=None) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) monkeypatch.setattr(ps, "prisma_client", None) monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) + ps.SpendCounterReseed, + "coalesced", + AsyncMock(side_effect=_reseed_fails), ) reservation = {"finalized": False} - with pytest.raises(RuntimeError, match="redis increment failed"): + with pytest.raises(RuntimeError, match="reseed failed"): await ps.increment_spend_counters( token="hashed-tok", team_id="t1", @@ -741,7 +732,19 @@ async def test_increment_spend_counters_failing_scope_propagates_after_siblings_ ) assert reservation["finalized"] is False - assert recorded == { + # every sibling scope settled (its warm-check ran) before the error propagated + assert set(warmed_keys) == { + "spend:key:hashed-tok", + "spend:team:t1", + "spend:team_member:u1:t1", + "spend:user:u1", + "spend:end_user:eu1", + "spend:tag:a", + "spend:org:org1", + } + # the surviving scopes' increments were still applied, in one pipeline call + fake_cache.redis_cache.async_increment_pipeline.assert_awaited_once() + assert applied == { "spend:key:hashed-tok": 5.0, "spend:team_member:u1:t1": 5.0, "spend:user:u1": 5.0, @@ -749,6 +752,7 @@ async def test_increment_spend_counters_failing_scope_propagates_after_siblings_ "spend:tag:a": 5.0, "spend:org:org1": 5.0, } + fake_cache.redis_cache.async_increment.assert_not_awaited() @pytest.mark.asyncio @@ -772,6 +776,108 @@ async def test_increment_spend_counters_zero_cost_is_noop_finalizes_reservation( assert reservation == {"finalized": True} assert fake_cache.redis_cache.async_increment.called is False + fake_cache.redis_cache.async_increment_pipeline.assert_not_awaited() + + +@pytest.mark.asyncio +async def test_increment_spend_counters_pipelines_all_scopes_in_one_redis_call( + monkeypatch, +): + """Every scope's increment must go out in a single async_increment_pipeline + call, not one INCRBYFLOAT round-trip per scope.""" + counter_cache = ps.DualCache() + fake_redis = AsyncMock() + fake_redis.async_get_cache = AsyncMock(return_value=1.0) # counters warm + + async def _pipeline(increment_list, **_): + return [1.5] * len(increment_list) + + fake_redis.async_increment_pipeline = AsyncMock(side_effect=_pipeline) + fake_redis.async_increment = AsyncMock() + fake_redis.get_ttl = MagicMock(return_value=None) + counter_cache.redis_cache = fake_redis + monkeypatch.setattr(ps, "spend_counter_cache", counter_cache) + monkeypatch.setattr(ps, "user_api_key_cache", ps.DualCache()) + monkeypatch.setattr(ps, "prisma_client", None) + + await ps.increment_spend_counters( + token="hashed", + team_id="team-1", + user_id="user-1", + response_cost=0.5, + org_id="org-1", + end_user_id="eu-1", + tags=["tag-a", "tag-b"], + ) + + fake_redis.async_increment_pipeline.assert_awaited_once() + assert fake_redis.async_increment.await_count == 0 + increment_list = fake_redis.async_increment_pipeline.await_args.kwargs["increment_list"] + expected_keys = { + "spend:key:hashed", + "spend:team:team-1", + "spend:team_member:user-1:team-1", + "spend:user:user-1", + "spend:end_user:eu-1", + "spend:tag:tag-a", + "spend:tag:tag-b", + "spend:org:org-1", + } + assert {op["key"] for op in increment_list} == expected_keys + assert all(op["increment_value"] == 0.5 for op in increment_list) + for key in expected_keys: + assert counter_cache.in_memory_cache.get_cache(key=key) == 1.5 + + +@pytest.mark.asyncio +async def test_increment_spend_counters_pipeline_failure_invalidates_all_counters( + monkeypatch, +): + """A failing pipeline must invalidate every pending counter so the next + request reseeds from the DB (which already holds this request's cost) + instead of trusting a value the write may have partially applied.""" + from redis.exceptions import MaxConnectionsError + + counter_cache = ps.DualCache() + pending_keys = ( + "spend:key:hashed", + "spend:team:team-1", + "spend:team_member:user-1:team-1", + "spend:user:user-1", + "spend:end_user:eu-1", + "spend:tag:tag-a", + "spend:tag:tag-b", + "spend:org:org-1", + ) + for key in pending_keys: + counter_cache.in_memory_cache.set_cache(key=key, value=1.0) + fake_redis = AsyncMock() + fake_redis.async_get_cache = AsyncMock(return_value=1.0) # counters warm + fake_redis.async_increment_pipeline = AsyncMock(side_effect=MaxConnectionsError()) + fake_redis.async_increment = AsyncMock() + fake_redis.async_delete_cache = AsyncMock() + fake_redis.get_ttl = MagicMock(return_value=None) + counter_cache.redis_cache = fake_redis + monkeypatch.setattr(ps, "spend_counter_cache", counter_cache) + monkeypatch.setattr(ps, "user_api_key_cache", ps.DualCache()) + monkeypatch.setattr(ps, "prisma_client", None) + + with pytest.raises(MaxConnectionsError): + await ps.increment_spend_counters( + token="hashed", + team_id="team-1", + user_id="user-1", + response_cost=0.5, + org_id="org-1", + end_user_id="eu-1", + tags=["tag-a", "tag-b"], + ) + + assert fake_redis.async_increment.await_count == 0 + deleted_keys = {call.kwargs["key"] for call in fake_redis.async_delete_cache.await_args_list} + assert deleted_keys == set(pending_keys) + for key in pending_keys: + assert counter_cache.in_memory_cache.get_cache(key=key) is None # --------------------------------------------------------------------------- @@ -781,9 +887,7 @@ async def test_increment_spend_counters_zero_cost_is_noop_finalizes_reservation( @pytest.mark.asyncio async def test_reconcile_budget_reservation_for_counter_update_returns_empty_set_when_none(): - result = await ps._reconcile_budget_reservation_for_counter_update( - budget_reservation=None, response_cost=1.0 - ) + result = await ps._reconcile_budget_reservation_for_counter_update(budget_reservation=None, response_cost=1.0) assert result == set() @@ -818,179 +922,151 @@ async def test_reconcile_budget_reservation_for_counter_update_failure_invalidat # --------------------------------------------------------------------------- -# _increment_end_user_and_tag_spend_counters +# _prepare_end_user_and_tag_spend_increments # --------------------------------------------------------------------------- @pytest.mark.asyncio -async def test_increment_end_user_and_tag_spend_counters_increments_each_unique_tag( +async def test_prepare_end_user_and_tag_spend_increments_returns_each_unique_tag( monkeypatch, ): - fake_cache = _make_spend_counter_cache( - redis_get_value=None, redis_increment_value=3.0 - ) + fake_cache = _make_spend_counter_cache(redis_get_value=1.0) fake_user_cache = _make_user_api_key_cache() monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) monkeypatch.setattr(ps, "prisma_client", None) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) - await ps._increment_end_user_and_tag_spend_counters( + pending = await ps._prepare_end_user_and_tag_spend_increments( end_user_id="eu1", tags=["a", "b", "a", "", None], response_cost=3.0, reserved_counter_keys=set(), ) - observed = { - "increment_calls": fake_cache.redis_cache.async_increment.call_count, - "in_memory_set_calls": fake_cache.in_memory_cache.set_cache.call_count, - "called": fake_cache.redis_cache.async_increment.called, - } - assert normalize(observed) == { - "increment_calls": 3, - "in_memory_set_calls": 3, - "called": True, + assert {item.counter_key for item in pending} == { + "spend:end_user:eu1", + "spend:tag:a", + "spend:tag:b", } + assert all(item.increment == 3.0 for item in pending) @pytest.mark.asyncio -async def test_increment_end_user_and_tag_spend_counters_no_end_user_no_tags_invalid_input_noop( +async def test_prepare_end_user_and_tag_spend_increments_no_end_user_no_tags_invalid_input_noop( monkeypatch, ): fake_cache = _make_spend_counter_cache() monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - await ps._increment_end_user_and_tag_spend_counters( + pending = await ps._prepare_end_user_and_tag_spend_increments( end_user_id=None, tags=None, response_cost=1.0, reserved_counter_keys=set(), ) + assert pending == () assert fake_cache.redis_cache.async_increment.called is False # --------------------------------------------------------------------------- -# _increment_org_spend_counter +# _prepare_org_spend_increment # --------------------------------------------------------------------------- @pytest.mark.asyncio -async def test_increment_org_spend_counter_increments_when_org_present(monkeypatch): - fake_cache = _make_spend_counter_cache( - redis_get_value=None, redis_increment_value=10.0 - ) +async def test_prepare_org_spend_increment_returns_pending_when_org_present(monkeypatch): + fake_cache = _make_spend_counter_cache(redis_get_value=1.0) fake_user_cache = _make_user_api_key_cache() monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) monkeypatch.setattr(ps, "prisma_client", None) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) - await ps._increment_org_spend_counter( + pending = await ps._prepare_org_spend_increment( org_id="org-1", response_cost=10.0, reserved_counter_keys=set(), ) - observed = { - "increment_called": fake_cache.redis_cache.async_increment.called, - "increment_calls": fake_cache.redis_cache.async_increment.call_count, - "counter_key_arg": fake_cache.redis_cache.async_increment.call_args.kwargs[ - "key" - ], - } - assert normalize(observed) == { - "increment_called": True, - "increment_calls": 1, - "counter_key_arg": "spend:org:org-1", - } + assert len(pending) == 1 + assert pending[0].counter_key == "spend:org:org-1" + assert pending[0].increment == 10.0 @pytest.mark.asyncio -async def test_increment_org_spend_counter_no_org_is_noop_invalid_id(monkeypatch): +async def test_prepare_org_spend_increment_no_org_is_noop_invalid_id(monkeypatch): fake_cache = _make_spend_counter_cache() monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - await ps._increment_org_spend_counter( + pending = await ps._prepare_org_spend_increment( org_id=None, response_cost=1.0, reserved_counter_keys=set(), ) + assert pending == () assert fake_cache.redis_cache.async_increment.called is False # --------------------------------------------------------------------------- -# _init_and_increment_unreserved_spend_counter +# _prepare_unreserved_spend_counter_increment # --------------------------------------------------------------------------- @pytest.mark.asyncio -async def test_init_and_increment_unreserved_spend_counter_skips_reserved_keys( +async def test_prepare_unreserved_spend_counter_increment_skips_reserved_keys( monkeypatch, ): fake_cache = _make_spend_counter_cache() monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - await ps._init_and_increment_unreserved_spend_counter( + pending = await ps._prepare_unreserved_spend_counter_increment( counter_key="spend:tag:x", source_cache_key="tag:x", increment=1.0, reserved_counter_keys={"spend:tag:x"}, ) + assert pending is None assert fake_cache.redis_cache.async_increment.called is False @pytest.mark.asyncio -async def test_init_and_increment_unreserved_spend_counter_proceeds_when_not_reserved( +async def test_prepare_unreserved_spend_counter_increment_proceeds_when_not_reserved( monkeypatch, ): - fake_cache = _make_spend_counter_cache( - redis_get_value=None, redis_increment_value=2.0 - ) + fake_cache = _make_spend_counter_cache(redis_get_value=None) fake_user_cache = _make_user_api_key_cache() + reseed = AsyncMock(return_value=None) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) monkeypatch.setattr(ps, "prisma_client", None) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", reseed) - await ps._init_and_increment_unreserved_spend_counter( + pending = await ps._prepare_unreserved_spend_counter_increment( counter_key="spend:tag:y", source_cache_key="tag:y", increment=2.0, reserved_counter_keys=set(), ) - observed = { - "increment_called": fake_cache.redis_cache.async_increment.called, - "redis_get_called": fake_cache.redis_cache.async_get_cache.called, - "reseed_consulted": True, - } - assert observed == { - "increment_called": True, - "redis_get_called": True, - "reseed_consulted": True, - } + assert pending is not None + assert pending.counter_key == "spend:tag:y" + assert pending.increment == 2.0 + assert fake_cache.redis_cache.async_get_cache.called is True + assert reseed.called is True # --------------------------------------------------------------------------- -# _init_and_increment_spend_counter +# _prepare_spend_counter_increment # --------------------------------------------------------------------------- @pytest.mark.asyncio -async def test_init_and_increment_spend_counter_warm_cache_skips_reseed(monkeypatch): - fake_cache = _make_spend_counter_cache( - redis_get_value=11.0, redis_increment_value=14.0 - ) +async def test_prepare_spend_counter_increment_warm_cache_skips_reseed(monkeypatch): + fake_cache = _make_spend_counter_cache(redis_get_value=11.0) fake_user_cache = _make_user_api_key_cache() monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) @@ -998,12 +1074,14 @@ async def test_init_and_increment_spend_counter_warm_cache_skips_reseed(monkeypa reseed = AsyncMock(return_value=None) monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", reseed) - await ps._init_and_increment_spend_counter( + pending = await ps._prepare_spend_counter_increment( counter_key="spend:key:k", source_cache_key="k", increment=3.0, ) + assert pending.counter_key == "spend:key:k" + assert pending.increment == 3.0 observed = { "reseed_called": reseed.called, "increment_called": fake_cache.redis_cache.async_increment.called, @@ -1011,23 +1089,21 @@ async def test_init_and_increment_spend_counter_warm_cache_skips_reseed(monkeypa } assert normalize(observed) == { "reseed_called": False, - "increment_called": True, + "increment_called": False, "in_memory_seeded_from_redis": True, } # --------------------------------------------------------------------------- -# _init_and_increment_window_spend_counter +# _prepare_window_spend_counter_increment # --------------------------------------------------------------------------- @pytest.mark.asyncio -async def test_init_and_increment_window_spend_counter_increments_when_initialized( +async def test_prepare_window_spend_counter_increment_returns_pending_when_initialized( monkeypatch, ): - fake_cache = _make_spend_counter_cache( - redis_get_value=0.0, redis_increment_value=5.0 - ) + fake_cache = _make_spend_counter_cache(redis_get_value=0.0) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "prisma_client", None) monkeypatch.setattr( @@ -1036,7 +1112,7 @@ async def test_init_and_increment_window_spend_counter_increments_when_initializ AsyncMock(return_value=0.0), ) - await ps._init_and_increment_window_spend_counter( + pending = await ps._prepare_window_spend_counter_increment( counter_key="spend:key:k:window:1d", entity_type="Key", entity_id="k", @@ -1045,26 +1121,19 @@ async def test_init_and_increment_window_spend_counter_increments_when_initializ increment=5.0, ) - observed = { - "redis_increment_called": fake_cache.redis_cache.async_increment.called, - "increment_calls": fake_cache.redis_cache.async_increment.call_count, - "in_memory_set_calls": fake_cache.in_memory_cache.set_cache.call_count, - } - assert normalize(observed) == { - "redis_increment_called": True, - "increment_calls": 1, - "in_memory_set_calls": 2, - } + assert pending is not None + assert pending.counter_key == "spend:key:k:window:1d" + assert pending.increment == 5.0 @pytest.mark.asyncio -async def test_init_and_increment_window_spend_counter_missing_window_start_invalid_skips( +async def test_prepare_window_spend_counter_increment_missing_window_start_invalid_skips( monkeypatch, ): fake_cache = _make_spend_counter_cache() monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - await ps._init_and_increment_window_spend_counter( + pending = await ps._prepare_window_spend_counter_increment( counter_key="spend:key:k:window:1d", entity_type="Key", entity_id="k", @@ -1073,6 +1142,7 @@ async def test_init_and_increment_window_spend_counter_missing_window_start_inva increment=5.0, ) + assert pending is None assert fake_cache.redis_cache.async_increment.called is False @@ -1114,16 +1184,12 @@ async def test_ensure_spend_counter_initialized_warm_skips_reseed_and_source( async def test_ensure_spend_counter_initialized_cold_seeds_from_source_cache( monkeypatch, ): - fake_cache = _make_spend_counter_cache( - redis_get_value=None, redis_increment_value=7.0 - ) + fake_cache = _make_spend_counter_cache(redis_get_value=None, redis_increment_value=7.0) fake_user_cache = _make_user_api_key_cache(get_value={"spend": 7.0}) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) monkeypatch.setattr(ps, "prisma_client", None) - monkeypatch.setattr( - ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None) - ) + monkeypatch.setattr(ps.SpendCounterReseed, "coalesced", AsyncMock(return_value=None)) await ps._ensure_spend_counter_initialized( counter_key="spend:user:u", @@ -1163,9 +1229,7 @@ async def test_get_source_cache_base_spend_reads_first_hit_from_list(monkeypatch fake_user_cache.async_get_cache = AsyncMock(side_effect=_get) monkeypatch.setattr(ps, "user_api_key_cache", fake_user_cache) - result = await ps._get_source_cache_base_spend( - source_cache_key=["miss", "hit-obj", "miss2"] - ) + result = await ps._get_source_cache_base_spend(source_cache_key=["miss", "hit-obj", "miss2"]) observed = { "result": result, @@ -1294,9 +1358,7 @@ async def test_increment_spend_counter_cache_redis_path_returns_new_value(monkey fake_cache = _make_spend_counter_cache(redis_increment_value=44.0) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) - result = await ps._increment_spend_counter_cache( - counter_key="spend:key:k", increment=4.0 - ) + result = await ps._increment_spend_counter_cache(counter_key="spend:key:k", increment=4.0) observed = { "result": result, @@ -1314,15 +1376,11 @@ async def test_increment_spend_counter_cache_redis_path_returns_new_value(monkey async def test_increment_spend_counter_cache_redis_error_raises_and_invalidates( monkeypatch, ): - fake_cache = _make_spend_counter_cache( - redis_increment_side_effect=RuntimeError("incr fail") - ) + fake_cache = _make_spend_counter_cache(redis_increment_side_effect=RuntimeError("incr fail")) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) with pytest.raises(RuntimeError): - await ps._increment_spend_counter_cache( - counter_key="spend:key:k", increment=1.0 - ) + await ps._increment_spend_counter_cache(counter_key="spend:key:k", increment=1.0) assert fake_cache.in_memory_cache.delete_cache.called is True assert fake_cache.redis_cache.async_delete_cache.called is True @@ -1343,9 +1401,7 @@ async def test_invalidate_spend_counter_deletes_in_memory_and_redis(monkeypatch) observed = { "in_memory_delete_called": fake_cache.in_memory_cache.delete_cache.called, "redis_delete_called": fake_cache.redis_cache.async_delete_cache.called, - "delete_args_key": fake_cache.redis_cache.async_delete_cache.call_args.kwargs[ - "key" - ], + "delete_args_key": fake_cache.redis_cache.async_delete_cache.call_args.kwargs["key"], } assert normalize(observed) == { "in_memory_delete_called": True, @@ -1357,9 +1413,7 @@ async def test_invalidate_spend_counter_deletes_in_memory_and_redis(monkeypatch) @pytest.mark.asyncio async def test_invalidate_spend_counter_swallows_redis_failure_no_raise(monkeypatch): fake_cache = _make_spend_counter_cache() - fake_cache.redis_cache.async_delete_cache = AsyncMock( - side_effect=RuntimeError("redis down") - ) + fake_cache.redis_cache.async_delete_cache = AsyncMock(side_effect=RuntimeError("redis down")) monkeypatch.setattr(ps, "spend_counter_cache", fake_cache) await ps._invalidate_spend_counter(counter_key="spend:key:k") diff --git a/tests/test_litellm/proxy/spend_tracking/test_budget_reservation_redis_failure.py b/tests/test_litellm/proxy/spend_tracking/test_budget_reservation_redis_failure.py index c123eeeed36..6165af4920d 100644 --- a/tests/test_litellm/proxy/spend_tracking/test_budget_reservation_redis_failure.py +++ b/tests/test_litellm/proxy/spend_tracking/test_budget_reservation_redis_failure.py @@ -41,6 +41,15 @@ class _FlakyRedisCache: self._store[key] = float(value) return True + async def async_increment_pipeline(self, increment_list, **kwargs): + results = [] + for op in increment_list: + results.append(await self.async_increment(op["key"], op["increment_value"])) + return results + + def get_ttl(self, **kwargs): + return None + @pytest.mark.asyncio async def test_direct_increment_runs_when_reservation_reconcile_hits_redis_failure( diff --git a/tests/test_litellm/proxy/test_budget_reservation.py b/tests/test_litellm/proxy/test_budget_reservation.py index 6dab054d8ea..40ebc03781c 100644 --- a/tests/test_litellm/proxy/test_budget_reservation.py +++ b/tests/test_litellm/proxy/test_budget_reservation.py @@ -2220,6 +2220,15 @@ class _ExpiringRedisCache: async def async_delete_cache(self, key: str, *args: object, **kwargs: object) -> None: self.store.pop(key, None) + async def async_increment_pipeline(self, increment_list, **kwargs): + results = [] + for op in increment_list: + results.append(await self.async_increment(op["key"], op["increment_value"])) + return results + + def get_ttl(self, **kwargs) -> None: + return None + @pytest.mark.asyncio async def test_reconcile_after_redis_counter_expiry_keeps_request_cost_enforced( diff --git a/tests/test_litellm/proxy/test_proxy_server.py b/tests/test_litellm/proxy/test_proxy_server.py index 54579e6cb7c..b0f38978727 100644 --- a/tests/test_litellm/proxy/test_proxy_server.py +++ b/tests/test_litellm/proxy/test_proxy_server.py @@ -7749,7 +7749,7 @@ async def test_increment_spend_counters_team_and_member(): @pytest.mark.asyncio -async def test_init_and_increment_spend_counter_reseeds_from_db_on_counter_miss(): +async def test_prepare_spend_counter_increment_reseeds_from_db_on_counter_miss(): """When the Redis counter is missing, the reseed path reads the authoritative spend from the DB (not a stale cache), so the next increment continues from the correct base value.""" @@ -7762,8 +7762,17 @@ async def test_init_and_increment_spend_counter_reseeds_from_db_on_counter_miss( recorded_increments.append({"key": key, "value": value, "ttl": ttl}) return value + async def record_pipeline(increment_list, **kwargs): + results = [] + for op in increment_list: + await record_increment(key=op["key"], value=op["increment_value"], ttl=op["ttl"]) + results.append(op["increment_value"]) + return results + fake_redis = AsyncMock() fake_redis.async_increment = AsyncMock(side_effect=record_increment) + fake_redis.async_increment_pipeline = AsyncMock(side_effect=record_pipeline) + fake_redis.get_ttl = MagicMock(return_value=None) fake_redis.async_get_cache = AsyncMock(return_value=None) # counter missing fake_redis.async_set_cache = AsyncMock(return_value=True) # SET NX wins counter_cache.redis_cache = fake_redis @@ -7782,7 +7791,10 @@ async def test_init_and_increment_spend_counter_reseeds_from_db_on_counter_miss( stale_cache.in_memory_cache.set_cache(key="team_id:team-9", value=stale_team) import litellm.proxy.proxy_server as ps - from litellm.proxy.proxy_server import _init_and_increment_spend_counter + from litellm.proxy.proxy_server import ( + _apply_spend_counter_increments, + _prepare_spend_counter_increment, + ) orig_user, orig_counter, orig_prisma = ( ps.user_api_key_cache, @@ -7793,11 +7805,12 @@ async def test_init_and_increment_spend_counter_reseeds_from_db_on_counter_miss( ps.spend_counter_cache = counter_cache ps.prisma_client = fake_prisma try: - await _init_and_increment_spend_counter( + pending = await _prepare_spend_counter_increment( counter_key="spend:team:team-9", source_cache_key="team_id:team-9", increment=1.5, ) + await _apply_spend_counter_increments(pending=(pending,)) fake_prisma.db.litellm_teamtable.find_unique.assert_awaited_once_with(where={"team_id": "team-9"}) # Seed uses SET NX with db_spend (42) — cross-pod safe, no INCR of 42. @@ -7976,7 +7989,10 @@ async def test_reseed_spend_from_db_skips_window_variant_keys(): @pytest.mark.asyncio async def test_window_spend_counter_reseeds_from_spend_logs_on_counter_miss(): from litellm.caching.dual_cache import DualCache - from litellm.proxy.proxy_server import _init_and_increment_window_spend_counter + from litellm.proxy.proxy_server import ( + _apply_spend_counter_increments, + _prepare_window_spend_counter_increment, + ) counter_cache = DualCache() window_start = datetime.now(timezone.utc) - timedelta(hours=1) @@ -7992,7 +8008,7 @@ async def test_window_spend_counter_reseeds_from_spend_logs_on_counter_miss(): ps.spend_counter_cache = counter_cache ps.prisma_client = fake_prisma try: - await _init_and_increment_window_spend_counter( + pending = await _prepare_window_spend_counter_increment( counter_key="spend:key:key-window:window:1h", entity_type="Key", entity_id="key-window", @@ -8000,6 +8016,7 @@ async def test_window_spend_counter_reseeds_from_spend_logs_on_counter_miss(): window_start=window_start, increment=0.5, ) + await _apply_spend_counter_increments(pending=(pending,) if pending is not None else ()) fake_prisma.db.litellm_spendlogs.group_by.assert_awaited_once_with( by=["api_key"], @@ -8015,7 +8032,10 @@ async def test_window_spend_counter_reseeds_from_spend_logs_on_counter_miss(): @pytest.mark.asyncio async def test_init_spend_counter_redis_clean_miss_skips_stale_in_memory(): from litellm.caching.dual_cache import DualCache - from litellm.proxy.proxy_server import _init_and_increment_spend_counter + from litellm.proxy.proxy_server import ( + _apply_spend_counter_increments, + _prepare_spend_counter_increment, + ) counter_cache = DualCache() counter_key = "spend:team:team-stale-local" @@ -8037,6 +8057,15 @@ async def test_init_spend_counter_redis_clean_miss_skips_stale_in_memory(): fake_redis.async_get_cache = AsyncMock(return_value=None) fake_redis.async_increment = AsyncMock(side_effect=redis_increment) fake_redis.async_set_cache = AsyncMock(side_effect=redis_set_cache) + + async def redis_increment_pipeline(increment_list, **_): + results = [] + for op in increment_list: + results.append(await redis_increment(key=op["key"], value=op["increment_value"])) + return results + + fake_redis.async_increment_pipeline = AsyncMock(side_effect=redis_increment_pipeline) + fake_redis.get_ttl = MagicMock(return_value=None) counter_cache.redis_cache = fake_redis db_row = MagicMock() @@ -8055,11 +8084,12 @@ async def test_init_spend_counter_redis_clean_miss_skips_stale_in_memory(): ps.prisma_client = fake_prisma ps.user_api_key_cache = DualCache() try: - await _init_and_increment_spend_counter( + pending = await _prepare_spend_counter_increment( counter_key=counter_key, source_cache_key="team_id:team-stale-local", increment=1.5, ) + await _apply_spend_counter_increments(pending=(pending,)) fake_prisma.db.litellm_teamtable.find_unique.assert_awaited_once_with(where={"team_id": "team-stale-local"}) # Seed via SET NX (42) + delta via INCRBYFLOAT (1.5) = 43.5. @@ -8074,7 +8104,10 @@ async def test_init_spend_counter_redis_clean_miss_skips_stale_in_memory(): @pytest.mark.asyncio async def test_window_spend_counter_redis_clean_miss_skips_stale_in_memory(): from litellm.caching.dual_cache import DualCache - from litellm.proxy.proxy_server import _init_and_increment_window_spend_counter + from litellm.proxy.proxy_server import ( + _apply_spend_counter_increments, + _prepare_window_spend_counter_increment, + ) counter_cache = DualCache() counter_key = "spend:key:key-window-stale-local:window:1h" @@ -8097,6 +8130,15 @@ async def test_window_spend_counter_redis_clean_miss_skips_stale_in_memory(): fake_redis.async_get_cache = AsyncMock(return_value=None) fake_redis.async_set_cache = AsyncMock(side_effect=redis_set_cache) fake_redis.async_increment = AsyncMock(side_effect=redis_increment) + + async def redis_increment_pipeline(increment_list, **_): + results = [] + for op in increment_list: + results.append(await redis_increment(key=op["key"], value=op["increment_value"])) + return results + + fake_redis.async_increment_pipeline = AsyncMock(side_effect=redis_increment_pipeline) + fake_redis.get_ttl = MagicMock(return_value=None) counter_cache.redis_cache = fake_redis fake_prisma = MagicMock() @@ -8111,7 +8153,7 @@ async def test_window_spend_counter_redis_clean_miss_skips_stale_in_memory(): ps.spend_counter_cache = counter_cache ps.prisma_client = fake_prisma try: - await _init_and_increment_window_spend_counter( + pending = await _prepare_window_spend_counter_increment( counter_key=counter_key, entity_type="Key", entity_id="key-window-stale-local", @@ -8119,6 +8161,7 @@ async def test_window_spend_counter_redis_clean_miss_skips_stale_in_memory(): window_start=window_start, increment=0.5, ) + await _apply_spend_counter_increments(pending=(pending,) if pending is not None else ()) fake_prisma.db.litellm_spendlogs.group_by.assert_awaited_once_with( by=["api_key"], @@ -8138,7 +8181,10 @@ async def test_window_spend_counter_redis_clean_miss_skips_stale_in_memory(): @pytest.mark.asyncio async def test_window_spend_counter_redis_concurrent_seed_does_not_double_seed(): from litellm.caching.dual_cache import DualCache - from litellm.proxy.proxy_server import _init_and_increment_window_spend_counter + from litellm.proxy.proxy_server import ( + _apply_spend_counter_increments, + _prepare_window_spend_counter_increment, + ) counter_cache = DualCache() counter_key = "spend:key:key-window-concurrent-seed:window:1h" @@ -8161,6 +8207,15 @@ async def test_window_spend_counter_redis_concurrent_seed_does_not_double_seed() fake_redis.async_get_cache = AsyncMock(side_effect=redis_get_cache) fake_redis.async_set_cache = AsyncMock(return_value=False) fake_redis.async_increment = AsyncMock(side_effect=redis_increment) + + async def redis_increment_pipeline(increment_list, **_): + results = [] + for op in increment_list: + results.append(await redis_increment(key=op["key"], value=op["increment_value"])) + return results + + fake_redis.async_increment_pipeline = AsyncMock(side_effect=redis_increment_pipeline) + fake_redis.get_ttl = MagicMock(return_value=None) counter_cache.redis_cache = fake_redis fake_prisma = MagicMock() @@ -8175,7 +8230,7 @@ async def test_window_spend_counter_redis_concurrent_seed_does_not_double_seed() ps.spend_counter_cache = counter_cache ps.prisma_client = fake_prisma try: - await _init_and_increment_window_spend_counter( + pending = await _prepare_window_spend_counter_increment( counter_key=counter_key, entity_type="Key", entity_id="key-window-concurrent-seed", @@ -8183,6 +8238,7 @@ async def test_window_spend_counter_redis_concurrent_seed_does_not_double_seed() window_start=window_start, increment=0.5, ) + await _apply_spend_counter_increments(pending=(pending,) if pending is not None else ()) fake_redis.async_set_cache.assert_awaited_once_with( key=counter_key, @@ -8199,7 +8255,7 @@ async def test_window_spend_counter_redis_concurrent_seed_does_not_double_seed() @pytest.mark.asyncio async def test_window_spend_counter_skips_invalid_window_start(): from litellm.caching.dual_cache import DualCache - from litellm.proxy.proxy_server import _init_and_increment_window_spend_counter + from litellm.proxy.proxy_server import _prepare_window_spend_counter_increment counter_cache = DualCache() @@ -8208,7 +8264,7 @@ async def test_window_spend_counter_skips_invalid_window_start(): orig_counter = ps.spend_counter_cache ps.spend_counter_cache = counter_cache try: - await _init_and_increment_window_spend_counter( + pending = await _prepare_window_spend_counter_increment( counter_key="spend:key:key-invalid-window:window:not-a-duration", entity_type="Key", entity_id="key-invalid-window", @@ -8216,6 +8272,7 @@ async def test_window_spend_counter_skips_invalid_window_start(): window_start=None, increment=0.5, ) + assert pending is None assert counter_cache.in_memory_cache.get_cache(key="spend:key:key-invalid-window:window:not-a-duration") is None finally: @@ -8279,6 +8336,9 @@ async def test_increment_spend_counters_finalizes_after_unreserved_increments(): async def assert_reservation_not_finalized_yet(**kwargs): assert budget_reservation["finalized"] is False incremented_counters.append(kwargs["counter_key"]) + return ps._PendingSpendIncrement( + counter_key=kwargs["counter_key"], increment=kwargs["increment"] + ) import litellm.proxy.proxy_server as ps @@ -8287,7 +8347,7 @@ async def test_increment_spend_counters_finalizes_after_unreserved_increments(): ps.user_api_key_cache = DualCache() try: with patch( - "litellm.proxy.proxy_server._init_and_increment_spend_counter", + "litellm.proxy.proxy_server._prepare_spend_counter_increment", new=AsyncMock(side_effect=assert_reservation_not_finalized_yet), ): await increment_spend_counters( @@ -8620,7 +8680,7 @@ async def test_get_current_spend_uses_db_zero_over_stale_fallback(): async def test_concurrent_read_and_write_paths_share_one_db_query(): """ The read path (`get_current_spend`) and the write path - (`_init_and_increment_spend_counter`) both reseed cold counters from + (`_prepare_spend_counter_increment`) both reseed cold counters from the DB. They must share the per-counter lock so a concurrent pre-call enforcement read and post-call increment for the same counter collapse to one DB query, not two. @@ -8629,7 +8689,7 @@ async def test_concurrent_read_and_write_paths_share_one_db_query(): from litellm.caching.dual_cache import DualCache from litellm.proxy.proxy_server import ( - _init_and_increment_spend_counter, + _prepare_spend_counter_increment, get_current_spend, ) @@ -8683,7 +8743,7 @@ async def test_concurrent_read_and_write_paths_share_one_db_query(): try: results = await _asyncio.gather( get_current_spend(counter_key=counter_key, fallback_spend=0.0), - _init_and_increment_spend_counter( + _prepare_spend_counter_increment( counter_key=counter_key, source_cache_key="ignored", increment=1.5, From 7c6e33ef70889445fab5650b1490ddf3c0448a11 Mon Sep 17 00:00:00 2001 From: "devin-ai-integration[bot]" <158243242+devin-ai-integration[bot]@users.noreply.github.com> Date: Wed, 9 Sep 2026 10:38:26 -0700 Subject: [PATCH 21/26] fix(proxy): ignore team_id="" on /key/update so team-less keys can be updated and imported (#40421) Co-authored-by: yassin Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/proxy/_types.py | 8 ++++++ .../test_key_management_endpoints.py | 26 +++++++++++++++++++ 2 files changed, 34 insertions(+) diff --git a/litellm/proxy/_types.py b/litellm/proxy/_types.py index df7ee8f508c..0e2fd50ac9f 100644 --- a/litellm/proxy/_types.py +++ b/litellm/proxy/_types.py @@ -3,6 +3,7 @@ import json import os from collections.abc import Callable, Mapping from datetime import datetime +from types import MappingProxyType from typing import TYPE_CHECKING, Any, Final, Literal, NamedTuple import httpx @@ -1294,6 +1295,13 @@ class UpdateKeyRequest(KeyRequestBase): rotation_interval: str | None = None organization_id: str | None = None + @model_validator(mode="before") + @classmethod + def drop_blank_team_id(cls, values: object) -> object: + if isinstance(values, Mapping) and values.get("team_id") == "": + return MappingProxyType({k: v for k, v in values.items() if k != "team_id"}) + return values + @field_validator("organization_id", mode="before") @classmethod def treat_cleared_organization_id_as_unset(cls, v: object) -> object: diff --git a/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py index a873a367eab..d2aeba18f7d 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_key_management_endpoints.py @@ -17975,6 +17975,32 @@ def test_key_request_blank_organization_id_is_unset(): assert UpdateKeyRequest(key="sk-1", organization_id="org-1").organization_id == "org-1" +def test_update_key_request_blank_team_id_is_not_a_team_change(): + from litellm.proxy._types import UpdateKeyRequest + from litellm.proxy.management_endpoints.key_management_endpoints import ( + is_different_team, + ) + + blank = UpdateKeyRequest(key="sk-1", team_id="", key_alias="renamed") + assert blank.team_id is None + assert "team_id" not in blank.model_dump(exclude_unset=True) + assert blank.model_dump(exclude_unset=True) == {"key": "sk-1", "key_alias": "renamed"} + assert is_different_team(data=blank, existing_key_row=LiteLLM_VerificationToken(token="hashed")) is False + assert ( + is_different_team(data=blank, existing_key_row=LiteLLM_VerificationToken(token="hashed", team_id="team-1")) + is False + ) + assert "team_id" in UpdateKeyRequest(key="sk-1", team_id=None).model_dump(exclude_unset=True) + assert UpdateKeyRequest(key="sk-1", team_id="team-1").team_id == "team-1" + assert ( + is_different_team( + data=UpdateKeyRequest(key="sk-1", team_id="team-1"), + existing_key_row=LiteLLM_VerificationToken(token="hashed"), + ) + is True + ) + + def test_key_generation_check_blank_team_id_uses_personal_permissions(monkeypatch): """key_generation_check with team_id="" must take the personal-key path instead of failing the team lookup with "Unable to find team object" (LIT-3925).""" From 699ae63b2a3270ab8b1068f8beba4a3bf71543b9 Mon Sep 17 00:00:00 2001 From: Clement Date: Thu, 10 Sep 2026 01:47:34 +0800 Subject: [PATCH 22/26] feat(router): support percentile-based TTFT routing (#40352) * feat(router): support percentile-based TTFT routing * fix(router): apply routing_strategy_args updates to the live selector Runtime routing_strategy_args updates (config reload, update_settings) only rebuilt the strategy selector when routing_strategy itself changed, so a newly added ttft_percentile sat unused until the proxy restarted. Also drops a comment that only restated the code it sat above. Claude-Session: https://claude.ai/code/session_01PmqjhFYcUh6vA72d8W9gdB * refactor(router): drop unreachable empty-samples guard in percentile latency _percentile_latency is only called behind use_ttft, which already requires a non-empty ttft sample list, so the early return was dead code and the one line Codecov flagged as uncovered on this patch. Claude-Session: https://claude.ai/code/session_01PmqjhFYcUh6vA72d8W9gdB * test(router): cover the no-selector path of a routing_strategy_args update simple-shuffle has no selector attribute to re-link, so the early return guards a setattr with a None attribute name. Dropping the guard makes the new test fail with "attribute name must be string, not 'NoneType'". Claude-Session: https://claude.ai/code/session_01PmqjhFYcUh6vA72d8W9gdB * fix(test): assert ValidationError on out-of-range ttft_percentile pytest.raises(ValueError) tripped PT011 for being too broad. Pydantic raises ValidationError for the gt/le constraint, so naming it satisfies the rule and pins the assertion to the constraint under test. Claude-Session: https://claude.ai/code/session_01PmqjhFYcUh6vA72d8W9gdB * fix(router): drop Final from a per-deployment loop variable basedpyright rejects "A Final variable cannot be assigned within a loop", which pushed reportGeneralTypeIssues one over its budget. selected_latency is rebound each iteration, so it matches its unannotated neighbours in the same loop. Claude-Session: https://claude.ai/code/session_01PmqjhFYcUh6vA72d8W9gdB * test(router): exempt _apply_updated_routing_strategy_args from the name scan The scan only reads test files with "router" in the filename, so it cannot see the update_settings tests in router_strategy/test_lowest_latency.py. Calling the private helper directly would test structure rather than behaviour, so it joins the existing entries ignored for the same reason. Claude-Session: https://claude.ai/code/session_01PmqjhFYcUh6vA72d8W9gdB --- litellm/router.py | 47 ++++- litellm/router_strategy/lowest_latency.py | 21 ++- .../router_code_coverage.py | 1 + .../router_strategy/test_lowest_latency.py | 174 +++++++++++++++++- 4 files changed, 225 insertions(+), 18 deletions(-) diff --git a/litellm/router.py b/litellm/router.py index 934a4ac86a9..2fbdf541487 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -1317,6 +1317,43 @@ class Router: if isinstance(litellm.input_callback, list): litellm.input_callback = [c for c in litellm.input_callback if id(c) not in selector_ids] + def _apply_updated_routing_strategy_args(self) -> None: + """ + Re-link the default group's selector to the current `routing_strategy_args`. + + Selectors freeze their `RoutingArgs` at construction, so a runtime args + update would otherwise keep serving the boot-time values until restart. + Latency/usage state survives the rebuild: it lives in the shared router + cache, not on the selector. + """ + strategy: Final = self._normalize_strategy(self.routing_strategy) + if strategy == "lar1": + from litellm.router_strategy.lar1_routing import apply_lar1_routing_strategy + + apply_lar1_routing_strategy(self, self.routing_strategy_args) + return + + attr: Final = self._DEFAULT_SELECTOR_ATTR_BY_STRATEGY.get(strategy or "") + current: Final = getattr(self, attr, None) if attr is not None else None + if attr is None or current is None: + return + + try: + rebuilt: Final = self._build_strategy_selector( + strategy=strategy or "", + routing_strategy_args=self.routing_strategy_args, + ) + except (TypeError, ValidationError): + verbose_router_logger.exception( + "Invalid routing_strategy_args %s for '%s'; keeping the previous ones", + self.routing_strategy_args, + strategy, + ) + return + + self._unregister_router_selectors((current,)) + setattr(self, attr, rebuilt) + def routing_strategy_init(self, routing_strategy: RoutingStrategy | str, routing_strategy_args: dict): verbose_router_logger.info("Routing strategy: %s", routing_strategy) self._validate_routing_strategy(routing_strategy) @@ -11847,7 +11884,7 @@ class Router: _existing_router_settings: Final = self.get_settings() rebuild_routing_groups = False - relink_lar1_from_args = False + routing_args_updated = False for var in kwargs: if var in RUNTIME_UPDATABLE_ROUTER_SETTINGS: if var in _int_settings: @@ -11886,15 +11923,13 @@ class Router: ) rebuild_routing_groups = True elif var == "routing_strategy_args": - relink_lar1_from_args = True + routing_args_updated = True setattr(self, var, value) else: verbose_router_logger.debug("Setting %s is not allowed", var) - if relink_lar1_from_args and self._normalize_strategy(self.routing_strategy) == "lar1": - from litellm.router_strategy.lar1_routing import apply_lar1_routing_strategy - - apply_lar1_routing_strategy(self, self.routing_strategy_args) + if routing_args_updated: + self._apply_updated_routing_strategy_args() if rebuild_routing_groups: self._init_routing_groups(self._routing_groups_input) diff --git a/litellm/router_strategy/lowest_latency.py b/litellm/router_strategy/lowest_latency.py index 805d4ff9080..e902192811c 100644 --- a/litellm/router_strategy/lowest_latency.py +++ b/litellm/router_strategy/lowest_latency.py @@ -3,8 +3,11 @@ import random from collections.abc import Sequence from datetime import datetime, timedelta +from math import ceil from typing import TYPE_CHECKING, Any, Final +from pydantic import Field + import litellm from litellm import ModelResponse, token_counter, verbose_logger from litellm.caching.caching import DualCache @@ -24,6 +27,7 @@ class RoutingArgs(LiteLLMPydanticObjectBase): ttl: float = 1 * 60 * 60 # 1 hour lowest_latency_buffer: float = 0 max_latency_list_size: int = 10 + ttft_percentile: float | None = Field(default=None, gt=0, le=1) def _average_latency(samples: Sequence[float]) -> float: @@ -32,6 +36,12 @@ def _average_latency(samples: Sequence[float]) -> float: return sum(samples) / len(samples) +def _percentile_latency(samples: Sequence[float], percentile: float) -> float: + values: Final = sorted(samples) + index: Final = ceil(len(values) * percentile) - 1 + return values[index] + + def _ttft_seconds(elapsed: timedelta | float) -> float: if isinstance(elapsed, timedelta): return elapsed.total_seconds() @@ -427,14 +437,17 @@ class LowestLatencyLoggingHandler(CustomLogger): item_rpm = item_map.get(precise_minute, {}).get("rpm", 0) item_tpm = item_map.get(precise_minute, {}).get("tpm", 0) - # get average latency or average ttft (depending on streaming/non-streaming) use_ttft = ( request_kwargs is not None and request_kwargs.get("stream", None) is not None and request_kwargs["stream"] is True and len(item_ttft_latency) > 0 ) - average_latency = _average_latency(item_ttft_latency if use_ttft else item_latency) + selected_latency = ( + _percentile_latency(item_ttft_latency, self.routing_args.ttft_percentile) + if use_ttft and self.routing_args.ttft_percentile is not None + else _average_latency(item_ttft_latency if use_ttft else item_latency) + ) # -------------- # # Debugging Logic @@ -443,7 +456,7 @@ class LowestLatencyLoggingHandler(CustomLogger): # this helps a user to debug why the router picked a specfic deployment # _deployment_api_base = _deployment.get("litellm_params", {}).get("api_base", "") if _deployment_api_base is not None: - _latency_per_deployment[_deployment_api_base] = average_latency + _latency_per_deployment[_deployment_api_base] = selected_latency # -------------- # # End of Debugging Logic # -------------- # @@ -453,7 +466,7 @@ class LowestLatencyLoggingHandler(CustomLogger): ): # if user passed in tpm / rpm in the model_list continue else: - potential_deployments.append((_deployment, average_latency)) + potential_deployments.append((_deployment, selected_latency)) if len(potential_deployments) == 0: return None diff --git a/tests/code_coverage_tests/router_code_coverage.py b/tests/code_coverage_tests/router_code_coverage.py index 0af29f069c6..a11f015743b 100644 --- a/tests/code_coverage_tests/router_code_coverage.py +++ b/tests/code_coverage_tests/router_code_coverage.py @@ -87,6 +87,7 @@ ignored_function_names = [ "_delete_claude_code_session_router_binding", # Tested through Redis cleanup failure in test_router.py "_resolve_claude_code_session_router", # Tested through Claude Code session routing in test_router.py "_get_claude_code_session_router_binding", # Tested through the two-worker session routing test in test_router.py + "_apply_updated_routing_strategy_args", # Tested via update_settings in test_lowest_latency.py (file lacks "router" in name) ] diff --git a/tests/test_litellm/router_strategy/test_lowest_latency.py b/tests/test_litellm/router_strategy/test_lowest_latency.py index 812d7bbff32..1a8614e3fca 100644 --- a/tests/test_litellm/router_strategy/test_lowest_latency.py +++ b/tests/test_litellm/router_strategy/test_lowest_latency.py @@ -8,11 +8,12 @@ import json from datetime import datetime, timedelta import pytest - +from pydantic import ValidationError import litellm from litellm.caching.caching import DualCache -from litellm.router_strategy.lowest_latency import LowestLatencyLoggingHandler +from litellm.router import Router +from litellm.router_strategy.lowest_latency import LowestLatencyLoggingHandler, RoutingArgs DEPLOYMENT_ID = "9876" KWARGS = { @@ -58,9 +59,9 @@ def test_sync_embedding_latency_is_json_serializable(): latencies = _recorded_latencies(cache) assert latencies, "expected a latency entry to be recorded" - assert all( - not isinstance(value, timedelta) for value in latencies - ), f"raw timedelta leaked into latency list: {latencies}" + assert all(not isinstance(value, timedelta) for value in latencies), ( + f"raw timedelta leaked into latency list: {latencies}" + ) assert latencies[-1] == pytest.approx(2.0) # the exact failure mode from production: redis cache sync json.dumps json.dumps({"latency": latencies}) @@ -84,9 +85,9 @@ async def test_async_embedding_latency_is_json_serializable(): latencies = _recorded_latencies(cache) assert latencies, "expected a latency entry to be recorded" - assert all( - not isinstance(value, timedelta) for value in latencies - ), f"raw timedelta leaked into latency list: {latencies}" + assert all(not isinstance(value, timedelta) for value in latencies), ( + f"raw timedelta leaked into latency list: {latencies}" + ) assert latencies[-1] == pytest.approx(3.0) json.dumps({"latency": latencies}) @@ -292,6 +293,85 @@ async def test_streaming_routing_ignores_per_token_ttft_samples_from_older_worke assert picked["model_info"]["id"] == FAST_TTFT_ID +@pytest.mark.asyncio +@pytest.mark.parametrize("sync_mode", [True, False], ids=["sync", "async"]) +@pytest.mark.parametrize( + ("ttft_percentile", "first_samples", "second_samples", "expected_id"), + [ + (None, [0.1, 0.1, 1.0], [0.3, 0.3, 0.3], SLOW_TTFT_ID), + (0.5, [0.1, 0.1, 1.0], [0.3, 0.3, 0.3], FAST_TTFT_ID), + (0.9, [0.1, 0.1, 0.1, 0.1, 1.5], [0.3, 0.3, 0.3, 0.3, 0.3], SLOW_TTFT_ID), + ], + ids=["default_average", "p50", "p90"], +) +async def test_streaming_ttft_ranking_percentile( + sync_mode: bool, + ttft_percentile: float | None, + first_samples: list[float], + second_samples: list[float], + expected_id: str, +): + cache = DualCache() + routing_args = {} if ttft_percentile is None else {"ttft_percentile": ttft_percentile} + handler = LowestLatencyLoggingHandler(router_cache=cache, routing_args=routing_args) + cache.set_cache( + key=f"{MODEL_GROUP}_map", + value={ + FAST_TTFT_ID: {"time_to_first_token_seconds": first_samples}, + SLOW_TTFT_ID: {"time_to_first_token_seconds": second_samples}, + }, + ) + + if sync_mode: + picked = handler.get_available_deployments( + model_group=MODEL_GROUP, + healthy_deployments=STREAMING_DEPLOYMENTS, + request_kwargs={"stream": True, "metadata": {}}, + ) + else: + picked = await handler.async_get_available_deployments( + model_group=MODEL_GROUP, + healthy_deployments=STREAMING_DEPLOYMENTS, + request_kwargs={"stream": True, "metadata": {}}, + ) + + assert picked is not None + assert picked["model_info"]["id"] == expected_id + + +@pytest.mark.parametrize("ttft_percentile", [0, -0.1, 1.1]) +def test_ttft_percentile_validation(ttft_percentile: float): + with pytest.raises(ValidationError): + RoutingArgs(ttft_percentile=ttft_percentile) + + +@pytest.mark.parametrize("ttft_percentile", [0.5, 0.9, 0.95, 1.0]) +def test_ttft_percentile_accepts_valid_values(ttft_percentile: float): + assert RoutingArgs(ttft_percentile=ttft_percentile).ttft_percentile == ttft_percentile + + +@pytest.mark.asyncio +async def test_ttft_percentile_does_not_change_non_streaming_routing(): + cache = DualCache() + handler = LowestLatencyLoggingHandler(router_cache=cache, routing_args={"ttft_percentile": 0.9}) + cache.set_cache( + key=f"{MODEL_GROUP}_map", + value={ + FAST_TTFT_ID: {"latency": [1.0], "time_to_first_token_seconds": [0.1]}, + SLOW_TTFT_ID: {"latency": [0.2], "time_to_first_token_seconds": [1.5]}, + }, + ) + + picked = await handler.async_get_available_deployments( + model_group=MODEL_GROUP, + healthy_deployments=STREAMING_DEPLOYMENTS, + request_kwargs={"stream": False, "metadata": {}}, + ) + + assert picked is not None + assert picked["model_info"]["id"] == SLOW_TTFT_ID + + @pytest.mark.asyncio @pytest.mark.parametrize( "cached_entry", @@ -318,3 +398,81 @@ async def test_async_get_available_deployments_treats_missing_samples_as_zero_la assert picked is not None assert picked["model_info"]["id"] == DEPLOYMENT_ID + + +def _latency_router(routing_strategy_args: dict) -> Router: + return Router( + model_list=[ + { + "model_name": MODEL_GROUP, + "litellm_params": {"model": f"openai/{MODEL_GROUP}", "api_key": "sk-fake"}, + "model_info": {"id": deployment_id}, + } + for deployment_id in (FAST_TTFT_ID, SLOW_TTFT_ID) + ], + routing_strategy="latency-based-routing", + routing_strategy_args=routing_strategy_args, + ) + + +def _seed_streaming_ttft(router: Router) -> None: + router.cache.set_cache( + key=f"{MODEL_GROUP}_map", + value={ + FAST_TTFT_ID: {"time_to_first_token_seconds": [0.1, 0.1, 1.0]}, + SLOW_TTFT_ID: {"time_to_first_token_seconds": [0.3, 0.3, 0.3]}, + }, + ) + + +async def _pick_streaming(router: Router) -> str: + picked = await router.async_get_available_deployment( + model=MODEL_GROUP, + request_kwargs={"stream": True, "metadata": {}}, + ) + return picked["model_info"]["id"] + + +@pytest.mark.asyncio +async def test_runtime_routing_strategy_args_update_applies_ttft_percentile(): + """A config reload that adds ttft_percentile must reach the live selector, + not sit unused until the proxy restarts.""" + router = _latency_router({"max_latency_list_size": 50}) + _seed_streaming_ttft(router) + + assert await _pick_streaming(router) == SLOW_TTFT_ID + + router.update_settings(routing_strategy_args={"max_latency_list_size": 50, "ttft_percentile": 0.5}) + + assert await _pick_streaming(router) == FAST_TTFT_ID + + +@pytest.mark.asyncio +async def test_runtime_routing_strategy_args_update_keeps_previous_args_when_invalid(): + router = _latency_router({"ttft_percentile": 0.5}) + _seed_streaming_ttft(router) + + router.update_settings(routing_strategy_args={"ttft_percentile": 5}) + + assert await _pick_streaming(router) == FAST_TTFT_ID + + +@pytest.mark.asyncio +async def test_runtime_routing_strategy_args_update_is_a_noop_without_a_selector(): + """simple-shuffle has no selector to re-link, so an args update must leave + the router alone instead of blowing up on a missing selector attribute.""" + router = Router( + model_list=[ + { + "model_name": MODEL_GROUP, + "litellm_params": {"model": f"openai/{MODEL_GROUP}", "api_key": "sk-fake"}, + "model_info": {"id": FAST_TTFT_ID}, + } + ], + routing_strategy="simple-shuffle", + ) + + router.update_settings(routing_strategy_args={"ttl": 5}) + + assert router.routing_strategy_args == {"ttl": 5} + assert await _pick_streaming(router) == FAST_TTFT_ID From c7163a80ddfd34da743b06d770d60aba05b292eb Mon Sep 17 00:00:00 2001 From: "devin-ai-integration[bot]" <158243242+devin-ai-integration[bot]@users.noreply.github.com> Date: Wed, 9 Sep 2026 10:48:06 -0700 Subject: [PATCH 23/26] perf(proxy): collapse per-worker SGR upserts into one statement per flush (#40362) Each proxy worker flushed one Prisma upsert per active (date, category, route) bucket every interval, so the Postgres primary saw workers x routes statements per interval across the deployment. A flush now builds a single multi-row INSERT ... ON CONFLICT DO UPDATE, and with use_redis_transaction_buffer on the workers push snapshots to a Redis list that one lease-holding pod folds and commits, so the whole deployment costs one statement per interval. The leader keeps popping until the list is empty so a deployment wider than the dequeue cap cannot build a backlog, and rows that fail both the commit and the Redis re-queue fall back to the leader's own accumulator instead of being lost. Resolves LIT-7371 Co-authored-by: yassin Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 1 + litellm/proxy/db/gateway_request_tracking.py | 220 ++++++++-- litellm/proxy/proxy_server.py | 14 +- .../proxy/db/test_gateway_request_tracking.py | 391 +++++++++++++++--- 4 files changed, 537 insertions(+), 89 deletions(-) diff --git a/litellm/constants.py b/litellm/constants.py index 108a914e9c1..f9389d22dea 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -335,6 +335,7 @@ DEFAULT_SSL_CIPHERS: Final = os.getenv( ########### v2 Architecture constants for managing writing updates to the database ########### REDIS_UPDATE_BUFFER_KEY: Final = "litellm_spend_update_buffer" +REDIS_GATEWAY_REQUESTS_BUFFER_KEY: Final = "litellm_gateway_requests_buffer" REDIS_DAILY_SPEND_UPDATE_BUFFER_KEY: Final = "litellm_daily_spend_update_buffer" REDIS_DAILY_TEAM_SPEND_UPDATE_BUFFER_KEY: Final = "litellm_daily_team_spend_update_buffer" REDIS_DAILY_ORG_SPEND_UPDATE_BUFFER_KEY: Final = "litellm_daily_org_spend_update_buffer" diff --git a/litellm/proxy/db/gateway_request_tracking.py b/litellm/proxy/db/gateway_request_tracking.py index bebd74e877c..c9ace68db33 100644 --- a/litellm/proxy/db/gateway_request_tracking.py +++ b/litellm/proxy/db/gateway_request_tracking.py @@ -10,13 +10,28 @@ strings rather than passing the raw path through. Nothing a caller sends can add a key, so the fold and the table it commits to are bounded by (days x routes) however much traffic arrives, and the response path carries no unbounded queue that would block once full. + +A flush commits its whole snapshot as one multi-row ``INSERT ... ON CONFLICT DO +UPDATE`` rather than one upsert per key, so a worker costs the primary one +statement per interval however many routes it served. With +``use_redis_transaction_buffer`` on, workers instead push their snapshot to a +Redis list and one lock-holding pod folds every entry and writes the table, so +the deployment as a whole costs the primary one statement per interval. """ -from dataclasses import asdict +import json +from collections.abc import AsyncIterator, Iterable from datetime import datetime, timezone -from typing import TYPE_CHECKING, Final +from itertools import chain +from types import MappingProxyType +from typing import TYPE_CHECKING, Final, TypeAlias + +from pydantic import TypeAdapter from litellm._logging import verbose_proxy_logger +from litellm.caching import RedisCache +from litellm.constants import MAX_REDIS_BUFFER_DEQUEUE_COUNT, REDIS_GATEWAY_REQUESTS_BUFFER_KEY +from litellm.proxy.db.db_transaction_queue.pod_lock_manager import PodLockManager from litellm.proxy.middleware.billable_request_metrics_middleware import BillableCategory from litellm.types.proxy.gateway_requests import ( GatewayRequestCounts, @@ -28,6 +43,15 @@ if TYPE_CHECKING: from litellm.proxy.utils import PrismaClient _EMPTY: Final = GatewayRequestCounts(successful_requests=0, failed_requests=0) +_TABLE: Final = '"LiteLLM_DailyGatewayRequests"' +_COLUMNS_PER_ROW: Final = 5 +_UTC_NOW: Final = "(NOW() AT TIME ZONE 'UTC')" +GATEWAY_REQUESTS_JOB_NAME: Final = "update_gateway_requests_job" + +_BufferedRows: TypeAlias = tuple[tuple[str, str, str, int, int], ...] +_BUFFERED_ROWS: Final = TypeAdapter(_BufferedRows) +_BUFFERED_ENTRIES: Final = TypeAdapter(tuple[str | bytes, ...]) +_NO_COUNTS: Final[GatewayRequestSnapshot] = MappingProxyType({}) def _utc_date() -> str: @@ -59,20 +83,54 @@ class GatewayRequestAccumulator: route) however long the database is unreachable. This buys at-least-once, not exactly-once, and the cost is worth stating. - The batch commits inside its context manager's ``__aexit__``, so a failure - raised after the transaction committed (a connection dropped while reading - the acknowledgement) restores counts that are already persisted, and the - next flush increments them a second time. Exactly-once would need a dedup - key the upserts could ignore on replay. For a traffic-volume metric a rare + The statement commits on the server before its acknowledgement is read, so + a failure raised after the commit (a connection dropped while reading the + acknowledgement) restores counts that are already persisted, and the next + flush increments them a second time. Exactly-once would need a dedup key + the upsert could ignore on replay. For a traffic-volume metric a rare overcount on a dropped acknowledgement beats losing a whole interval to every database blip, so the trade is deliberate. """ - for key, counts in snapshot.items(): - existing = self._counts.get(key, _EMPTY) - self._counts[key] = GatewayRequestCounts( - successful_requests=existing.successful_requests + counts.successful_requests, - failed_requests=existing.failed_requests + counts.failed_requests, - ) + self._counts = dict(fold_counts(chain(self._counts.items(), snapshot.items()))) # mutable-ok: fold replaced + + +def fold_counts(items: Iterable[tuple[GatewayRequestKey, GatewayRequestCounts]]) -> GatewayRequestSnapshot: + """Sum counts key-wise; the result stays bounded by (date x category x route).""" + folded: Final[dict[GatewayRequestKey, GatewayRequestCounts]] = {} # mutable-ok: local fold returned once + for key, counts in items: + existing = folded.get(key, _EMPTY) + folded[key] = GatewayRequestCounts( + successful_requests=existing.successful_requests + counts.successful_requests, + failed_requests=existing.failed_requests + counts.failed_requests, + ) + return folded + + +def build_gateway_requests_upsert(snapshot: GatewayRequestSnapshot) -> tuple[str, tuple[str | int, ...]]: + """ + One ``INSERT ... ON CONFLICT DO UPDATE`` that increments every (date, category, + route) in the snapshot. Rows are ordered by the conflict key so concurrent + writers lock rows in the same order and cannot deadlock. + """ + ordered: Final = sorted(snapshot.items(), key=lambda item: (item[0].date, item[0].category, item[0].route)) + rows: Final = ", ".join( + f"(${base + 1}::text, ${base + 2}::text, ${base + 3}::text, ${base + 4}::bigint, ${base + 5}::bigint, {_UTC_NOW})" + for base in range(0, len(ordered) * _COLUMNS_PER_ROW, _COLUMNS_PER_ROW) + ) + sql: Final = ( + f'INSERT INTO {_TABLE} ("date", "category", "route", "successful_requests", "failed_requests", "updated_at")\n' + f"VALUES {rows}\n" + 'ON CONFLICT ("date", "category", "route") DO UPDATE SET\n' + f' "successful_requests" = {_TABLE}."successful_requests" + EXCLUDED."successful_requests",\n' + f' "failed_requests" = {_TABLE}."failed_requests" + EXCLUDED."failed_requests",\n' + f' "updated_at" = {_UTC_NOW}' + ) + params: Final[tuple[str | int, ...]] = tuple( + value + for key, counts in ordered + for value in (key.date, key.category, key.route, counts.successful_requests, counts.failed_requests) + ) + return sql, params async def commit_gateway_requests_to_db( @@ -80,50 +138,130 @@ async def commit_gateway_requests_to_db( prisma_client: "PrismaClient", snapshot: GatewayRequestSnapshot, ) -> None: - """Upsert one incrementing row per (date, category, route).""" + """Increment every (date, category, route) in the snapshot with a single statement.""" if not snapshot: return - ordered: Final = sorted(snapshot.items(), key=lambda item: (item[0].date, item[0].category, item[0].route)) + sql, params = build_gateway_requests_upsert(snapshot) + await prisma_client.db.execute_raw(sql, *params) # pyright: ignore[reportAny] # untyped prisma client - # pyright: ignore[reportAny] on both lines -- prisma's generated client is untyped, - # so .db and every table action off it resolve to Any at this boundary. The dict - # literals below are the shape prisma's generated inputs require. - async with prisma_client.db.batch_() as batcher: # pyright: ignore[reportAny] # untyped prisma client - for key, counts in ordered: - columns = asdict(key) - batcher.litellm_dailygatewayrequests.upsert( # pyright: ignore[reportAny] # untyped prisma client - where={"date_category_route": columns}, # mutable-ok: prisma input is dict-shaped - data={ # mutable-ok: prisma input is dict-shaped - "create": { # mutable-ok: prisma input is dict-shaped - **columns, - "successful_requests": counts.successful_requests, - "failed_requests": counts.failed_requests, - }, - "update": { # mutable-ok: prisma input is dict-shaped - "successful_requests": {"increment": counts.successful_requests}, # mutable-ok: as above - "failed_requests": {"increment": counts.failed_requests}, # mutable-ok: as above - }, - }, + verbose_proxy_logger.debug( + "Gateway request tracking - committed %d aggregated rows in one statement", len(snapshot) + ) + + +class GatewayRequestRedisBuffer: + """ + Folds every worker's snapshot through one Redis list so a single pod per + interval writes the table, mirroring the spend writer's transaction buffer. + + Each entry is one worker's snapshot as JSON rows; the lock holder pops them, + sums them, and commits one statement. A commit failure pushes the summed + rows back so the next holder retries, keeping the at-least-once guarantee. + If that push fails too, the rows go back to the holder's own accumulator so + they ride along with its next flush instead of vanishing with the pop. + """ + + def __init__(self, *, redis_cache: RedisCache, pod_lock_manager: PodLockManager) -> None: + self._redis_cache: Final = redis_cache + self._pod_lock_manager: Final = pod_lock_manager + + async def push(self, snapshot: GatewayRequestSnapshot) -> None: + if not snapshot: + return + rows: Final[_BufferedRows] = tuple( + (key.date, key.category, key.route, counts.successful_requests, counts.failed_requests) + for key, counts in snapshot.items() + ) + await self._redis_cache.async_rpush(key=REDIS_GATEWAY_REQUESTS_BUFFER_KEY, values=(json.dumps(rows),)) + + async def _pop_batch(self) -> tuple[str | bytes, ...]: + popped: Final[object] = await self._redis_cache.async_lpop( # pyright: ignore[reportAny] # redis returns Any + key=REDIS_GATEWAY_REQUESTS_BUFFER_KEY, count=MAX_REDIS_BUFFER_DEQUEUE_COUNT + ) + if not popped: + return () + return _BUFFERED_ENTRIES.validate_python(popped if isinstance(popped, list) else (popped,)) + + async def _pop_all(self) -> AsyncIterator[str | bytes]: + while True: + batch = await self._pop_batch() + for entry in batch: + yield entry + if len(batch) < MAX_REDIS_BUFFER_DEQUEUE_COUNT: + return + + async def pop(self) -> GatewayRequestSnapshot: + entries: Final = tuple([entry async for entry in self._pop_all()]) + return fold_counts( + ( + GatewayRequestKey(date=date, category=category, route=route), + GatewayRequestCounts(successful_requests=succeeded, failed_requests=failed), ) + for entry in entries + for date, category, route, succeeded, failed in _BUFFERED_ROWS.validate_json(entry) + ) - verbose_proxy_logger.debug("Gateway request tracking - committed %d aggregated rows", len(ordered)) + async def commit_if_leader(self, prisma_client: "PrismaClient") -> GatewayRequestSnapshot: + """ + Drain the list and write it as one statement, but only on the pod holding the job lock. + + The lock is a lease, never released: the holder re-enters it on every flush and + keeps committing alone until the TTL lapses, so the primary sees one statement + per flush interval deployment-wide instead of one per worker. + + Returns the popped rows that could be neither committed nor re-queued, for the + caller to keep in memory. Empty on success. + """ + if not await self._pod_lock_manager.acquire_lock(cronjob_id=GATEWAY_REQUESTS_JOB_NAME): + return _NO_COUNTS + buffered: Final = await self.pop() + try: + await commit_gateway_requests_to_db(prisma_client=prisma_client, snapshot=buffered) + except Exception: # noqa: BLE001 -- a failed commit must not stop the scheduler + verbose_proxy_logger.warning( + "Gateway request tracking - failed to commit %d buffered rows, re-queuing to Redis for the next flush", + len(buffered), + exc_info=True, + ) + return await self._requeue(buffered) + return _NO_COUNTS + + async def _requeue(self, snapshot: GatewayRequestSnapshot) -> GatewayRequestSnapshot: + try: + await self.push(snapshot) + except Exception: # noqa: BLE001 -- the rows go back to the caller's accumulator instead + verbose_proxy_logger.warning( + "Gateway request tracking - Redis re-queue failed, keeping %d rows in memory for the next flush", + len(snapshot), + exc_info=True, + ) + return snapshot + return _NO_COUNTS async def flush_gateway_requests( prisma_client: "PrismaClient", accumulator: GatewayRequestAccumulator, + redis_buffer: GatewayRequestRedisBuffer | None = None, ) -> None: """ Scheduler entrypoint. Never raises: a metering failure must not kill the job. + With ``redis_buffer`` the snapshot goes to Redis and only the lease holder + writes to Postgres. Shutdown passes no buffer so a departing worker writes its + own counts directly instead of parking them behind a lease it may not hold. + ``CancelledError`` is deliberately not caught, so a flush cancelled during shutdown drops its snapshot rather than restoring counts onto an accumulator the process is about to discard. """ snapshot: Final = accumulator.drain() try: - await commit_gateway_requests_to_db(prisma_client=prisma_client, snapshot=snapshot) + if redis_buffer is None: + await commit_gateway_requests_to_db(prisma_client=prisma_client, snapshot=snapshot) + else: + await redis_buffer.push(snapshot) except Exception: # noqa: BLE001 -- a failed flush must not stop the scheduler accumulator.restore(snapshot) verbose_proxy_logger.warning( @@ -131,3 +269,13 @@ async def flush_gateway_requests( len(snapshot), exc_info=True, ) + return + if redis_buffer is None: + return + try: + accumulator.restore(await redis_buffer.commit_if_leader(prisma_client)) + except Exception: # noqa: BLE001 -- entries still in Redis are drained by the next flush + verbose_proxy_logger.warning( + "Gateway request tracking - leader drain failed, buffered rows stay in Redis for the next flush", + exc_info=True, + ) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index c241e66049b..09e43eb74e1 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -424,6 +424,7 @@ from litellm.proxy.db.exception_handler import ( ) from litellm.proxy.db.gateway_request_tracking import ( GatewayRequestAccumulator, + GatewayRequestRedisBuffer, flush_gateway_requests, ) from litellm.proxy.db.proxy_worker_heartbeat import ( @@ -2355,6 +2356,17 @@ open_telemetry_logger: OpenTelemetry | None = None gateway_request_accumulator: Final = GatewayRequestAccumulator() ### INITIALIZE GLOBAL LOGGING OBJECT ### proxy_logging_obj: ProxyLogging = ProxyLogging(user_api_key_cache=user_api_key_cache, premium_user=premium_user) + + +def _gateway_request_redis_buffer() -> GatewayRequestRedisBuffer | None: + """Shares the spend writer's transaction-buffer Redis and pod lock when use_redis_transaction_buffer is on.""" + writer: Final = proxy_logging_obj.db_spend_update_writer + redis_cache: Final = writer.redis_update_buffer.redis_cache + if redis_cache is None or not writer.redis_update_buffer._should_commit_spend_updates_to_redis(): + return None + return GatewayRequestRedisBuffer(redis_cache=redis_cache, pod_lock_manager=writer.pod_lock_manager) + + ### REDIS QUEUE ### async_result: Final = None celery_app_conn: Final = None @@ -9633,7 +9645,7 @@ class ProxyStartupEvent: flush_gateway_requests, "interval", seconds=batch_writing_interval, - args=(prisma_client, gateway_request_accumulator), + args=(prisma_client, gateway_request_accumulator, _gateway_request_redis_buffer()), id="update_gateway_requests_job", replace_existing=True, misfire_grace_time=APSCHEDULER_MISFIRE_GRACE_TIME, diff --git a/tests/test_litellm/proxy/db/test_gateway_request_tracking.py b/tests/test_litellm/proxy/db/test_gateway_request_tracking.py index 93a11a914cb..045261e2d53 100644 --- a/tests/test_litellm/proxy/db/test_gateway_request_tracking.py +++ b/tests/test_litellm/proxy/db/test_gateway_request_tracking.py @@ -8,8 +8,11 @@ from datetime import datetime, timezone import pytest +from litellm.constants import MAX_REDIS_BUFFER_DEQUEUE_COUNT, REDIS_GATEWAY_REQUESTS_BUFFER_KEY from litellm.proxy.db.gateway_request_tracking import ( + GATEWAY_REQUESTS_JOB_NAME, GatewayRequestAccumulator, + GatewayRequestRedisBuffer, commit_gateway_requests_to_db, flush_gateway_requests, ) @@ -83,40 +86,47 @@ def test_drain_snapshot_is_not_mutated_by_later_records(): # ── commit ──────────────────────────────────────────────────────────────────── -class FakeTable: - def __init__(self) -> None: - self.upserts: list[dict] = [] - - def upsert(self, *, where: dict, data: dict) -> None: - self.upserts.append({"where": where, "data": data}) - - -class FakeBatcher: - def __init__(self, table: FakeTable) -> None: - self.litellm_dailygatewayrequests = table - - async def __aenter__(self) -> "FakeBatcher": - return self - - async def __aexit__(self, *args: object) -> bool: - return False - - class FakeDB: - def __init__(self, table: FakeTable) -> None: - self._table = table + def __init__(self) -> None: + self.statements: list[tuple[str, tuple[object, ...]]] = [] - def batch_(self) -> FakeBatcher: - return FakeBatcher(self._table) + async def execute_raw(self, query: str, *args: object) -> int: + self.statements.append((query, args)) + return len(args) // 5 class FakePrismaClient: def __init__(self) -> None: - self.table = FakeTable() - self.db = FakeDB(self.table) + self.db = FakeDB() -def test_commit_upserts_one_incrementing_row_per_key(): +def _rows_written(client: FakePrismaClient) -> list[tuple[object, ...]]: + """Every (date, category, route, successful, failed) tuple the database received, in statement order.""" + return [params[i : i + 5] for _, params in client.db.statements for i in range(0, len(params), 5)] + + +def test_commit_increments_with_a_single_statement_for_the_whole_snapshot(): + """One statement per flush is the whole point: the previous per-key upsert cost + the primary (workers x routes) statements per interval.""" + client = FakePrismaClient() + snapshot = { + GatewayRequestKey(date="2026-08-01", category="llm", route=route): ( + GatewayRequestCounts(successful_requests=7, failed_requests=2) + ) + for route in ("/chat/completions", "/embeddings", "/responses", "/v1/messages", "/mcp") + } + + asyncio.run(commit_gateway_requests_to_db(prisma_client=client, snapshot=snapshot)) + + assert len(client.db.statements) == 1 + sql, params = client.db.statements[0] + assert sql.count("ON CONFLICT") == 1 + assert sql.count("(NOW() AT TIME ZONE 'UTC'))") == 5 + assert len(params) == 25 + + +def test_commit_sql_adds_to_the_existing_row_instead_of_replacing_it(): + """A worker only knows its own share; the SQL must add EXCLUDED onto the stored count.""" client = FakePrismaClient() snapshot = { GatewayRequestKey(date="2026-08-01", category="llm", route="/chat/completions"): ( @@ -126,20 +136,36 @@ def test_commit_upserts_one_incrementing_row_per_key(): asyncio.run(commit_gateway_requests_to_db(prisma_client=client, snapshot=snapshot)) - assert len(client.table.upserts) == 1 - written = client.table.upserts[0] - assert written["where"] == { - "date_category_route": { - "date": "2026-08-01", - "category": "llm", - "route": "/chat/completions", - } + sql, params = client.db.statements[0] + assert 'INSERT INTO "LiteLLM_DailyGatewayRequests"' in sql + assert 'ON CONFLICT ("date", "category", "route") DO UPDATE SET' in sql + assert ( + '"successful_requests" = "LiteLLM_DailyGatewayRequests"."successful_requests" + EXCLUDED."successful_requests"' + in sql + ) + assert '"failed_requests" = "LiteLLM_DailyGatewayRequests"."failed_requests" + EXCLUDED."failed_requests"' in sql + assert params == ("2026-08-01", "llm", "/chat/completions", 7, 2) + + +def test_commit_placeholders_line_up_with_params(): + """$n positions are generated per row; a drift here silently swaps a route for a count.""" + client = FakePrismaClient() + snapshot = { + GatewayRequestKey(date="2026-08-01", category="llm", route="/chat/completions"): ( + GatewayRequestCounts(successful_requests=1, failed_requests=0) + ), + GatewayRequestKey(date="2026-08-01", category="mcp", route="/mcp"): ( + GatewayRequestCounts(successful_requests=0, failed_requests=3) + ), } - assert written["data"]["update"] == { - "successful_requests": {"increment": 7}, - "failed_requests": {"increment": 2}, - } - assert written["data"]["create"]["successful_requests"] == 7 + + asyncio.run(commit_gateway_requests_to_db(prisma_client=client, snapshot=snapshot)) + + sql, params = client.db.statements[0] + assert "($1::text, $2::text, $3::text, $4::bigint, $5::bigint," in sql + assert "($6::text, $7::text, $8::text, $9::bigint, $10::bigint," in sql + assert "$11" not in sql + assert params == ("2026-08-01", "llm", "/chat/completions", 1, 0, "2026-08-01", "mcp", "/mcp", 0, 3) def test_commit_is_deterministically_ordered(): @@ -154,17 +180,14 @@ def test_commit_is_deterministically_ordered(): asyncio.run(commit_gateway_requests_to_db(prisma_client=client, snapshot=snapshot)) - written_order = [ - (row["where"]["date_category_route"]["date"], row["where"]["date_category_route"]["category"]) - for row in client.table.upserts - ] + written_order = [(row[0], row[1]) for row in _rows_written(client)] assert written_order == [("2026-08-01", "llm"), ("2026-08-01", "mcp"), ("2026-08-02", "llm")] def test_commit_skips_the_database_entirely_when_nothing_accumulated(): client = FakePrismaClient() asyncio.run(commit_gateway_requests_to_db(prisma_client=client, snapshot={})) - assert client.table.upserts == [] + assert client.db.statements == [] # ── flush ───────────────────────────────────────────────────────────────────── @@ -177,12 +200,12 @@ def test_flush_drains_and_commits(): asyncio.run(flush_gateway_requests(client, acc)) - assert len(client.table.upserts) == 1 + assert len(client.db.statements) == 1 assert acc.drain() == {} class ExplodingDB: - def batch_(self): + async def execute_raw(self, query: str, *args: object) -> int: raise RuntimeError("db gone") @@ -208,10 +231,7 @@ def test_failed_flush_keeps_counts_for_the_next_attempt(): client = FakePrismaClient() asyncio.run(flush_gateway_requests(client, acc)) - assert client.table.upserts[0]["data"]["update"] == { - "successful_requests": {"increment": 1}, - "failed_requests": {"increment": 1}, - } + assert _rows_written(client) == [(_today(), "llm", "/chat/completions", 1, 1)] def test_restored_counts_merge_with_requests_recorded_meanwhile(): @@ -223,5 +243,272 @@ def test_restored_counts_merge_with_requests_recorded_meanwhile(): client = FakePrismaClient() asyncio.run(flush_gateway_requests(client, acc)) - assert len(client.table.upserts) == 1 - assert client.table.upserts[0]["data"]["update"]["successful_requests"] == {"increment": 2} + assert _rows_written(client) == [(_today(), "llm", "/chat/completions", 2, 0)] + + +class ExplodingDBWithInFlightRequest: + """Fails the write after a request has been recorded while it was in flight.""" + + def __init__(self, accumulator: GatewayRequestAccumulator) -> None: + self.accumulator = accumulator + + async def execute_raw(self, query: str, *args: object) -> int: + _record(self.accumulator, 500) + raise RuntimeError("db gone") + + +class ExplodingClientWithInFlightRequest: + def __init__(self, accumulator: GatewayRequestAccumulator) -> None: + self.db = ExplodingDBWithInFlightRequest(accumulator) + + +def test_restore_keeps_requests_recorded_while_the_failed_write_was_in_flight(): + acc = GatewayRequestAccumulator() + _record(acc, 200) + asyncio.run(flush_gateway_requests(ExplodingClientWithInFlightRequest(acc), acc)) + + client = FakePrismaClient() + asyncio.run(flush_gateway_requests(client, acc)) + + assert _rows_written(client) == [(_today(), "llm", "/chat/completions", 1, 1)] + + +# ── redis buffer ────────────────────────────────────────────────────────────── + + +class FakeRedis: + def __init__(self) -> None: + self.lists: dict[str, list[str]] = {} + + async def async_rpush(self, key: str, values: list[str]) -> int: + self.lists.setdefault(key, []).extend(values) + return len(self.lists[key]) + + async def async_lpop(self, key: str, count: int) -> list[str] | None: + queue = self.lists.get(key, []) + if not queue: + return None + popped, self.lists[key] = queue[:count], queue[count:] + return popped + + +class FakePodLock: + def __init__(self, *, leader: bool) -> None: + self.leader = leader + self.held: list[str] = [] + self.released: list[str] = [] + + async def acquire_lock(self, cronjob_id: str) -> bool: + self.held.append(cronjob_id) + return self.leader + + async def release_lock(self, cronjob_id: str) -> None: + self.released.append(cronjob_id) + + +class FakeLease: + """Redis-side view of the job lock: SET NX by pod id, re-entrant for the holder, freed only by release or TTL.""" + + def __init__(self) -> None: + self.holder: str | None = None + + +class FakeLeasePodLock: + def __init__(self, lease: FakeLease, pod_id: str) -> None: + self.lease = lease + self.pod_id = pod_id + + async def acquire_lock(self, cronjob_id: str) -> bool: + if self.lease.holder is None: + self.lease.holder = self.pod_id + return self.lease.holder == self.pod_id + + async def release_lock(self, cronjob_id: str) -> None: + if self.lease.holder == self.pod_id: + self.lease.holder = None + + +def _buffer(redis: FakeRedis, *, leader: bool) -> tuple[GatewayRequestRedisBuffer, FakePodLock]: + lock = FakePodLock(leader=leader) + return GatewayRequestRedisBuffer(redis_cache=redis, pod_lock_manager=lock), lock # pyright: ignore[reportArgumentType] # duck-typed fakes + + +def test_non_leader_workers_push_to_redis_and_never_touch_the_database(): + redis = FakeRedis() + client = FakePrismaClient() + for _ in range(3): + acc = GatewayRequestAccumulator() + _record(acc, 200) + buffer, _ = _buffer(redis, leader=False) + asyncio.run(flush_gateway_requests(client, acc, buffer)) + + assert client.db.statements == [] + assert len(redis.lists[REDIS_GATEWAY_REQUESTS_BUFFER_KEY]) == 3 + + +def test_leader_folds_every_workers_snapshot_into_one_statement(): + """Fifty workers each flushing the same routes must cost the primary one statement, not fifty.""" + redis = FakeRedis() + client = FakePrismaClient() + for _ in range(50): + acc = GatewayRequestAccumulator() + _record(acc, 200) + _record(acc, 500, route="/responses") + buffer, _ = _buffer(redis, leader=False) + asyncio.run(flush_gateway_requests(client, acc, buffer)) + + leader_acc = GatewayRequestAccumulator() + _record(leader_acc, 200) + leader, lock = _buffer(redis, leader=True) + asyncio.run(flush_gateway_requests(client, leader_acc, leader)) + + assert len(client.db.statements) == 1 + assert _rows_written(client) == [ + (_today(), "llm", "/chat/completions", 51, 0), + (_today(), "llm", "/responses", 0, 50), + ] + assert redis.lists[REDIS_GATEWAY_REQUESTS_BUFFER_KEY] == [] + assert lock.held == [GATEWAY_REQUESTS_JOB_NAME] + assert lock.released == [] + + +def test_leader_keeps_the_lease_so_staggered_pods_cost_one_statement_per_interval(): + """Pods flush on their own clocks; without the lease each one would win the lock in turn and commit alone.""" + redis = FakeRedis() + client = FakePrismaClient() + lease = FakeLease() + pods = tuple( + GatewayRequestRedisBuffer(redis_cache=redis, pod_lock_manager=FakeLeasePodLock(lease, f"pod-{i}")) # pyright: ignore[reportArgumentType] # duck-typed fakes + for i in range(4) + ) + + for _interval in range(3): + for pod in pods: + acc = GatewayRequestAccumulator() + _record(acc, 200) + asyncio.run(flush_gateway_requests(client, acc, pod)) + + assert lease.holder == "pod-0" + assert len(client.db.statements) == 3 + assert [row[3] for row in _rows_written(client)] == [1, 4, 4] + assert len(redis.lists[REDIS_GATEWAY_REQUESTS_BUFFER_KEY]) == 3 + + +def test_leader_drains_a_backlog_deeper_than_one_capped_pop(): + """More workers than MAX_REDIS_BUFFER_DEQUEUE_COUNT must not leave a growing tail queued behind the cap.""" + redis = FakeRedis() + client = FakePrismaClient() + workers = MAX_REDIS_BUFFER_DEQUEUE_COUNT * 2 + 1 + for _ in range(workers): + acc = GatewayRequestAccumulator() + _record(acc, 200) + buffer, _ = _buffer(redis, leader=False) + asyncio.run(flush_gateway_requests(client, acc, buffer)) + + leader, _ = _buffer(redis, leader=True) + asyncio.run(flush_gateway_requests(client, GatewayRequestAccumulator(), leader)) + + assert len(client.db.statements) == 1 + assert _rows_written(client) == [(_today(), "llm", "/chat/completions", workers, 0)] + assert redis.lists[REDIS_GATEWAY_REQUESTS_BUFFER_KEY] == [] + + +def test_leader_with_nothing_buffered_writes_nothing(): + redis = FakeRedis() + client = FakePrismaClient() + leader, lock = _buffer(redis, leader=True) + + asyncio.run(flush_gateway_requests(client, GatewayRequestAccumulator(), leader)) + + assert client.db.statements == [] + assert lock.released == [] + + +def test_leader_requeues_to_redis_when_the_database_commit_fails(): + """Counts popped from Redis are gone from every worker; a failed commit must put them back.""" + redis = FakeRedis() + acc = GatewayRequestAccumulator() + _record(acc, 200) + _record(acc, 200) + leader, lock = _buffer(redis, leader=True) + + asyncio.run(flush_gateway_requests(ExplodingClient(), acc, leader)) + + assert len(redis.lists[REDIS_GATEWAY_REQUESTS_BUFFER_KEY]) == 1 + assert lock.released == [] + assert acc.drain() == {} + + client = FakePrismaClient() + retry, _ = _buffer(redis, leader=True) + asyncio.run(flush_gateway_requests(client, GatewayRequestAccumulator(), retry)) + assert _rows_written(client) == [(_today(), "llm", "/chat/completions", 2, 0)] + + +class ExplodingRedis(FakeRedis): + async def async_rpush(self, key: str, values: list[str]) -> int: + raise RuntimeError("redis gone") + + +class UnreadableRedis(FakeRedis): + async def async_lpop(self, key: str, count: int) -> list[str] | None: + raise RuntimeError("redis gone mid-flush") + + +class UnwritableRedis(FakeRedis): + """Pops succeed, pushes fail: a Redis that went read-only between the leader's pop and its re-queue.""" + + async def async_rpush(self, key: str, values: list[str]) -> int: + raise RuntimeError("redis read-only") + + +def test_leader_keeps_popped_counts_in_memory_when_both_the_database_and_the_requeue_fail(): + """The pop removed the only copy; if Redis will not take it back the leader itself must carry it.""" + redis = FakeRedis() + worker_acc = GatewayRequestAccumulator() + _record(worker_acc, 200) + _record(worker_acc, 200) + worker, _ = _buffer(redis, leader=False) + asyncio.run(flush_gateway_requests(FakePrismaClient(), worker_acc, worker)) + + degraded = UnwritableRedis() + degraded.lists = redis.lists + leader_acc = GatewayRequestAccumulator() + leader, _ = _buffer(degraded, leader=True) + asyncio.run(flush_gateway_requests(ExplodingClient(), leader_acc, leader)) + assert degraded.lists[REDIS_GATEWAY_REQUESTS_BUFFER_KEY] == [] + + client = FakePrismaClient() + retry, _ = _buffer(redis, leader=True) + asyncio.run(flush_gateway_requests(client, leader_acc, retry)) + assert _rows_written(client) == [(_today(), "llm", "/chat/completions", 2, 0)] + + +def test_leader_whose_redis_read_fails_leaves_the_pushed_rows_for_the_next_flush(): + """The scheduler job must not raise, and nothing is popped so nothing needs restoring anywhere.""" + redis = UnreadableRedis() + acc = GatewayRequestAccumulator() + _record(acc, 200) + client = FakePrismaClient() + leader, _ = _buffer(redis, leader=True) + + asyncio.run(flush_gateway_requests(client, acc, leader)) + + assert client.db.statements == [] + assert acc.drain() == {} + assert len(redis.lists[REDIS_GATEWAY_REQUESTS_BUFFER_KEY]) == 1 + + +def test_failed_redis_push_keeps_counts_locally_for_the_next_flush(): + acc = GatewayRequestAccumulator() + _record(acc, 200) + _record(acc, 500) + buffer, lock = _buffer(ExplodingRedis(), leader=True) + + asyncio.run(flush_gateway_requests(FakePrismaClient(), acc, buffer)) + + assert lock.held == [] + assert acc.drain() == { + GatewayRequestKey(date=_today(), category="llm", route="/chat/completions"): ( + GatewayRequestCounts(successful_requests=1, failed_requests=1) + ) + } From 6195fbf5f3aaa656d916f0a7c3013ada34e87752 Mon Sep 17 00:00:00 2001 From: Animesh Kumar Date: Wed, 9 Sep 2026 23:24:12 +0530 Subject: [PATCH 24/26] test: isolate bedrock aws tests from ambient SSL env vars Nine cases assert the sts client is built with verify=True, but get_ssl_verify reads SSL_CERT_FILE and SSL_VERIFY, so the argument depended on the ambient environment. The published images set SSL_CERT_FILE, so the suite failed there while passing in CI. Fixes #40357 --- tests/test_litellm/llms/bedrock/test_base_aws_llm.py | 8 ++++++++ 1 file changed, 8 insertions(+) diff --git a/tests/test_litellm/llms/bedrock/test_base_aws_llm.py b/tests/test_litellm/llms/bedrock/test_base_aws_llm.py index f854d806bdc..a1c28f36e70 100644 --- a/tests/test_litellm/llms/bedrock/test_base_aws_llm.py +++ b/tests/test_litellm/llms/bedrock/test_base_aws_llm.py @@ -38,6 +38,14 @@ def flush_shared_bedrock_iam_cache(): yield +@pytest.fixture(autouse=True) +def _clean_ssl_env(monkeypatch): + """get_ssl_verify reads these, so the sts client's verify= would otherwise depend on + the ambient environment. The published images set SSL_CERT_FILE.""" + for env_var in ("SSL_CERT_FILE", "SSL_VERIFY"): + monkeypatch.delenv(env_var, raising=False) + + def test_base_aws_llm_instances_share_process_wide_iam_cache(): """Regression LIT-2662: new instances must reuse iam_cache (Bedrock passthrough is per-request).""" first = BaseAWSLLM() From c82c9cbcedf43661521270740273d6437d76c098 Mon Sep 17 00:00:00 2001 From: tin-berri Date: Wed, 9 Sep 2026 11:04:11 -0700 Subject: [PATCH 25/26] fix(router): strip encrypted reasoning on an auto-router tier change instead of a 503 (#40280) A Responses API follow-up that replays reasoning.encrypted_content is pinned to the deployment that minted it. Behind an auto-router the pre-routing hook rebinds the model to the tier it picked before the candidate pool is built, so a turn that classifies into a different tier never finds the origin and the affinity check raised its fail-fast 503, whose text claims a cooldown that does not exist When the deployment that minted the reasoning is not a member of the model group this turn is routed to, strip the encrypted reasoning (keeping any readable summary, string or block form) and dispatch to the routed group. Membership is tested by deployment id against the candidate set the router itself resolved for the route (routing group, model_name, team, and pattern alike), not by model-group name, so an alias, a provider-qualified spelling, a team-public name, or a pattern route of the same group is not misread as a tier change. An unknown origin (a removed deployment, or a forged/unauthenticated marker) is handled the same as a cross-group one and its reasoning is stripped, so a real cross-group id and a nonexistent id return the same response and cannot be used to enumerate deployment ids. Unavailability within the origin's own group keeps the existing 429/503 fail-fast, so the cooldown contract is unchanged Resolves LIT-7195 Claude-Session: https://claude.ai/code/session_01KAumQbhzk6jdWWHFLA8Jar Co-authored-by: Claude Opus 4.8 --- litellm/responses/utils.py | 43 ++ litellm/router.py | 34 ++ .../encrypted_content_affinity_check.py | 73 ++- .../test_encrypted_content_affinity_check.py | 561 ++++++++++++------ tests/test_litellm/test_router.py | 42 ++ 5 files changed, 557 insertions(+), 196 deletions(-) diff --git a/litellm/responses/utils.py b/litellm/responses/utils.py index 540d492beec..599e978df6a 100644 --- a/litellm/responses/utils.py +++ b/litellm/responses/utils.py @@ -543,6 +543,49 @@ class ResponsesAPIRequestUtils: return request_input + @staticmethod + def strip_encrypted_reasoning_from_input(request_input: object) -> None: + """Drop reasoning items the routed deployment cannot decrypt, keeping their readable summary. + + Mutates ``request_input`` in place: the router's fallback snapshot shares this + list object, so a rebound list would replay the stripped items on the fallback hop. + """ + if not isinstance(request_input, list): + return + items: Final = cast(list[object], request_input) # cast-ok: untyped client json + stripped: Final = tuple(ResponsesAPIRequestUtils._without_encrypted_reasoning(item) for item in items) + items[:] = (item for item in stripped if item is not None) # rebind-ok: list shared with fallback snapshot + + @staticmethod + def _without_encrypted_reasoning(item: object) -> object | None: + if not isinstance(item, dict): + return item + reasoning: Final = cast(Mapping[str, object], item) # cast-ok: untyped client json + if reasoning.get("type") != "reasoning" or not reasoning.get("encrypted_content"): + return reasoning + readable: Final = any( + ResponsesAPIRequestUtils._has_readable_text(reasoning.get(key)) for key in ("summary", "content") + ) + if not readable: + return None + kept: Final[dict[str, object]] = { # mutable-ok: request item rebuilt without the undecryptable keys + key: value for key, value in reasoning.items() if key not in ("encrypted_content", "id") + } + return kept + + @staticmethod + def _has_readable_text(value: object) -> bool: + """A reasoning item's ``summary``/``content`` carries readable text: a non-empty string, or a + list holding at least one block with a non-empty ``text`` field (summary_text / output_text).""" + if isinstance(value, str): + return bool(value.strip()) + if isinstance(value, list): + return any( + isinstance(block, dict) and bool(cast(Mapping[str, object], block).get("text")) # cast-ok: untyped json + for block in value + ) + return False + @staticmethod def _build_responses_api_response_id( custom_llm_provider: str | None, diff --git a/litellm/router.py b/litellm/router.py index 2fbdf541487..68ace283949 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -11167,6 +11167,40 @@ class Router: return ids + def get_candidate_model_ids_for_route(self, model: str, team_id: str | None = None) -> frozenset[str]: + """ + Deployment ids that could serve ``model`` for ``team_id``, unioned across the paths + the router resolves a route through: ``model_group_alias``, a routing group, the + ``model_name`` and team indexes, and wildcard pattern routes. Read-only and + side-effect-free, unlike ``_common_checks_available_deployment`` which also applies + fallbacks and can raise. Lets a pre-call check tell a genuine cross-group route from + same-group unavailability without re-deriving that precedence at the call site, and + without leaking deployment ids into request kwargs bound for the provider. + """ + resolved: Final = self._get_model_from_alias(model=model) or model + routing_group_members: Final = self._get_routing_group_deployments(model=resolved, team_id=team_id) + if routing_group_members is not None: + return self._deployment_ids(routing_group_members) + if resolved in self.model_names: + return self._deployment_ids(self._get_all_deployments(model_name=resolved, team_id=team_id)) + team_router: Final = self.team_pattern_routers.get(team_id) if team_id is not None else None + return self._deployment_ids( + ( + *self._get_all_deployments(model_name=resolved, team_id=team_id), + *(self.pattern_router.route(resolved) or ()), + *((team_router.route(resolved) or ()) if team_router is not None else ()), + ) + ) + + @staticmethod + def _deployment_ids(deployments: Sequence[Mapping[str, object]]) -> frozenset[str]: + return frozenset( + str(model_info["id"]) + for deployment in deployments + for model_info in (deployment.get("model_info"),) + if isinstance(model_info, Mapping) and model_info.get("id") is not None + ) + def has_model_id(self, candidate_id: str) -> bool: """ O(1) membership check for a deployment ID without allocating large lists. diff --git a/litellm/router_utils/pre_call_checks/encrypted_content_affinity_check.py b/litellm/router_utils/pre_call_checks/encrypted_content_affinity_check.py index b623e31ce06..d10153881d9 100644 --- a/litellm/router_utils/pre_call_checks/encrypted_content_affinity_check.py +++ b/litellm/router_utils/pre_call_checks/encrypted_content_affinity_check.py @@ -37,13 +37,13 @@ Safe to enable globally: """ import time +from collections.abc import Mapping from typing import TYPE_CHECKING, Final, Optional, Protocol, cast import httpx from litellm._logging import verbose_router_logger from litellm.exceptions import ( - BadRequestError, RateLimitError, ServiceUnavailableError, ) @@ -158,6 +158,23 @@ class EncryptedContentAffinityCheck(CustomLogger): return deployment return None + @staticmethod + def _request_team_id(request_kwargs: Mapping[str, object]) -> str | None: + containers: Final = (request_kwargs.get("metadata"), request_kwargs.get("litellm_metadata")) + team_ids: Final = (c.get("user_api_key_team_id") for c in containers if isinstance(c, Mapping)) + return next((tid for tid in team_ids if isinstance(tid, str)), None) + + def _routed_group_candidate_model_ids(self, request_kwargs: Mapping[str, object], model: str) -> frozenset[str]: + """ + Deployment ids that could serve this turn's routed ``model``, as the router + resolves a route (model_group_alias / routing group / model_name / team / + pattern). Delegates to the router so the full precedence is not re-derived here + and no deployment ids are written into request kwargs bound for the provider. + """ + if self.router is None: + return frozenset() + return self.router.get_candidate_model_ids_for_route(model=model, team_id=self._request_team_id(request_kwargs)) + @staticmethod def _encryption_boundary_key( litellm_params: object, @@ -225,10 +242,14 @@ class EncryptedContentAffinityCheck(CustomLogger): """ If the request ``input`` contains litellm-encoded item IDs, decode the embedded ``model_id`` and pin the request to that deployment. Raises - ``RateLimitError`` / ``ServiceUnavailableError`` / ``BadRequestError`` - when the originating deployment is unavailable and no encryption-boundary - peer exists, rather than dispatching a doomed request to a non-peer - deployment. The 429/503 split mirrors the originating cooldown's status: + ``RateLimitError`` / ``ServiceUnavailableError`` when the originating + deployment is a member of the routed model group but currently unavailable + and no encryption-boundary peer exists, rather than dispatching a doomed + request to a non-peer deployment. When the origin is not a member of the + routed group (an auto-router tier change, a model switch with no peer, a + removed deployment, or an unknown/forged marker), the encrypted reasoning is + stripped and the request dispatches with its readable history instead. The + 429/503 split mirrors the originating cooldown's status: a 429-induced cooldown surfaces as 429 (with ``Retry-After`` set to the remaining cooldown window) so OpenAI-compatible clients back off and retry after the deployment is eligible again. @@ -285,12 +306,34 @@ class EncryptedContentAffinityCheck(CustomLogger): request_kwargs["_encrypted_content_affinity_pinned"] = True return boundary_matches - # Dispatching to a non-peer would guarantee an upstream - # `invalid_encrypted_content` 400, so fail fast with a clearer error. + # The origin cannot serve this turn's routed group and no peer shares the boundary, so its + # encrypted reasoning can never decrypt here. Strip it, keep the readable history, and dispatch + # to the routed group instead of failing. Membership is tested by deployment id against the set + # the router actually resolved for this route, not by model-group name, so an alias, a + # provider-qualified spelling, a team-public name, or a pattern route of the same group is not + # mistaken for a tier change. An unknown origin (a removed deployment, or a forged marker) is + # treated the same as a cross-group one, which also denies an authenticated caller a + # deployment-id existence oracle: a real cross-group id and a nonexistent id both strip and + # dispatch rather than returning distinguishable responses. Only a genuine same-group member + # that is currently unavailable falls through to the fail-fast, preserving the cooldown contract. + routed_group_model_ids: Final = ( + self._routed_group_candidate_model_ids(request_kwargs, model) if originating is not None else frozenset() + ) + if str(model_id) not in routed_group_model_ids: + verbose_router_logger.debug( + "EncryptedContentAffinityCheck: model_id=%s is not a candidate for the routed group %s; " + "forwarding without its encrypted reasoning", + model_id, + model, + ) + ResponsesAPIRequestUtils.strip_encrypted_reasoning_from_input(request_input) + return typed_healthy_deployments + + # The origin is a member of the routed group but currently unavailable (cooled down); fail fast + # rather than dispatching to a non-peer, which would guarantee an upstream 400. raise await self._unavailable_origin_error( model=model, model_id=model_id, - originating=originating, parent_otel_span=parent_otel_span, ) @@ -298,25 +341,11 @@ class EncryptedContentAffinityCheck(CustomLogger): self, model: str, model_id: str, - originating: Deployment | None, parent_otel_span: Span | None, ) -> Exception: # Public error messages intentionally omit the originating ``model_id`` so # an authenticated caller forging encrypted-content markers cannot use the # error surface to enumerate which deployment IDs exist on this router. - if originating is None: - return BadRequestError( - message=( - "The deployment that produced this encrypted_content is no " - "longer configured on this router, and no deployment on the " - "same encryption boundary is available. Re-issue the request " - "without the stale encrypted_content items, or restore the " - "originating deployment." - ), - model=model, - llm_provider="", - ) - cooldown: Final = await self._get_origin_cooldown(model_id=model_id, parent_otel_span=parent_otel_span) if cooldown is not None and str(cooldown.get("status_code")) == "429": diff --git a/tests/test_litellm/router_utils/pre_call_checks/test_encrypted_content_affinity_check.py b/tests/test_litellm/router_utils/pre_call_checks/test_encrypted_content_affinity_check.py index dac991a41c4..3961c8d74c2 100644 --- a/tests/test_litellm/router_utils/pre_call_checks/test_encrypted_content_affinity_check.py +++ b/tests/test_litellm/router_utils/pre_call_checks/test_encrypted_content_affinity_check.py @@ -16,12 +16,10 @@ The mechanism works without any cache and supports two encoding strategies: """ import time -from typing import List, Optional from unittest.mock import AsyncMock, patch import pytest - import litellm from litellm.responses.utils import ResponsesAPIRequestUtils from litellm.types.llms.openai import ResponsesAPIResponse @@ -68,9 +66,7 @@ class TestEncryptedItemIdCodec: def test_roundtrip(self): model_id = "deployment-1" original_item_id = "rs_abc123def456" - encoded = ResponsesAPIRequestUtils._build_encrypted_item_id( - model_id, original_item_id - ) + encoded = ResponsesAPIRequestUtils._build_encrypted_item_id(model_id, original_item_id) assert encoded.startswith("encitem_") decoded = ResponsesAPIRequestUtils._decode_encrypted_item_id(encoded) assert decoded is not None @@ -81,9 +77,7 @@ class TestEncryptedItemIdCodec: """Decoding must succeed even if base64 padding (=) was stripped in transit.""" model_id = "gpt-5.1-codex-openai-2" original_item_id = "rs_0efb96cb222403210069a01d5d52588196a9dc394ffdb89d00" - encoded = ResponsesAPIRequestUtils._build_encrypted_item_id( - model_id, original_item_id - ) + encoded = ResponsesAPIRequestUtils._build_encrypted_item_id(model_id, original_item_id) # Strip any trailing '=' to simulate what happens in transit stripped = encoded.rstrip("=") decoded = ResponsesAPIRequestUtils._decode_encrypted_item_id(stripped) @@ -100,9 +94,7 @@ class TestEncryptedItemIdCodec: """item_id values containing ';' must survive the roundtrip.""" model_id = "deployment-1" original_item_id = "rs_part1;part2;part3" - encoded = ResponsesAPIRequestUtils._build_encrypted_item_id( - model_id, original_item_id - ) + encoded = ResponsesAPIRequestUtils._build_encrypted_item_id(model_id, original_item_id) decoded = ResponsesAPIRequestUtils._decode_encrypted_item_id(encoded) assert decoded is not None assert decoded["item_id"] == original_item_id @@ -118,11 +110,7 @@ class TestUpdateEncryptedContentItemIds: {"id": "rs_xyz", "type": "reasoning", "encrypted_content": "secret"}, ], } - result = ( - ResponsesAPIRequestUtils._update_encrypted_content_item_ids_in_response( - response, model_id - ) - ) + result = ResponsesAPIRequestUtils._update_encrypted_content_item_ids_in_response(response, model_id) # Plain message item untouched assert result["output"][0]["id"] == "msg_abc" # Reasoning item with encrypted_content gets encoded @@ -133,16 +121,8 @@ class TestUpdateEncryptedContentItemIds: assert decoded["item_id"] == "rs_xyz" def test_no_op_when_model_id_is_none(self): - response = { - "output": [ - {"id": "rs_xyz", "type": "reasoning", "encrypted_content": "secret"} - ] - } - result = ( - ResponsesAPIRequestUtils._update_encrypted_content_item_ids_in_response( - response, None - ) - ) + response = {"output": [{"id": "rs_xyz", "type": "reasoning", "encrypted_content": "secret"}]} + result = ResponsesAPIRequestUtils._update_encrypted_content_item_ids_in_response(response, None) assert result["output"][0]["id"] == "rs_xyz" @@ -151,9 +131,7 @@ class TestEncryptedContentWrapping: """Test wrapping encrypted_content with model_id metadata.""" model_id = "deployment-1" original_content = "gAAAAABpnW_yEYmSNEyOG_original_encrypted_data" - wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id( - original_content, model_id - ) + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id(original_content, model_id) assert wrapped.startswith("litellm_enc:") assert wrapped != original_content @@ -170,9 +148,7 @@ class TestEncryptedContentWrapping: ( model_id, content, - ) = ResponsesAPIRequestUtils._unwrap_encrypted_content_with_model_id( - plain_content - ) + ) = ResponsesAPIRequestUtils._unwrap_encrypted_content_with_model_id(plain_content) assert model_id is None assert content == plain_content @@ -189,11 +165,7 @@ class TestEncryptedContentWrapping: }, ], } - result = ( - ResponsesAPIRequestUtils._update_encrypted_content_item_ids_in_response( - response, model_id - ) - ) + result = ResponsesAPIRequestUtils._update_encrypted_content_item_ids_in_response(response, model_id) assert result["output"][0].get("encrypted_content") is None wrapped = result["output"][1]["encrypted_content"] assert wrapped.startswith("litellm_enc:") @@ -210,19 +182,13 @@ class TestRestoreEncryptedContentItemIds: def test_restores_encoded_ids(self): model_id = "deployment-1" original_id = "rs_encrypted_item_456" - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - model_id, original_id - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id(model_id, original_id) request_input = [ {"type": "message", "id": "msg_abc123", "role": "assistant"}, {"type": "reasoning", "id": encoded_id, "encrypted_content": "secret"}, ] - restored = ( - ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input( - request_input - ) - ) + restored = ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input(request_input) assert restored[0]["id"] == "msg_abc123" assert restored[1]["id"] == original_id @@ -230,33 +196,21 @@ class TestRestoreEncryptedContentItemIds: """Test that wrapped encrypted_content is unwrapped before forwarding.""" model_id = "deployment-1" original_content = "gAAAAABpnW_yEYmSNEyOG_original" - wrapped_content = ( - ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id( - original_content, model_id - ) - ) + wrapped_content = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id(original_content, model_id) request_input = [ {"type": "reasoning", "encrypted_content": wrapped_content}, ] - restored = ( - ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input( - request_input - ) - ) + restored = ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input(request_input) assert restored[0]["encrypted_content"] == original_content def test_no_op_for_plain_string_input(self): - result = ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input( - "Hello world" - ) + result = ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input("Hello world") assert result == "Hello world" def test_no_op_for_unencoded_ids(self): request_input = [{"type": "message", "id": "msg_plain"}] - result = ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input( - request_input - ) + result = ResponsesAPIRequestUtils._restore_encrypted_content_item_ids_in_input(request_input) assert result[0]["id"] == "msg_plain" @@ -283,9 +237,7 @@ async def test_encrypted_content_affinity_tracks_and_routes(): "id": "msg_abc123", "status": "completed", "role": "assistant", - "content": [ - {"type": "output_text", "text": "Hello!", "annotations": []} - ], + "content": [{"type": "output_text", "text": "Hello!", "annotations": []}], }, { "type": "reasoning", @@ -347,9 +299,9 @@ async def test_encrypted_content_affinity_tracks_and_routes(): # The response must have rewritten the encrypted item's ID to encoded form encoded_item_id = _extract_encoded_item_id(first_response) - assert encoded_item_id.startswith( - "encitem_" - ), f"Expected output item ID to be rewritten to encitem_... but got {encoded_item_id!r}" + assert encoded_item_id.startswith("encitem_"), ( + f"Expected output item ID to be rewritten to encitem_... but got {encoded_item_id!r}" + ) # Verify the encoded ID decodes back to the correct deployment + original ID decoded = ResponsesAPIRequestUtils._decode_encrypted_item_id(encoded_item_id) @@ -371,9 +323,9 @@ async def test_encrypted_content_affinity_tracks_and_routes(): ) second_model_id = second_response._hidden_params["model_id"] - assert ( - second_model_id == first_model_id - ), f"Expected affinity to route to {first_model_id}, but got {second_model_id}" + assert second_model_id == first_model_id, ( + f"Expected affinity to route to {first_model_id}, but got {second_model_id}" + ) @pytest.mark.asyncio @@ -478,9 +430,7 @@ async def test_encrypted_content_affinity_bypasses_rpm_limits(): # Extract encoded item ID from the first response output encoded_item_id = _extract_encoded_item_id(first_response) - assert encoded_item_id.startswith( - "encitem_" - ), f"Expected encitem_... but got {encoded_item_id!r}" + assert encoded_item_id.startswith("encitem_"), f"Expected encitem_... but got {encoded_item_id!r}" # Follow-up with the encoded item ID — should pin to same deployment second_response = await router.aresponses( @@ -628,17 +578,13 @@ async def test_encrypted_content_affinity_with_wrapped_content_no_id(): if hasattr(first_item, "encrypted_content") else first_item.get("encrypted_content") ) - assert wrapped_content.startswith( - "litellm_enc:" - ), f"Expected wrapped content but got {wrapped_content[:50]}..." + assert wrapped_content.startswith("litellm_enc:"), f"Expected wrapped content but got {wrapped_content[:50]}..." # Verify we can extract model_id from wrapped content ( extracted_model_id, _, - ) = ResponsesAPIRequestUtils._unwrap_encrypted_content_with_model_id( - wrapped_content - ) + ) = ResponsesAPIRequestUtils._unwrap_encrypted_content_with_model_id(wrapped_content) assert extracted_model_id == first_model_id # Second request: use wrapped encrypted_content WITHOUT an ID (Codex behavior) @@ -653,9 +599,9 @@ async def test_encrypted_content_affinity_with_wrapped_content_no_id(): ) second_model_id = second_response._hidden_params["model_id"] - assert ( - second_model_id == first_model_id - ), f"Expected affinity to route to {first_model_id}, but got {second_model_id}" + assert second_model_id == first_model_id, ( + f"Expected affinity to route to {first_model_id}, but got {second_model_id}" + ) def test_encrypted_content_wrapping_preserves_original_content(): @@ -664,13 +610,9 @@ def test_encrypted_content_wrapping_preserves_original_content(): This is critical for streaming responses where content must round-trip correctly. """ model_id = "test-deployment-1" - original_encrypted_content = ( - "gAAAAABpnW_yEYmSNEyOG_streaming_test_content_with_special_chars==+/" - ) + original_encrypted_content = "gAAAAABpnW_yEYmSNEyOG_streaming_test_content_with_special_chars==+/" - wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id( - original_encrypted_content, model_id - ) + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id(original_encrypted_content, model_id) assert wrapped.startswith("litellm_enc:") assert wrapped != original_encrypted_content @@ -691,9 +633,7 @@ def test_encrypted_content_wrapping_with_multiple_semicolons(): model_id = "deployment-with-semicolons" original_content = "gAAAAAB;some;content;with;semicolons" - wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id( - original_content, model_id - ) + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id(original_content, model_id) ( extracted_model_id, @@ -764,9 +704,7 @@ async def test_encrypted_content_affinity_preserves_litellm_metadata_for_respons request_kwargs=request_kwargs, ) - assert ( - request_kwargs["litellm_metadata"]["encrypted_content_affinity_enabled"] is True - ) + assert request_kwargs["litellm_metadata"]["encrypted_content_affinity_enabled"] is True assert request_kwargs["litellm_metadata"]["model_info"] == {"id": "dep-1"} @@ -777,9 +715,7 @@ def test_encrypted_content_wrapping_empty_string(): model_id = "test-deployment" original_content = "" - wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id( - original_content, model_id - ) + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id(original_content, model_id) assert wrapped.startswith("litellm_enc:") @@ -1132,9 +1068,7 @@ def test_boundary_key_accepts_pydantic_litellm_params_instance(): "api_key": "fake-azure-resource-key-a", } - pydantic_key = EncryptedContentAffinityCheck._encryption_boundary_key( - pydantic_params - ) + pydantic_key = EncryptedContentAffinityCheck._encryption_boundary_key(pydantic_params) plain_key = EncryptedContentAffinityCheck._encryption_boundary_key(plain_params) assert pydantic_key is not None @@ -1161,18 +1095,8 @@ def test_boundary_key_rejects_non_dict_like_inputs(): for bad in (None, [], "not a dict", 42, object()): assert EncryptedContentAffinityCheck._encryption_boundary_key(bad) is None - assert ( - EncryptedContentAffinityCheck._encryption_boundary_key( - {"api_base": "", "api_key": "k"} - ) - is None - ) - assert ( - EncryptedContentAffinityCheck._encryption_boundary_key( - {"api_base": "https://x"} - ) - is None - ) + assert EncryptedContentAffinityCheck._encryption_boundary_key({"api_base": "", "api_key": "k"}) is None + assert EncryptedContentAffinityCheck._encryption_boundary_key({"api_base": "https://x"}) is None # --------------------------------------------------------------------------- @@ -1180,10 +1104,11 @@ def test_boundary_key_rejects_non_dict_like_inputs(): # --------------------------------------------------------------------------- -def _make_originating_mock(api_base: str, api_key: str): +def _make_originating_mock(api_base: str, api_key: str, model_name: str = "gpt-5.4"): from unittest.mock import MagicMock originating = MagicMock() + originating.model_name = model_name originating.litellm_params.model_dump.return_value = { "api_base": api_base, "api_key": api_key, @@ -1192,19 +1117,23 @@ def _make_originating_mock(api_base: str, api_key: str): def _make_router_mock_with_cooldown( - originating, cooldown_entries: Optional[List[tuple]] = None + originating, + cooldown_entries: list[tuple] | None = None, + routed_group_model_ids: list[str] | None = None, ): """ Build a MagicMock router whose ``cooldown_cache.async_get_active_cooldowns`` - returns ``cooldown_entries`` (defaulting to ``[]`` — no active cooldown). + returns ``cooldown_entries`` (defaulting to ``[]`` — no active cooldown), and + whose ``get_candidate_model_ids_for_route`` returns ``routed_group_model_ids`` + (the deployment ids the router resolves for the routed model; defaulting to ``[]`` + — origin absent from the routed group, i.e. a tier change). """ from unittest.mock import AsyncMock, MagicMock mock_router = MagicMock() mock_router.get_deployment.return_value = originating - mock_router.cooldown_cache.async_get_active_cooldowns = AsyncMock( - return_value=list(cooldown_entries or []) - ) + mock_router.cooldown_cache.async_get_active_cooldowns = AsyncMock(return_value=list(cooldown_entries or [])) + mock_router.get_candidate_model_ids_for_route.return_value = frozenset(routed_group_model_ids or []) return mock_router @@ -1235,15 +1164,15 @@ async def test_affinity_raises_service_unavailable_when_origin_cooled_for_non_42 }, ) ], + routed_group_model_ids=["deployment-a-cooled", "deployment-b"], ) check = EncryptedContentAffinityCheck(router=mock_router) - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-a-cooled", "rs_test" - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-a-cooled", "rs_test") healthy_only_b = [ { "model_info": {"id": "deployment-b"}, + "model_name": "gpt-5.4", "litellm_params": { "api_base": "https://account-b.openai.azure.com/", "api_key": "key-b", @@ -1297,15 +1226,15 @@ async def test_affinity_raises_rate_limit_with_retry_after_when_origin_cooled_fo }, ) ], + routed_group_model_ids=["deployment-a-cooled-429", "deployment-b"], ) check = EncryptedContentAffinityCheck(router=mock_router) - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-a-cooled-429", "rs_test" - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-a-cooled-429", "rs_test") healthy_only_b = [ { "model_info": {"id": "deployment-b"}, + "model_name": "gpt-5.4", "litellm_params": { "api_base": "https://account-b.openai.azure.com/", "api_key": "key-b", @@ -1345,15 +1274,16 @@ async def test_affinity_raises_service_unavailable_when_origin_filtered_without_ ) originating = _make_originating_mock("https://account-a.openai.azure.com/", "key-a") - mock_router = _make_router_mock_with_cooldown(originating, cooldown_entries=[]) + mock_router = _make_router_mock_with_cooldown( + originating, cooldown_entries=[], routed_group_model_ids=["deployment-a-filtered", "deployment-b"] + ) check = EncryptedContentAffinityCheck(router=mock_router) - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-a-filtered", "rs_test" - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-a-filtered", "rs_test") healthy_only_b = [ { "model_info": {"id": "deployment-b"}, + "model_name": "gpt-5.4", "litellm_params": { "api_base": "https://account-b.openai.azure.com/", "api_key": "key-b", @@ -1377,15 +1307,18 @@ async def test_affinity_raises_service_unavailable_when_origin_filtered_without_ @pytest.mark.asyncio -async def test_affinity_raises_bad_request_when_origin_removed(): +async def test_affinity_strips_and_dispatches_when_origin_is_unknown_or_removed(): """ - Originating deployment was removed from the router config and no boundary - peer is available. This is permanent (the stale encrypted_content cannot - be honored), so surface a 400 with actionable text. + A removed deployment, or a forged/unknown affinity marker, resolves to no + originating deployment. It is handled like a cross-group origin: the encrypted + reasoning is stripped and the request dispatches with its readable history, + rather than returning a distinguishable error. That uniform handling denies an + authenticated caller a deployment-id existence oracle, an existing cross-group id + and a nonexistent id both strip and proceed, so responses cannot be told apart. + The membership lookup is skipped entirely when the origin is unknown. """ from unittest.mock import MagicMock - from litellm.exceptions import BadRequestError from litellm.router_utils.pre_call_checks.encrypted_content_affinity_check import ( EncryptedContentAffinityCheck, ) @@ -1394,12 +1327,11 @@ async def test_affinity_raises_bad_request_when_origin_removed(): mock_router.get_deployment.return_value = None check = EncryptedContentAffinityCheck(router=mock_router) - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-removed", "rs_test" - ) - healthy_only_b = [ + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id("gAAAAA-blob", "deployment-removed") + routed_pool = [ { "model_info": {"id": "deployment-b"}, + "model_name": "gpt-5.4", "litellm_params": { "api_base": "https://account-b.openai.azure.com/", "api_key": "key-b", @@ -1408,18 +1340,28 @@ async def test_affinity_raises_bad_request_when_origin_removed(): } ] request_kwargs = { - "input": [{"id": encoded_id, "type": "reasoning"}], + "litellm_metadata": {}, + "input": [ + {"role": "user", "content": "why is the sky blue?"}, + { + "type": "reasoning", + "encrypted_content": wrapped, + "summary": [{"type": "summary_text", "text": "scattering"}], + }, + {"role": "user", "content": "and sunsets?"}, + ], } - with pytest.raises(BadRequestError) as excinfo: - await check.async_filter_deployments( - model="gpt-5.4", - healthy_deployments=healthy_only_b, - messages=None, - request_kwargs=request_kwargs, - ) + result = await check.async_filter_deployments( + model="gpt-5.4", + healthy_deployments=routed_pool, + messages=None, + request_kwargs=request_kwargs, + ) - assert "deployment-removed" not in str(excinfo.value) + assert result is routed_pool + assert not any(isinstance(item, dict) and item.get("encrypted_content") for item in request_kwargs["input"]) + mock_router.get_candidate_model_ids_for_route.assert_not_called() @pytest.mark.asyncio @@ -1444,9 +1386,7 @@ async def test_affinity_does_not_raise_when_boundary_peer_available(): mock_router.get_deployment.return_value = originating check = EncryptedContentAffinityCheck(router=mock_router) - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-a", "rs_test" - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-a", "rs_test") peer = { "model_info": {"id": "deployment-a-peer"}, "litellm_params": { @@ -1490,9 +1430,7 @@ async def test_model_group_affinity_config_enables_encrypted_content_affinity(): }, target_deployment, ] - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-b", "rs_test" - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-b", "rs_test") request_kwargs = { "input": [{"type": "reasoning", "id": encoded_id}], "litellm_metadata": {}, @@ -1536,9 +1474,7 @@ async def test_model_group_affinity_config_does_not_disable_global_encrypted_con }, target_deployment, ] - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-b", "rs_test" - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-b", "rs_test") request_kwargs = { "input": [{"type": "reasoning", "id": encoded_id}], "litellm_metadata": {}, @@ -1600,15 +1536,9 @@ async def test_model_group_encrypted_content_affinity_overrides_global_deploymen try: callbacks = router.optional_callbacks or [] - deployment_callback = next( - cb for cb in callbacks if isinstance(cb, DeploymentAffinityCheck) - ) - encrypted_content_callback = next( - cb for cb in callbacks if isinstance(cb, EncryptedContentAffinityCheck) - ) - assert callbacks.index(encrypted_content_callback) < callbacks.index( - deployment_callback - ) + deployment_callback = next(cb for cb in callbacks if isinstance(cb, DeploymentAffinityCheck)) + encrypted_content_callback = next(cb for cb in callbacks if isinstance(cb, EncryptedContentAffinityCheck)) + assert callbacks.index(encrypted_content_callback) < callbacks.index(deployment_callback) assert encrypted_content_callback.enable_global_affinity is False cache_key = DeploymentAffinityCheck.get_affinity_cache_key( @@ -1620,9 +1550,7 @@ async def test_model_group_encrypted_content_affinity_overrides_global_deploymen value={"model_id": "deployment-a"}, ttl=60, ) - encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id( - "deployment-b", "rs_test" - ) + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-b", "rs_test") request_kwargs = { "input": [ { @@ -1643,16 +1571,301 @@ async def test_model_group_encrypted_content_affinity_overrides_global_deploymen ) assert after_deployment_affinity == [deployment_a, deployment_b] - after_encrypted_content_affinity = ( - await encrypted_content_callback.async_filter_deployments( - model=model_group, - healthy_deployments=after_deployment_affinity, - messages=None, - request_kwargs=request_kwargs, - ) + after_encrypted_content_affinity = await encrypted_content_callback.async_filter_deployments( + model=model_group, + healthy_deployments=after_deployment_affinity, + messages=None, + request_kwargs=request_kwargs, ) assert after_encrypted_content_affinity == [deployment_b] assert request_kwargs.get("_encrypted_content_affinity_pinned") is True finally: router.discard() + + +class TestStripEncryptedReasoningFromInput: + def test_keeps_summary_and_drops_encrypted_content_and_id(self): + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id("gAAAAA-blob", "deployment-a") + encoded_id = ResponsesAPIRequestUtils._build_encrypted_item_id("deployment-a", "rs_1") + request_input = [ + {"role": "user", "content": "first turn"}, + { + "type": "reasoning", + "id": encoded_id, + "encrypted_content": wrapped, + "summary": [{"type": "summary_text", "text": "thought about it"}], + }, + {"type": "reasoning", "id": encoded_id, "encrypted_content": wrapped}, + {"type": "reasoning", "encrypted_content": wrapped, "summary": []}, + {"type": "message", "id": "msg_1", "role": "assistant", "content": "hi"}, + {"role": "user", "content": "second turn"}, + ] + ResponsesAPIRequestUtils.strip_encrypted_reasoning_from_input(request_input) + assert request_input == [ + {"role": "user", "content": "first turn"}, + { + "type": "reasoning", + "summary": [{"type": "summary_text", "text": "thought about it"}], + }, + {"type": "message", "id": "msg_1", "role": "assistant", "content": "hi"}, + {"role": "user", "content": "second turn"}, + ] + + def test_keeps_string_form_summary_when_stripping(self): + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id("gAAAAA-blob", "deployment-a") + request_input = [ + {"type": "reasoning", "encrypted_content": wrapped, "summary": "plain string thought"}, + { + "type": "reasoning", + "encrypted_content": wrapped, + "content": [{"type": "output_text", "text": "in content"}], + }, + {"type": "reasoning", "encrypted_content": wrapped, "summary": "", "content": []}, + ] + ResponsesAPIRequestUtils.strip_encrypted_reasoning_from_input(request_input) + assert request_input == [ + {"type": "reasoning", "summary": "plain string thought"}, + {"type": "reasoning", "content": [{"type": "output_text", "text": "in content"}]}, + ] + + def test_leaves_input_untouched_when_no_encrypted_reasoning(self): + request_input = [ + {"role": "user", "content": "first turn"}, + {"type": "reasoning", "summary": [{"type": "summary_text", "text": "no blob here"}]}, + {"role": "user", "content": "second turn"}, + ] + before = [dict(item) for item in request_input] + ResponsesAPIRequestUtils.strip_encrypted_reasoning_from_input(request_input) + assert request_input == before + + +def _cross_group_request_kwargs(): + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id("gAAAAA-blob", "deployment-a") + return { + "litellm_metadata": {}, + "input": [ + {"role": "user", "content": "ZEBRA: why is the sky blue?"}, + { + "type": "reasoning", + "encrypted_content": wrapped, + "summary": [{"type": "summary_text", "text": "scattering"}], + }, + {"type": "message", "role": "assistant", "content": "Rayleigh scattering."}, + {"role": "user", "content": "KIWI: and sunsets?"}, + ], + } + + +@pytest.mark.asyncio +async def test_affinity_strips_encrypted_reasoning_when_routed_to_another_model_group(): + """ + An auto-router tier change (or a model switch with no boundary peer): the + routed pool holds no deployment of the origin's model group. The origin is + healthy, so a 503 would be wrong; the follow-up dispatches to the routed + pool with the origin's encrypted reasoning stripped and its summary kept. + """ + from litellm.router_utils.pre_call_checks.encrypted_content_affinity_check import ( + EncryptedContentAffinityCheck, + ) + + originating = _make_originating_mock(None, "key-a", model_name="gpt-reasoning-tier") + mock_router = _make_router_mock_with_cooldown( + originating, cooldown_entries=[], routed_group_model_ids=["deployment-b"] + ) + check = EncryptedContentAffinityCheck(router=mock_router) + routed_pool = [ + { + "model_info": {"id": "deployment-b"}, + "model_name": "gpt-simple-tier", + "litellm_params": { + "api_base": "https://gateway.example/v1", + "api_key": "key-b", + "model": "openai/gpt-5-nano", + }, + } + ] + request_kwargs = _cross_group_request_kwargs() + original_input = request_kwargs["input"] + + result = await check.async_filter_deployments( + model="gpt-simple-tier", + healthy_deployments=routed_pool, + messages=None, + request_kwargs=request_kwargs, + ) + + assert result is routed_pool + assert "_encrypted_content_affinity_pinned" not in request_kwargs + assert request_kwargs["litellm_metadata"]["encrypted_content_affinity_enabled"] is True + assert request_kwargs["input"] is original_input + assert [item.get("type") or item["role"] for item in original_input] == [ + "user", + "reasoning", + "message", + "user", + ] + assert original_input[1] == { + "type": "reasoning", + "summary": [{"type": "summary_text", "text": "scattering"}], + } + assert not any(isinstance(item, dict) and item.get("encrypted_content") for item in original_input) + + +@pytest.mark.asyncio +async def test_affinity_fails_fast_within_the_origins_own_group(): + """ + Negative class for the tier-change discriminator: the routed group IS the + origin's group (a same-group cooldown, not a tier change), so even with a + healthy non-origin sibling that cannot decrypt the content, the request + still fails fast and the encrypted reasoning is left intact rather than + stripped. Preserves the LIT-3051 cooldown contract. + """ + from litellm.exceptions import ServiceUnavailableError + from litellm.router_utils.pre_call_checks.encrypted_content_affinity_check import ( + EncryptedContentAffinityCheck, + ) + + originating = _make_originating_mock( + "https://account-a.openai.azure.com/", "key-a", model_name="gpt-reasoning-tier" + ) + mock_router = _make_router_mock_with_cooldown( + originating, cooldown_entries=[], routed_group_model_ids=["deployment-a", "deployment-a2"] + ) + check = EncryptedContentAffinityCheck(router=mock_router) + sibling_pool = [ + { + "model_info": {"id": "deployment-a2"}, + "model_name": "gpt-reasoning-tier", + "litellm_params": { + "api_base": "https://account-a2.openai.azure.com/", + "api_key": "key-a2", + "model": "azure/gpt-5.4", + }, + } + ] + request_kwargs = _cross_group_request_kwargs() + + with pytest.raises(ServiceUnavailableError): + await check.async_filter_deployments( + model="gpt-reasoning-tier", + healthy_deployments=sibling_pool, + messages=None, + request_kwargs=request_kwargs, + ) + + assert request_kwargs["input"][1].get("encrypted_content") + + +@pytest.mark.asyncio +async def test_affinity_does_not_strip_when_group_is_spelled_differently_but_same_by_id(): + """ + The discriminator must key on deployment-id membership, not on the model-group + name string. Here the origin's configured group is spelled ``openai/gpt-5.4-mini`` + while the routed group is the canonical ``gpt-5.4-mini``: same group, different + spelling. A name compare (``originating.model_name != model``) would read this as + a tier change and strip the reasoning it did not have to. Because the origin's id + is a member of the routed group, this is a same-group cooldown instead: the request + fails fast and the encrypted reasoning is left intact. + """ + from litellm.exceptions import ServiceUnavailableError + from litellm.router_utils.pre_call_checks.encrypted_content_affinity_check import ( + EncryptedContentAffinityCheck, + ) + + originating = _make_originating_mock(None, "key-a", model_name="openai/gpt-5.4-mini") + mock_router = _make_router_mock_with_cooldown( + originating, cooldown_entries=[], routed_group_model_ids=["deployment-mini-a", "deployment-mini-b"] + ) + check = EncryptedContentAffinityCheck(router=mock_router) + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id("gAAAAA-blob", "deployment-mini-a") + sibling_pool = [ + { + "model_info": {"id": "deployment-mini-b"}, + "model_name": "gpt-5.4-mini", + "litellm_params": { + "api_base": "https://gateway.example/v1", + "api_key": "key-b", + "model": "openai/gpt-5.4-mini", + }, + } + ] + request_kwargs = { + "litellm_metadata": {}, + "input": [ + {"role": "user", "content": "why is the sky blue?"}, + { + "type": "reasoning", + "encrypted_content": wrapped, + "summary": [{"type": "summary_text", "text": "scattering"}], + }, + {"role": "user", "content": "and sunsets?"}, + ], + } + + with pytest.raises(ServiceUnavailableError): + await check.async_filter_deployments( + model="gpt-5.4-mini", + healthy_deployments=sibling_pool, + messages=None, + request_kwargs=request_kwargs, + ) + + assert request_kwargs["input"][1].get("encrypted_content") + + +@pytest.mark.asyncio +async def test_affinity_honors_router_candidate_ids_for_team_and_pattern_routes(): + """ + The exact `model_name` index does not include team-public or pattern routes, so a + same-group cooldown reached only through one of those would be misread as a tier change + and stripped. The check asks the router for the candidate ids it resolves for the route + (`get_candidate_model_ids_for_route`), which covers those paths, rather than the bare + index. Here that set marks the origin as a candidate, so the request fails fast with its + reasoning intact, and the routed group and team are passed through to the router. + """ + from litellm.exceptions import ServiceUnavailableError + from litellm.router_utils.pre_call_checks.encrypted_content_affinity_check import ( + EncryptedContentAffinityCheck, + ) + + originating = _make_originating_mock(None, "key-a", model_name="model_name_teamA_uuid") + mock_router = _make_router_mock_with_cooldown( + originating, cooldown_entries=[], routed_group_model_ids=["deployment-team-a", "deployment-team-b"] + ) + check = EncryptedContentAffinityCheck(router=mock_router) + wrapped = ResponsesAPIRequestUtils._wrap_encrypted_content_with_model_id("gAAAAA-blob", "deployment-team-a") + sibling_pool = [ + { + "model_info": {"id": "deployment-team-b"}, + "model_name": "team-public-model", + "litellm_params": { + "api_base": "https://gateway.example/v1", + "api_key": "key-b", + "model": "openai/gpt-5.4-mini", + }, + } + ] + request_kwargs = { + "litellm_metadata": {"user_api_key_team_id": "teamA"}, + "input": [ + {"role": "user", "content": "why is the sky blue?"}, + { + "type": "reasoning", + "encrypted_content": wrapped, + "summary": [{"type": "summary_text", "text": "scattering"}], + }, + {"role": "user", "content": "and sunsets?"}, + ], + } + + with pytest.raises(ServiceUnavailableError): + await check.async_filter_deployments( + model="team-public-model", + healthy_deployments=sibling_pool, + messages=None, + request_kwargs=request_kwargs, + ) + + assert request_kwargs["input"][1].get("encrypted_content") + mock_router.get_candidate_model_ids_for_route.assert_called_once_with(model="team-public-model", team_id="teamA") diff --git a/tests/test_litellm/test_router.py b/tests/test_litellm/test_router.py index bc79c5f6589..80da922724d 100644 --- a/tests/test_litellm/test_router.py +++ b/tests/test_litellm/test_router.py @@ -14728,3 +14728,45 @@ def test_router_stays_quiet_when_a_deployment_drop_params_is_a_flag(value, caplo ) assert "is not a flag value" not in caplog.text + + +def test_get_candidate_model_ids_for_route_covers_model_name_and_pattern(): + """ + get_candidate_model_ids_for_route resolves a route the way the router does, so a + pre-call check can tell a genuine cross-group route from same-group unavailability. + A concrete model group returns its member ids; a wildcard/pattern deployment is + included for a concrete model it matches, which the bare model_name index misses. + Regression guard for the LIT-7195 tier-change discriminator's team/pattern gaps. + """ + router = Router( + model_list=[ + { + "model_name": "grp", + "litellm_params": {"model": "openai/gpt-4o", "api_key": "sk-a", "api_base": "https://x.invalid"}, + "model_info": {"id": "dep-a"}, + }, + { + "model_name": "grp", + "litellm_params": {"model": "openai/gpt-4o-mini", "api_key": "sk-b", "api_base": "https://x.invalid"}, + "model_info": {"id": "dep-b"}, + }, + { + "model_name": "openai/*", + "litellm_params": {"model": "openai/*", "api_key": "sk-c", "api_base": "https://x.invalid"}, + "model_info": {"id": "dep-wild"}, + }, + ] + ) + + assert router.get_candidate_model_ids_for_route(model="grp") == frozenset({"dep-a", "dep-b"}) + assert "dep-wild" in router.get_candidate_model_ids_for_route(model="openai/gpt-4o-some-new-model") + + +def test_deployment_ids_stringifies_ids_and_skips_entries_without_a_model_info_id(): + deployments = ( + {"model_info": {"id": "a"}}, + {"model_info": {"id": 2}}, + {"model_info": {}}, + {"no_model_info": True}, + ) + assert Router._deployment_ids(deployments) == frozenset({"a", "2"}) From 096984bfc2f146e6e4bb6c259e8daab885a5e5f6 Mon Sep 17 00:00:00 2001 From: "devin-ai-integration[bot]" <158243242+devin-ai-integration[bot]@users.noreply.github.com> Date: Wed, 9 Sep 2026 11:19:29 -0700 Subject: [PATCH 26/26] fix(proxy): pin multi-root CA bundle to the server's root before handing it to Prisma (#40428) Co-authored-by: yassin Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/proxy/db/db_url_settings.py | 107 +++++++++- .../proxy/db/test_db_url_settings.py | 199 ++++++++++++++++-- 2 files changed, 282 insertions(+), 24 deletions(-) diff --git a/litellm/proxy/db/db_url_settings.py b/litellm/proxy/db/db_url_settings.py index 4a0231ad9df..d1e4b3e92b8 100644 --- a/litellm/proxy/db/db_url_settings.py +++ b/litellm/proxy/db/db_url_settings.py @@ -34,12 +34,20 @@ writer's connection params (pool size, timeouts, pgbouncer mode) for the ones the reader URL does not pin itself. """ +import _ssl +import hashlib import os +import socket +import ssl +import struct +import sys +import tempfile import urllib.parse -from collections.abc import Mapping +from collections.abc import Callable, Mapping, Sequence from functools import partial +from pathlib import Path from types import MappingProxyType -from typing import Annotated, Final, cast +from typing import Annotated, Final, Protocol, TypeAlias, cast from pydantic import AliasChoices, BeforeValidator, Field from pydantic_settings import BaseSettings, SettingsConfigDict @@ -126,21 +134,100 @@ def add_missing_query_params(url: str, params: Mapping[str, str | int | float]) LIBPQ_VERIFY_SSLMODES: Final[frozenset[str]] = frozenset({"verify-ca", "verify-full"}) +PEM_CERT_HEADER: Final = b"-----BEGIN CERTIFICATE-----" +PG_SSL_REQUEST: Final = struct.pack("!ii", 8, 80877103) +TLS_PROBE_TIMEOUT_SECONDS: Final = 10.0 + +RootCertResolver: TypeAlias = Callable[[str, str, int], str] # mutable-ok: Callable parameter syntax -def translate_libpq_ssl_params(url: str) -> str: +class _VerifiedChainSource(Protocol): + def get_verified_chain(self) -> Sequence[_ssl.Certificate] | None: ... + + +def _verified_chain_der(tls: ssl.SSLSocket) -> tuple[bytes, ...]: + if sys.version_info >= (3, 13): + return tuple(tls.get_verified_chain()) + legacy: Final = cast( # cast-ok: the stub omits _sslobj, the C object has get_verified_chain since 3.10 + "_VerifiedChainSource | None", + tls._sslobj, # pyright: ignore[reportAttributeAccessIssue, reportUnknownMemberType] # public API only from 3.13 + ) + chain: Final = () if legacy is None else legacy.get_verified_chain() or () + return tuple(cert.public_bytes(_ssl.ENCODING_DER) for cert in chain) + + +def _server_trust_anchor(cafile: str, host: str, port: int) -> bytes | None: + try: + context: Final = ssl.create_default_context(cafile=cafile) + with socket.create_connection((host, port), timeout=TLS_PROBE_TIMEOUT_SECONDS) as raw: + raw.sendall(PG_SSL_REQUEST) + if raw.recv(1) != b"S": + return None + with context.wrap_socket(raw, server_hostname=host) as tls: + chain: Final = _verified_chain_der(tls) + except (OSError, ValueError): + return None + return chain[-1] if chain else None + + +def pin_bundle_root(cert_path: str, host: str, port: int) -> str: + """Reduce a multi-root CA bundle to the one root that verifies ``host``. + + Prisma's ``sslcert`` loads a single PEM certificate (native-tls + ``Certificate::from_pem``), so pointing it at a bundle such as the AWS RDS + global bundle trusts only the first of its 108 regional roots and the + handshake fails with "unable to get local issuer certificate" for every + other region. A single-certificate file is returned as is. For a bundle, + one verifying handshake (chain and hostname, whole bundle as trust store) + identifies the trust anchor the server actually chains to, which is + written to a single-certificate file for Prisma. If the probe fails the + bundle path is returned unchanged, so Prisma fails closed exactly as + before rather than trusting anything the bundle would not. + """ + try: + if Path(cert_path).read_bytes().count(PEM_CERT_HEADER) < 2: + return cert_path + except OSError: + return cert_path + root: Final = _server_trust_anchor(cert_path, host, port) + if root is None: + return cert_path + pinned: Final = Path(tempfile.gettempdir()) / f"litellm-sslcert-{hashlib.sha256(root).hexdigest()[:16]}.pem" + return str(pinned) if _replace_file(pinned, ssl.DER_cert_to_PEM_cert(root)) else cert_path + + +def _replace_file(target: Path, content: str) -> bool: + """Write ``content`` to a private temp file and rename it over ``target``, so + readers never see a partial file and a symlink planted at ``target`` is + replaced rather than followed.""" + try: + fd, staged = tempfile.mkstemp(dir=target.parent, prefix=f"{target.name}.") + except OSError: + return False + try: + with os.fdopen(fd, "w") as handle: + handle.write(content) + os.replace(staged, target) + except OSError: + Path(staged).unlink(missing_ok=True) + return False + return True + + +def translate_libpq_ssl_params(url: str, resolve_root_cert: RootCertResolver = pin_bundle_root) -> str: """Rewrite libpq's certificate-verification params into Prisma's dialect. Prisma's engine only knows ``sslmode=disable|prefer|require``, ``sslcert`` - (the CA bundle) and ``sslaccept=strict``. It silently discards + (a single CA certificate) and ``sslaccept=strict``. It silently discards ``sslrootcert`` and downgrades ``sslmode=verify-ca`` / ``verify-full`` to ``prefer``, so a URL copied from libpq / RDS docs connects over TLS with no certificate check at all. ``verify-ca`` and ``verify-full`` both become ``require`` (Prisma has no CA-only mode), ``sslrootcert`` becomes - ``sslcert``, and either one turns on ``sslaccept=strict`` (chain and - hostname), matching libpq where a root cert makes ``require`` verify. - Prisma params the operator pinned themselves win; anything else is left - untouched. + ``sslcert`` (run through ``resolve_root_cert``, which pins a multi-root + bundle down to the server's root), and either one turns on + ``sslaccept=strict`` (chain and hostname), matching libpq where a root + cert makes ``require`` verify. Prisma params the operator pinned + themselves win; anything else is left untouched. """ parsed: Final = urllib.parse.urlsplit(url) pairs: Final = tuple(urllib.parse.parse_qsl(parsed.query, keep_blank_values=True)) @@ -154,7 +241,9 @@ def translate_libpq_ssl_params(url: str) -> str: if key != "sslrootcert" ) root_cert: Final = tuple( - ("sslcert", value) for key, value in pairs if key == "sslrootcert" and "sslcert" not in keys + ("sslcert", resolve_root_cert(value, parsed.hostname or "", parsed.port or int(DEFAULT_POSTGRES_PORT))) + for key, value in pairs + if key == "sslrootcert" and "sslcert" not in keys ) strict: Final = () if "sslaccept" in keys else (("sslaccept", "strict"),) query: Final = urllib.parse.urlencode(translated + root_cert + strict) diff --git a/tests/test_litellm/proxy/db/test_db_url_settings.py b/tests/test_litellm/proxy/db/test_db_url_settings.py index ba342342366..875dca4bee3 100644 --- a/tests/test_litellm/proxy/db/test_db_url_settings.py +++ b/tests/test_litellm/proxy/db/test_db_url_settings.py @@ -11,15 +11,30 @@ clobber a pre-existing ``DATABASE_URL_READ_REPLICA``. A pre-existing ``DATABASE_URL`` (password auth) is likewise left untouched. """ +import datetime +import hashlib import os +import socket +import ssl +import tempfile +import threading import urllib.parse +from collections.abc import Iterator +from dataclasses import dataclass +from pathlib import Path +from typing import Final from unittest.mock import patch import pytest +from cryptography import x509 +from cryptography.hazmat.primitives import hashes, serialization +from cryptography.hazmat.primitives.asymmetric import ec from pydantic import ValidationError from litellm.proxy.db.db_url_settings import ( + PG_SSL_REQUEST, DatabaseURLSettings, + translate_libpq_ssl_params, unsupported_db_scheme, unsupported_db_scheme_message, ) @@ -381,9 +396,7 @@ def test_writer_password_is_percent_encoded(monkeypatch): def test_writer_url_not_clobbered_when_already_set(monkeypatch): """An operator-pinned DATABASE_URL (e.g. helm's $(VAR) assembly) always wins over the discrete fields.""" - monkeypatch.setenv( - "DATABASE_URL", "postgresql://pinned:url@db.example.com:5432/litellm_db" - ) + monkeypatch.setenv("DATABASE_URL", "postgresql://pinned:url@db.example.com:5432/litellm_db") monkeypatch.setenv("DATABASE_HOST", "writer.example.com") monkeypatch.setenv("DATABASE_USER", "litellm") monkeypatch.setenv("DATABASE_NAME", "litellm_db") @@ -515,9 +528,7 @@ def test_apply_to_env_rejects_pinned_sqlite_direct_url(monkeypatch): def test_apply_to_env_rejects_pinned_non_postgres_reader(monkeypatch): monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@writer.example.com:5432/db") - monkeypatch.setenv( - "DATABASE_URL_READ_REPLICA", "mysql://u:p@reader.example.com:3306/db" - ) + monkeypatch.setenv("DATABASE_URL_READ_REPLICA", "mysql://u:p@reader.example.com:3306/db") with pytest.raises(RuntimeError, match=r"DATABASE_URL_READ_REPLICA.*mysql"): _apply() @@ -542,15 +553,11 @@ def test_reader_inherits_writer_connection_params(monkeypatch): "DATABASE_URL", "postgresql://u:p@writer.example.com:5432/db?connection_limit=3&pool_timeout=20&pgbouncer=true", ) - monkeypatch.setenv( - "DATABASE_URL_READ_REPLICA", "postgresql://u:p@reader.example.com:5432/db" - ) + monkeypatch.setenv("DATABASE_URL_READ_REPLICA", "postgresql://u:p@reader.example.com:5432/db") _apply() - query = urllib.parse.parse_qs( - urllib.parse.urlsplit(os.environ["DATABASE_URL_READ_REPLICA"]).query - ) + query = urllib.parse.parse_qs(urllib.parse.urlsplit(os.environ["DATABASE_URL_READ_REPLICA"]).query) assert query["connection_limit"] == ["3"] assert query["pool_timeout"] == ["20"] assert query["pgbouncer"] == ["true"] @@ -568,9 +575,7 @@ def test_reader_keeps_its_own_pinned_connection_params(monkeypatch): _apply() - query = urllib.parse.parse_qs( - urllib.parse.urlsplit(os.environ["DATABASE_URL_READ_REPLICA"]).query - ) + query = urllib.parse.parse_qs(urllib.parse.urlsplit(os.environ["DATABASE_URL_READ_REPLICA"]).query) assert query["connection_limit"] == ["50"] assert query["pool_timeout"] == ["20"] @@ -776,6 +781,170 @@ def test_libpq_verify_full_and_sslrootcert_become_prisma_strict_sslcert(monkeypa } +def _issue_cert( + subject: str, issuer: x509.Certificate | None, issuer_key: ec.EllipticCurvePrivateKey | None, ca: bool +) -> tuple[x509.Certificate, ec.EllipticCurvePrivateKey]: + key: Final = ec.generate_private_key(ec.SECP256R1()) + name: Final = x509.Name((x509.NameAttribute(x509.NameOID.COMMON_NAME, subject),)) + now: Final = datetime.datetime.now(datetime.timezone.utc) + builder: Final = ( + x509.CertificateBuilder() + .subject_name(name) + .issuer_name(issuer.subject if issuer else name) + .public_key(key.public_key()) + .serial_number(x509.random_serial_number()) + .not_valid_before(now - datetime.timedelta(minutes=5)) + .not_valid_after(now + datetime.timedelta(days=1)) + .add_extension(x509.BasicConstraints(ca=ca, path_length=None), critical=True) + .add_extension(x509.SubjectAlternativeName((x509.DNSName("localhost"),)), critical=False) + ) + return builder.sign(issuer_key or key, hashes.SHA256()), key + + +def _pem(cert: x509.Certificate) -> bytes: + return cert.public_bytes(serialization.Encoding.PEM) + + +class _TlsPostgresStub: + """Answers one libpq ``SSLRequest`` with ``S`` and serves ``leaf + intermediate``.""" + + def __init__(self, chain_pem: Path, key_pem: Path) -> None: + self.context: Final = ssl.SSLContext(ssl.PROTOCOL_TLS_SERVER) + self.context.load_cert_chain(str(chain_pem), str(key_pem)) + self.listener: Final = socket.create_server(("127.0.0.1", 0)) + self.port: Final[int] = self.listener.getsockname()[1] + self.thread: Final = threading.Thread(target=self._serve, daemon=True) + self.thread.start() + + def _serve(self) -> None: + with self.listener: + while True: + try: + conn: socket.socket = self.listener.accept()[0] + except OSError: + return + with conn: + try: + if conn.recv(8) == PG_SSL_REQUEST: + conn.sendall(b"S") + with self.context.wrap_socket(conn, server_side=True) as tls: + tls.recv(1) + except OSError: + continue + + +@dataclass(frozen=True, slots=True) +class _RdsLikePki: + bundle: Path + wrong_bundle: Path + root: Path + port: int + + +@pytest.fixture +def rds_like_pki(tmp_path: Path) -> Iterator[_RdsLikePki]: + """An RDS-shaped trust setup: the server sends leaf + intermediate, the + bundle holds only self-signed roots, and the right root is not first.""" + root, root_key = _issue_cert("Real Root CA", None, None, ca=True) + decoys: Final = tuple(_issue_cert(f"Decoy Root CA {i}", None, None, ca=True)[0] for i in range(3)) + intermediate, intermediate_key = _issue_cert("Intermediate CA", root, root_key, ca=True) + leaf, leaf_key = _issue_cert("localhost", intermediate, intermediate_key, ca=False) + chain_pem: Final = tmp_path / "server-chain.pem" + chain_pem.write_bytes(_pem(leaf) + _pem(intermediate)) + key_pem: Final = tmp_path / "server.key" + key_pem.write_bytes( + leaf_key.private_bytes( + serialization.Encoding.PEM, serialization.PrivateFormat.PKCS8, serialization.NoEncryption() + ) + ) + bundle: Final = tmp_path / "global-bundle.pem" + bundle.write_bytes(b"".join(_pem(decoy) for decoy in decoys) + _pem(root)) + wrong_bundle: Final = tmp_path / "wrong-bundle.pem" + wrong_bundle.write_bytes(b"".join(_pem(decoy) for decoy in decoys)) + root_pem: Final = tmp_path / "root.pem" + root_pem.write_bytes(_pem(root)) + stub: Final = _TlsPostgresStub(chain_pem, key_pem) + yield _RdsLikePki(bundle=bundle, wrong_bundle=wrong_bundle, root=root_pem, port=stub.port) + stub.listener.close() + + +def _params(url: str) -> tuple[tuple[str, str], ...]: + return tuple(urllib.parse.parse_qsl(urllib.parse.urlsplit(url).query, keep_blank_values=True)) + + +def test_multi_root_bundle_is_pinned_to_the_root_the_server_chains_to( + monkeypatch: pytest.MonkeyPatch, rds_like_pki: _RdsLikePki +): + """Prisma's ``sslcert`` loads only the first certificate of the file, so + handing it the whole RDS bundle trusts one region's root and fails with + "unable to get local issuer certificate" everywhere else. The URL Prisma + receives must point at a single-certificate file holding the server's root.""" + monkeypatch.setenv( + "DATABASE_URL", + f"postgresql://u:p@localhost:{rds_like_pki.port}/litellm_db?sslmode=verify-full&sslrootcert={rds_like_pki.bundle}", + ) + + _apply() + + (sslmode, sslcert, sslaccept, _) = _params(os.environ["DATABASE_URL"]) + assert (sslmode, sslaccept) == (("sslmode", "require"), ("sslaccept", "strict")) + assert sslcert[0] == "sslcert" and sslcert[1] != str(rds_like_pki.bundle) + assert Path(sslcert[1]).read_bytes() == rds_like_pki.root.read_bytes() + + +def test_pinned_root_replaces_a_planted_symlink_instead_of_writing_through_it( + monkeypatch: pytest.MonkeyPatch, rds_like_pki: _RdsLikePki, tmp_path: Path +): + """The pinned file has a predictable name in a shared temp dir, so a symlink + planted there must not redirect the write onto its target.""" + monkeypatch.setattr(tempfile, "tempdir", str(tmp_path)) + root_der: Final = x509.load_pem_x509_certificate(rds_like_pki.root.read_bytes()).public_bytes( + serialization.Encoding.DER + ) + pinned: Final = tmp_path / f"litellm-sslcert-{hashlib.sha256(root_der).hexdigest()[:16]}.pem" + victim: Final = tmp_path / "victim.txt" + victim.write_text("untouched") + pinned.symlink_to(victim) + monkeypatch.setenv( + "DATABASE_URL", + f"postgresql://u:p@localhost:{rds_like_pki.port}/litellm_db?sslmode=verify-full&sslrootcert={rds_like_pki.bundle}", + ) + + _apply() + + assert ("sslcert", str(pinned)) in _params(os.environ["DATABASE_URL"]) + assert victim.read_text() == "untouched" + assert not pinned.is_symlink() and pinned.read_bytes() == rds_like_pki.root.read_bytes() + + +def test_bundle_without_the_servers_root_is_passed_through_unchanged( + monkeypatch: pytest.MonkeyPatch, rds_like_pki: _RdsLikePki +): + """Nothing in the bundle verifies the server, so no root is pinned and + Prisma keeps rejecting the connection instead of trusting a root the + operator never shipped.""" + monkeypatch.setenv( + "DATABASE_URL", + f"postgresql://u:p@localhost:{rds_like_pki.port}/litellm_db" + f"?sslmode=verify-full&sslrootcert={rds_like_pki.wrong_bundle}", + ) + + _apply() + + assert ("sslcert", str(rds_like_pki.wrong_bundle)) in _params(os.environ["DATABASE_URL"]) + + +def test_root_cert_resolver_receives_the_urls_host_and_default_port(): + def resolver(cert_path: str, host: str, port: int) -> str: + return f"/pinned/{host}/{port}{cert_path}" + + url: Final = translate_libpq_ssl_params( + "postgresql://u:p@db.example.com/litellm_db?sslmode=verify-full&sslrootcert=/certs/bundle.pem", resolver + ) + + assert ("sslcert", "/pinned/db.example.com/5432/certs/bundle.pem") in _params(url) + + def test_libpq_verify_ca_becomes_prisma_strict(monkeypatch): monkeypatch.setenv("DATABASE_URL", "postgresql://u:p@db.example.com:5432/litellm_db?sslmode=verify-ca")