From 0608f0a00f2c76b50640ed7f4559f4c8551fdc44 Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 02:17:02 +0000 Subject: [PATCH 01/12] fix: reject unknown runtime router settings Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 22 +++++++++++ litellm/proxy/proxy_server.py | 21 +++++++++- litellm/router.py | 30 ++++---------- litellm/types/router.py | 29 +++++++------- .../proxy/proxy_server/test_routes_config.py | 39 +++++++++++++++++++ .../test_router_retry_policy_update.py | 21 +++++++++- ui/litellm-dashboard/src/lib/http/schema.d.ts | 2 + 7 files changed, 125 insertions(+), 39 deletions(-) diff --git a/litellm/constants.py b/litellm/constants.py index 1bd977dd9a9..a7506cb6378 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -38,6 +38,28 @@ DEFAULT_MAX_TOKENS: Final = int(os.getenv("DEFAULT_MAX_TOKENS", 4096)) DEFAULT_ALLOWED_FAILS: Final = int(os.getenv("DEFAULT_ALLOWED_FAILS", 3)) DEFAULT_REDIS_SYNC_INTERVAL: Final = int(os.getenv("DEFAULT_REDIS_SYNC_INTERVAL", 1)) DEFAULT_COOLDOWN_TIME_SECONDS: Final = int(os.getenv("DEFAULT_COOLDOWN_TIME_SECONDS", 5)) +RUNTIME_UPDATABLE_ROUTER_SETTINGS: Final[frozenset[str]] = frozenset( + { + "routing_strategy_args", + "routing_strategy", + "routing_groups", + "allowed_fails", + "cooldown_time", + "num_retries", + "timeout", + "max_retries", + "retry_after", + "fallbacks", + "context_window_fallbacks", + "retry_policy", + "model_group_retry_policy", + "model_group_alias", + "enable_weighted_failover", + "enable_tag_filtering", + "tag_routing_prefix", + "optional_pre_call_checks", + } +) DEFAULT_REPLICATE_POLLING_RETRIES: Final = int(os.getenv("DEFAULT_REPLICATE_POLLING_RETRIES", 5)) DEFAULT_REPLICATE_POLLING_DELAY_SECONDS: Final = int(os.getenv("DEFAULT_REPLICATE_POLLING_DELAY_SECONDS", 1)) DEFAULT_IMAGE_TOKEN_COUNT: Final = int(os.getenv("DEFAULT_IMAGE_TOKEN_COUNT", 250)) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index 77a80ea0052..9de6b38265a 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -39,7 +39,7 @@ from typing import ( import anyio import websockets import websockets.exceptions -from pydantic import BaseModel, Json, JsonValue, ValidationError +from pydantic import BaseModel, Json, JsonValue, TypeAdapter, ValidationError from typing_extensions import NotRequired, ReadOnly, assert_never from litellm._uuid import uuid @@ -60,6 +60,7 @@ from litellm.constants import ( LITELLM_SETTINGS_SAFE_DB_OVERRIDES, LITELLM_UI_ALLOW_HEADERS, LITELLM_UI_SESSION_DURATION, + RUNTIME_UPDATABLE_ROUTER_SETTINGS, ) from litellm.litellm_core_utils.litellm_logging import ( _init_custom_logger_compatible_class, @@ -16207,6 +16208,7 @@ async def invitation_delete( ) async def update_config( config_info: ConfigYAML, + request: Request, user_api_key_dict: UserAPIKeyAuth = Depends(user_api_key_auth), ): """ @@ -16218,6 +16220,23 @@ async def update_config( a side effect of an unrelated update. """ global llm_router, llm_model_list, general_settings, proxy_config, proxy_logging_obj, master_key, prisma_client + request_body: Final[Mapping[str, JsonValue]] = TypeAdapter(Mapping[str, JsonValue]).validate_python( + await request.json() + ) + raw_router_settings: Final = request_body.get("router_settings") + if isinstance(raw_router_settings, dict): + unsupported_router_settings: Final = sorted(set(raw_router_settings) - RUNTIME_UPDATABLE_ROUTER_SETTINGS) + if unsupported_router_settings: + raise HTTPException( + status_code=400, + detail={ + "error": ( + f"Unsupported router settings: {', '.join(unsupported_router_settings)} " + "are not runtime-updatable router settings" + ) + }, + ) + try: if user_api_key_dict.user_role != LitellmUserRoles.PROXY_ADMIN: raise HTTPException(status_code=403, detail="Only proxy admins can update config") diff --git a/litellm/router.py b/litellm/router.py index 23d8907fb49..fc4f815170d 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -50,6 +50,7 @@ from litellm.constants import ( DEFAULT_HEALTH_CHECK_INTERVAL, DEFAULT_HEALTH_CHECK_STALENESS_MULTIPLIER, DEFAULT_MAX_LRU_CACHE_SIZE, + RUNTIME_UPDATABLE_ROUTER_SETTINGS, SESSION_DEPLOYMENT_AFFINITY_TTL_METADATA_KEY, ) from litellm.integrations.custom_logger import CustomLogger @@ -2072,6 +2073,10 @@ class Router: if _callback is None: continue + if self.optional_callbacks is not None and any( + isinstance(callback, type(_callback)) for callback in self.optional_callbacks + ): + continue if self.optional_callbacks is None: self.optional_callbacks = [] self.optional_callbacks.append(_callback) @@ -11331,27 +11336,6 @@ class Router: """ Update the router settings. """ - # only the following settings are allowed to be configured - _allowed_settings: Final = [ - "routing_strategy_args", - "routing_strategy", - "routing_groups", - "allowed_fails", - "cooldown_time", - "num_retries", - "timeout", - "max_retries", - "retry_after", - "fallbacks", - "context_window_fallbacks", - "retry_policy", - "model_group_retry_policy", - "model_group_alias", - "enable_weighted_failover", - "enable_tag_filtering", - "tag_routing_prefix", - ] - _int_settings: Final = [ "timeout", "num_retries", @@ -11364,13 +11348,15 @@ class Router: rebuild_routing_groups = False relink_lar1_from_args = False for var in kwargs: - if var in _allowed_settings: + if var in RUNTIME_UPDATABLE_ROUTER_SETTINGS: if var in _int_settings: _casted_value = int(kwargs[var]) setattr(self, var, _casted_value) elif var == "routing_groups": self._routing_groups_input = kwargs[var] rebuild_routing_groups = True + elif var == "optional_pre_call_checks": + self.add_optional_pre_call_checks(kwargs[var]) elif var == "retry_policy": value = kwargs[var] if isinstance(value, dict): diff --git a/litellm/types/router.py b/litellm/types/router.py index e0957383aac..2a5f264cee3 100644 --- a/litellm/types/router.py +++ b/litellm/types/router.py @@ -106,6 +106,20 @@ class RetryPolicy(BaseModel): InternalServerErrorRetries: int | None = None +OptionalPreCallChecks = list[ + Literal[ + "prompt_caching", + "router_budget_limiting", + "responses_api_deployment_check", + "deployment_affinity", + "session_affinity", + "forward_client_headers_by_model_group", + "enforce_model_rate_limits", + "encrypted_content_affinity", + ] +] + + class UpdateRouterConfig(BaseModel): """ Set of params that you can modify via `router.update_settings()`. @@ -128,6 +142,7 @@ class UpdateRouterConfig(BaseModel): model_group_alias: dict[str, str | dict] | None = {} enable_tag_filtering: bool | None = None tag_routing_prefix: str | None = None + optional_pre_call_checks: OptionalPreCallChecks | None = None model_config = ConfigDict(protected_namespaces=()) @@ -869,20 +884,6 @@ class FallbackAccessCheck(Protocol): async def __call__(self, *, model: str, request_kwargs: Mapping[str, object], llm_router: "Router") -> bool: ... -OptionalPreCallChecks = list[ - Literal[ - "prompt_caching", - "router_budget_limiting", - "responses_api_deployment_check", - "deployment_affinity", - "session_affinity", - "forward_client_headers_by_model_group", - "enforce_model_rate_limits", - "encrypted_content_affinity", - ] -] - - class LiteLLM_RouterFileObject(TypedDict, total=False): """ Tracking the litellm params hash, used for mapping the file id to the right model diff --git a/tests/test_litellm/proxy/proxy_server/test_routes_config.py b/tests/test_litellm/proxy/proxy_server/test_routes_config.py index ad3c470acf3..0df8fb663e2 100644 --- a/tests/test_litellm/proxy/proxy_server/test_routes_config.py +++ b/tests/test_litellm/proxy/proxy_server/test_routes_config.py @@ -60,6 +60,45 @@ def test_config_update_happy_admin(client, auth_as, mock_prisma, monkeypatch): assert normalize(response.json()) == {"message": "Config updated successfully"} +def test_config_update_persists_optional_pre_call_checks(client, auth_as, mock_prisma, monkeypatch): + from litellm.proxy import proxy_server as ps + from litellm.proxy._types import LitellmUserRoles + + table = _install_litellm_config(mock_prisma) + monkeypatch.setattr(ps, "prisma_client", mock_prisma) + fake_proxy_config = MagicMock() + fake_proxy_config.add_deployment = AsyncMock() + monkeypatch.setattr(ps, "proxy_config", fake_proxy_config) + + with auth_as(LitellmUserRoles.PROXY_ADMIN): + response = client.post( + "/config/update", + json={"router_settings": {"optional_pre_call_checks": ["prompt_caching"]}}, + ) + + assert response.status_code == 200 + persisted = json.loads(table.upsert.call_args.kwargs["data"]["create"]["param_value"]) + assert persisted["optional_pre_call_checks"] == ["prompt_caching"] + + +def test_config_update_rejects_unknown_router_setting(client, auth_as, mock_prisma, monkeypatch): + from litellm.proxy import proxy_server as ps + from litellm.proxy._types import LitellmUserRoles + + table = _install_litellm_config(mock_prisma) + monkeypatch.setattr(ps, "prisma_client", mock_prisma) + + with auth_as(LitellmUserRoles.PROXY_ADMIN): + response = client.post( + "/config/update", + json={"router_settings": {"optional_precall_checks": ["prompt_caching"]}}, + ) + + assert response.status_code == 400 + assert "optional_precall_checks" in response.json()["detail"]["error"] + table.upsert.assert_not_called() + + def test_config_update_non_admin_forbidden(client, auth_as, mock_prisma, monkeypatch): """POST /config/update by a non-admin caller is rejected; the error surfaces as a ProxyException with the admin-only message.""" diff --git a/tests/test_litellm/test_router_retry_policy_update.py b/tests/test_litellm/test_router_retry_policy_update.py index 1b98b8c1ae8..1b014cd8401 100644 --- a/tests/test_litellm/test_router_retry_policy_update.py +++ b/tests/test_litellm/test_router_retry_policy_update.py @@ -26,8 +26,8 @@ from unittest.mock import AsyncMock, MagicMock import pytest from pydantic import ValidationError - import litellm +from litellm.router_utils.pre_call_checks.prompt_caching_deployment_check import PromptCachingDeploymentCheck from litellm.types.router import RetryPolicy, UpdateRouterConfig # --------------------------------------------------------------------------- @@ -100,6 +100,19 @@ def _build_router() -> litellm.Router: ) +def test_update_settings_adds_optional_pre_call_check_once(): + router = _build_router() + + router.update_settings(num_retries=7, optional_pre_call_checks=["prompt_caching"]) + router.update_settings(optional_pre_call_checks=["prompt_caching"]) + + prompt_caching_callbacks = [ + callback for callback in router.optional_callbacks if isinstance(callback, PromptCachingDeploymentCheck) + ] + assert len(prompt_caching_callbacks) == 1 + assert router.num_retries == 7 + + def test_update_settings_persists_retry_policy_dict(): """When the proxy's ``_add_router_settings_from_db_config`` calls ``llm_router.update_settings(retry_policy={...})`` after reading the @@ -228,7 +241,7 @@ async def test_config_update_persists_and_reads_back_retry_policy(monkeypatch): """The exact global retry_policy save the UI performs must survive the real ``/config/update`` -> DB -> apply -> ``/get/config/callbacks`` path, not snap back to the ``num_retries`` fallback the ticket reported.""" - import litellm.proxy.proxy_server as proxy_server + from litellm.proxy import proxy_server from litellm.proxy._types import ConfigYAML, LitellmUserRoles, UserAPIKeyAuth router = _build_router() @@ -255,8 +268,12 @@ async def test_config_update_persists_and_reads_back_retry_policy(monkeypatch): RateLimitErrorRetries=7, ) ) + request = MagicMock() + request.json = AsyncMock(return_value={"router_settings": {"retry_policy": posted.model_dump()}}) + await proxy_server.update_config( config_info=ConfigYAML(router_settings=posted), + request=request, user_api_key_dict=UserAPIKeyAuth(user_role=LitellmUserRoles.PROXY_ADMIN, api_key="sk-1234"), ) diff --git a/ui/litellm-dashboard/src/lib/http/schema.d.ts b/ui/litellm-dashboard/src/lib/http/schema.d.ts index 6f044fec3f3..bde7fd611d5 100644 --- a/ui/litellm-dashboard/src/lib/http/schema.d.ts +++ b/ui/litellm-dashboard/src/lib/http/schema.d.ts @@ -37473,6 +37473,8 @@ export interface components { } | null; /** Num Retries */ num_retries?: number | null; + /** Optional Pre Call Checks */ + optional_pre_call_checks?: ("prompt_caching" | "router_budget_limiting" | "responses_api_deployment_check" | "deployment_affinity" | "session_affinity" | "forward_client_headers_by_model_group" | "enforce_model_rate_limits" | "encrypted_content_affinity")[] | null; /** Retry After */ retry_after?: number | null; retry_policy?: components["schemas"]["RetryPolicy"] | null; From cb511f70ccbee8cc9257b52cc6ad5d7721219ce3 Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 02:20:51 +0000 Subject: [PATCH 02/12] fix: preserve config update authorization order Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 44 +++++++++---------- litellm/proxy/proxy_server.py | 34 +++++++------- .../proxy/proxy_server/test_routes_config.py | 19 +++++++- .../test_router_retry_policy_update.py | 3 +- 4 files changed, 59 insertions(+), 41 deletions(-) diff --git a/litellm/constants.py b/litellm/constants.py index a7506cb6378..5b44e8b5f51 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -9,6 +9,28 @@ DEFAULT_HEALTH_CHECK_PROMPT: Final = str(os.getenv("DEFAULT_HEALTH_CHECK_PROMPT" AZURE_DEFAULT_RESPONSES_API_VERSION: Final = str(os.getenv("AZURE_DEFAULT_RESPONSES_API_VERSION", "preview")) ROUTER_MAX_FALLBACKS: Final = int(os.getenv("ROUTER_MAX_FALLBACKS", 5)) ROUTER_FALLBACK_ERROR_DETAIL_MAX_CHARS: Final = 2000 +RUNTIME_UPDATABLE_ROUTER_SETTINGS: Final[frozenset[str]] = frozenset( + { + "routing_strategy_args", + "routing_strategy", + "routing_groups", + "allowed_fails", + "cooldown_time", + "num_retries", + "timeout", + "max_retries", + "retry_after", + "fallbacks", + "context_window_fallbacks", + "retry_policy", + "model_group_retry_policy", + "model_group_alias", + "enable_weighted_failover", + "enable_tag_filtering", + "tag_routing_prefix", + "optional_pre_call_checks", + } +) DEFAULT_BATCH_SIZE: Final = int(os.getenv("DEFAULT_BATCH_SIZE", 512)) DEFAULT_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_FLUSH_INTERVAL_SECONDS", 5)) DEFAULT_S3_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_S3_FLUSH_INTERVAL_SECONDS", 10)) @@ -38,28 +60,6 @@ DEFAULT_MAX_TOKENS: Final = int(os.getenv("DEFAULT_MAX_TOKENS", 4096)) DEFAULT_ALLOWED_FAILS: Final = int(os.getenv("DEFAULT_ALLOWED_FAILS", 3)) DEFAULT_REDIS_SYNC_INTERVAL: Final = int(os.getenv("DEFAULT_REDIS_SYNC_INTERVAL", 1)) DEFAULT_COOLDOWN_TIME_SECONDS: Final = int(os.getenv("DEFAULT_COOLDOWN_TIME_SECONDS", 5)) -RUNTIME_UPDATABLE_ROUTER_SETTINGS: Final[frozenset[str]] = frozenset( - { - "routing_strategy_args", - "routing_strategy", - "routing_groups", - "allowed_fails", - "cooldown_time", - "num_retries", - "timeout", - "max_retries", - "retry_after", - "fallbacks", - "context_window_fallbacks", - "retry_policy", - "model_group_retry_policy", - "model_group_alias", - "enable_weighted_failover", - "enable_tag_filtering", - "tag_routing_prefix", - "optional_pre_call_checks", - } -) DEFAULT_REPLICATE_POLLING_RETRIES: Final = int(os.getenv("DEFAULT_REPLICATE_POLLING_RETRIES", 5)) DEFAULT_REPLICATE_POLLING_DELAY_SECONDS: Final = int(os.getenv("DEFAULT_REPLICATE_POLLING_DELAY_SECONDS", 1)) DEFAULT_IMAGE_TOKEN_COUNT: Final = int(os.getenv("DEFAULT_IMAGE_TOKEN_COUNT", 250)) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index 9de6b38265a..dbeb8486539 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -16220,27 +16220,27 @@ async def update_config( a side effect of an unrelated update. """ global llm_router, llm_model_list, general_settings, proxy_config, proxy_logging_obj, master_key, prisma_client - request_body: Final[Mapping[str, JsonValue]] = TypeAdapter(Mapping[str, JsonValue]).validate_python( - await request.json() - ) - raw_router_settings: Final = request_body.get("router_settings") - if isinstance(raw_router_settings, dict): - unsupported_router_settings: Final = sorted(set(raw_router_settings) - RUNTIME_UPDATABLE_ROUTER_SETTINGS) - if unsupported_router_settings: - raise HTTPException( - status_code=400, - detail={ - "error": ( - f"Unsupported router settings: {', '.join(unsupported_router_settings)} " - "are not runtime-updatable router settings" - ) - }, - ) - try: if user_api_key_dict.user_role != LitellmUserRoles.PROXY_ADMIN: raise HTTPException(status_code=403, detail="Only proxy admins can update config") + request_body: Final[Mapping[str, JsonValue]] = TypeAdapter(Mapping[str, JsonValue]).validate_python( + await request.json() + ) + raw_router_settings: Final = request_body.get("router_settings") + if isinstance(raw_router_settings, dict): + unsupported_router_settings: Final = sorted(set(raw_router_settings) - RUNTIME_UPDATABLE_ROUTER_SETTINGS) + if unsupported_router_settings: + raise HTTPException( + status_code=400, + detail={ + "error": ( + f"Unsupported router settings: {', '.join(unsupported_router_settings)} " + "are not runtime-updatable router settings" + ) + }, + ) + if prisma_client is None: raise Exception("No DB Connected") diff --git a/tests/test_litellm/proxy/proxy_server/test_routes_config.py b/tests/test_litellm/proxy/proxy_server/test_routes_config.py index 0df8fb663e2..4b0954c350a 100644 --- a/tests/test_litellm/proxy/proxy_server/test_routes_config.py +++ b/tests/test_litellm/proxy/proxy_server/test_routes_config.py @@ -95,10 +95,27 @@ def test_config_update_rejects_unknown_router_setting(client, auth_as, mock_pris ) assert response.status_code == 400 - assert "optional_precall_checks" in response.json()["detail"]["error"] + assert "optional_precall_checks" in response.json()["error"]["message"] table.upsert.assert_not_called() +def test_config_update_unknown_router_setting_non_admin_forbidden(client, auth_as, mock_prisma, monkeypatch): + from litellm.proxy import proxy_server as ps + from litellm.proxy._types import LitellmUserRoles + + _install_litellm_config(mock_prisma) + monkeypatch.setattr(ps, "prisma_client", mock_prisma) + + with auth_as(LitellmUserRoles.INTERNAL_USER): + response = client.post( + "/config/update", + json={"router_settings": {"optional_precall_checks": ["prompt_caching"]}}, + ) + + assert response.status_code == 403 + assert "admin" in response.json()["error"]["message"].lower() + + def test_config_update_non_admin_forbidden(client, auth_as, mock_prisma, monkeypatch): """POST /config/update by a non-admin caller is rejected; the error surfaces as a ProxyException with the admin-only message.""" diff --git a/tests/test_litellm/test_router_retry_policy_update.py b/tests/test_litellm/test_router_retry_policy_update.py index 1b014cd8401..e386eebf3d9 100644 --- a/tests/test_litellm/test_router_retry_policy_update.py +++ b/tests/test_litellm/test_router_retry_policy_update.py @@ -26,6 +26,7 @@ from unittest.mock import AsyncMock, MagicMock import pytest from pydantic import ValidationError + import litellm from litellm.router_utils.pre_call_checks.prompt_caching_deployment_check import PromptCachingDeploymentCheck from litellm.types.router import RetryPolicy, UpdateRouterConfig @@ -241,7 +242,7 @@ async def test_config_update_persists_and_reads_back_retry_policy(monkeypatch): """The exact global retry_policy save the UI performs must survive the real ``/config/update`` -> DB -> apply -> ``/get/config/callbacks`` path, not snap back to the ``num_retries`` fallback the ticket reported.""" - from litellm.proxy import proxy_server + import litellm.proxy.proxy_server as proxy_server from litellm.proxy._types import ConfigYAML, LitellmUserRoles, UserAPIKeyAuth router = _build_router() From 29f0110fe08d5b8f4798e20eed6d2368e4cada2d Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 02:29:41 +0000 Subject: [PATCH 03/12] test: pass request to config update test Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- tests/proxy_unit_tests/test_proxy_server.py | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/tests/proxy_unit_tests/test_proxy_server.py b/tests/proxy_unit_tests/test_proxy_server.py index 47554913419..54cce9cdd78 100644 --- a/tests/proxy_unit_tests/test_proxy_server.py +++ b/tests/proxy_unit_tests/test_proxy_server.py @@ -3076,7 +3076,9 @@ async def test_update_config_success_callback_normalization(): admin_user = UserAPIKeyAuth( user_role=LitellmUserRoles.PROXY_ADMIN, api_key="sk-test" ) - await proxy_server.update_config(config_update, user_api_key_dict=admin_user) + request = MagicMock() + request.json = AsyncMock(return_value={"litellm_settings": {"success_callback": ["SQS", "sQs"]}}) + await proxy_server.update_config(config_update, request=request, user_api_key_dict=admin_user) assert ( "litellm_settings" in upserted From e67f98feb1cb1758e253beaf005eaaad44ab3abe Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 02:44:17 +0000 Subject: [PATCH 04/12] fix: reconcile runtime pre-call checks Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/router.py | 30 +++++++++++++- .../test_router_retry_policy_update.py | 40 +++++++++++++++++++ 2 files changed, 69 insertions(+), 1 deletion(-) diff --git a/litellm/router.py b/litellm/router.py index fc4f815170d..e2865542e89 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -355,6 +355,13 @@ _PreRoutingStrategyT = TypeVar("_PreRoutingStrategyT") _ALIAS_PARAMS_NEVER_FORWARDED: Final = frozenset({"model", "api_base", "api_key", "api_version"}) _ALIAS_MARKER_FORWARDED_PARAMS_KWARG: Final = "_alias_marker_forwarded_params" +_RUNTIME_TOGGLEABLE_PRE_CALL_CHECKS: Final[Mapping[str, type[CustomLogger]]] = MappingProxyType( + { + "prompt_caching": PromptCachingDeploymentCheck, + "enforce_model_rate_limits": ModelRateLimitingCheck, + } +) + def _stream_chunks_have_generated_content(chunks: Sequence[ModelResponseStream]) -> bool: for chunk in chunks: @@ -2082,6 +2089,27 @@ class Router: self.optional_callbacks.append(_callback) litellm.logging_callback_manager.add_litellm_callback(_callback) + def set_optional_pre_call_checks(self, optional_pre_call_checks: OptionalPreCallChecks | None) -> None: + if optional_pre_call_checks is None: + return + requested: Final = frozenset(optional_pre_call_checks) + for name, callback_cls in _RUNTIME_TOGGLEABLE_PRE_CALL_CHECKS.items(): + if name not in requested: + self._remove_optional_callbacks_of_type(callback_cls) + self.add_optional_pre_call_checks(optional_pre_call_checks) + + def _remove_optional_callbacks_of_type(self, callback_cls: type[CustomLogger]) -> None: + if self.optional_callbacks is None: + return + removed: Final = [cb for cb in self.optional_callbacks if isinstance(cb, callback_cls)] + if not removed: + return + self.optional_callbacks = [cb for cb in self.optional_callbacks if not isinstance(cb, callback_cls)] + for cb in removed: + litellm.logging_callback_manager.remove_callback_from_list_by_object( + litellm.callbacks, cb, require_self=False + ) + def print_deployment(self, deployment: dict): """ returns a copy of the deployment with the api key masked @@ -11356,7 +11384,7 @@ class Router: self._routing_groups_input = kwargs[var] rebuild_routing_groups = True elif var == "optional_pre_call_checks": - self.add_optional_pre_call_checks(kwargs[var]) + self.set_optional_pre_call_checks(kwargs[var]) elif var == "retry_policy": value = kwargs[var] if isinstance(value, dict): diff --git a/tests/test_litellm/test_router_retry_policy_update.py b/tests/test_litellm/test_router_retry_policy_update.py index e386eebf3d9..2c23d0da7e7 100644 --- a/tests/test_litellm/test_router_retry_policy_update.py +++ b/tests/test_litellm/test_router_retry_policy_update.py @@ -28,6 +28,8 @@ from pydantic import ValidationError import litellm +from litellm.router_strategy.budget_limiter import RouterBudgetLimiting +from litellm.router_utils.pre_call_checks.model_rate_limit_check import ModelRateLimitingCheck from litellm.router_utils.pre_call_checks.prompt_caching_deployment_check import PromptCachingDeploymentCheck from litellm.types.router import RetryPolicy, UpdateRouterConfig @@ -114,6 +116,44 @@ def test_update_settings_adds_optional_pre_call_check_once(): assert router.num_retries == 7 +def test_update_settings_clears_omitted_toggleable_pre_call_checks(): + router = _build_router() + + router.update_settings(optional_pre_call_checks=["prompt_caching"]) + router.update_settings(optional_pre_call_checks=[]) + + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in (router.optional_callbacks or [])) + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in litellm.callbacks) + + +def test_update_settings_replaces_toggleable_pre_call_checks(): + router = _build_router() + + router.update_settings(optional_pre_call_checks=["prompt_caching"]) + router.update_settings(optional_pre_call_checks=["enforce_model_rate_limits"]) + + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in (router.optional_callbacks or [])) + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in litellm.callbacks) + assert any(isinstance(callback, ModelRateLimitingCheck) for callback in (router.optional_callbacks or [])) + + +@pytest.mark.asyncio +async def test_update_settings_preserves_router_budget_limiting_when_omitted(monkeypatch): + async def _disable_periodic_sync(*args, **kwargs): + return None + + monkeypatch.setattr( + "litellm.router_strategy.budget_limiter.RouterBudgetLimiting.periodic_sync_in_memory_spend_with_redis", + _disable_periodic_sync, + ) + router = _build_router() + + router.add_optional_pre_call_checks(["router_budget_limiting"]) + router.update_settings(optional_pre_call_checks=[]) + + assert any(isinstance(callback, RouterBudgetLimiting) for callback in (router.optional_callbacks or [])) + + def test_update_settings_persists_retry_policy_dict(): """When the proxy's ``_add_router_settings_from_db_config`` calls ``llm_router.update_settings(retry_policy={...})`` after reading the From 7b86b7f4cd7576aeba56d9721f28002a4e5c6383 Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 02:45:45 +0000 Subject: [PATCH 05/12] test: isolate router callback state Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- tests/test_litellm/test_router_retry_policy_update.py | 9 +++++++++ 1 file changed, 9 insertions(+) diff --git a/tests/test_litellm/test_router_retry_policy_update.py b/tests/test_litellm/test_router_retry_policy_update.py index 2c23d0da7e7..db710f76887 100644 --- a/tests/test_litellm/test_router_retry_policy_update.py +++ b/tests/test_litellm/test_router_retry_policy_update.py @@ -21,6 +21,7 @@ This file pins both halves of the fix. import json from dataclasses import dataclass +from typing import Final from unittest.mock import AsyncMock, MagicMock import pytest @@ -33,6 +34,14 @@ from litellm.router_utils.pre_call_checks.model_rate_limit_check import ModelRat from litellm.router_utils.pre_call_checks.prompt_caching_deployment_check import PromptCachingDeploymentCheck from litellm.types.router import RetryPolicy, UpdateRouterConfig + +@pytest.fixture(autouse=True) +def isolate_litellm_callbacks(): + callbacks_before: Final = litellm.callbacks.copy() + yield + litellm.callbacks = callbacks_before + + # --------------------------------------------------------------------------- # UpdateRouterConfig schema membership (LIT-3152 part 1) # --------------------------------------------------------------------------- From 3dea586d3be34c52156fe764051dd8d750d4570b Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 02:48:09 +0000 Subject: [PATCH 06/12] test: cover runtime callback reconciliation Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../test_router_retry_policy_update.py | 20 +++++++++++++++++++ 1 file changed, 20 insertions(+) diff --git a/tests/test_litellm/test_router_retry_policy_update.py b/tests/test_litellm/test_router_retry_policy_update.py index db710f76887..9e0bb0b9bef 100644 --- a/tests/test_litellm/test_router_retry_policy_update.py +++ b/tests/test_litellm/test_router_retry_policy_update.py @@ -135,6 +135,26 @@ def test_update_settings_clears_omitted_toggleable_pre_call_checks(): assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in litellm.callbacks) +def test_set_optional_pre_call_checks_reconciles_callback_types(): + router = _build_router() + + router.set_optional_pre_call_checks(["prompt_caching"]) + router.set_optional_pre_call_checks([]) + + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in (router.optional_callbacks or [])) + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in litellm.callbacks) + + +def test_remove_optional_pre_call_check_removes_local_and_global_callbacks(): + router = _build_router() + + router.set_optional_pre_call_checks(["prompt_caching"]) + router._remove_optional_callbacks_of_type(PromptCachingDeploymentCheck) + + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in (router.optional_callbacks or [])) + assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in litellm.callbacks) + + def test_update_settings_replaces_toggleable_pre_call_checks(): router = _build_router() From 70a4f74a0d65bd89a9e9127076efd4301b987a29 Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 02:56:23 +0000 Subject: [PATCH 07/12] test: allow callback state fixture mutation Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- tests/test_litellm/test_router_retry_policy_update.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/tests/test_litellm/test_router_retry_policy_update.py b/tests/test_litellm/test_router_retry_policy_update.py index 9e0bb0b9bef..26251290176 100644 --- a/tests/test_litellm/test_router_retry_policy_update.py +++ b/tests/test_litellm/test_router_retry_policy_update.py @@ -39,7 +39,7 @@ from litellm.types.router import RetryPolicy, UpdateRouterConfig def isolate_litellm_callbacks(): callbacks_before: Final = litellm.callbacks.copy() yield - litellm.callbacks = callbacks_before + litellm.callbacks = callbacks_before # test-quality-ok: required callback-state restoration fixture # --------------------------------------------------------------------------- From 1d3e26fd98b40d40f498e14b2470e8acc79fb9f6 Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 03:28:04 +0000 Subject: [PATCH 08/12] fix: preserve shared optional callbacks Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/router.py | 19 +++++---- .../test_router_retry_policy_update.py | 41 ++++++++++++++++++- 2 files changed, 50 insertions(+), 10 deletions(-) diff --git a/litellm/router.py b/litellm/router.py index e2865542e89..e7588d8ad5a 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -2099,16 +2099,19 @@ class Router: self.add_optional_pre_call_checks(optional_pre_call_checks) def _remove_optional_callbacks_of_type(self, callback_cls: type[CustomLogger]) -> None: - if self.optional_callbacks is None: + if self.optional_callbacks is None or not any(type(cb) is callback_cls for cb in self.optional_callbacks): return - removed: Final = [cb for cb in self.optional_callbacks if isinstance(cb, callback_cls)] - if not removed: + self.optional_callbacks = [cb for cb in self.optional_callbacks if type(cb) is not callback_cls] + if any( + router is not self and any(type(cb) is callback_cls for cb in (router.optional_callbacks or [])) + for router in tuple(_live_routers) + ): return - self.optional_callbacks = [cb for cb in self.optional_callbacks if not isinstance(cb, callback_cls)] - for cb in removed: - litellm.logging_callback_manager.remove_callback_from_list_by_object( - litellm.callbacks, cb, require_self=False - ) + for cb in tuple(litellm.callbacks): + if type(cb) is callback_cls: + litellm.logging_callback_manager.remove_callback_from_list_by_object( + litellm.callbacks, cb, require_self=False + ) def print_deployment(self, deployment: dict): """ diff --git a/tests/test_litellm/test_router_retry_policy_update.py b/tests/test_litellm/test_router_retry_policy_update.py index 26251290176..be568134763 100644 --- a/tests/test_litellm/test_router_retry_policy_update.py +++ b/tests/test_litellm/test_router_retry_policy_update.py @@ -151,8 +151,45 @@ def test_remove_optional_pre_call_check_removes_local_and_global_callbacks(): router.set_optional_pre_call_checks(["prompt_caching"]) router._remove_optional_callbacks_of_type(PromptCachingDeploymentCheck) - assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in (router.optional_callbacks or [])) - assert not any(isinstance(callback, PromptCachingDeploymentCheck) for callback in litellm.callbacks) + assert not any(type(callback) is PromptCachingDeploymentCheck for callback in (router.optional_callbacks or [])) + assert not any(type(callback) is PromptCachingDeploymentCheck for callback in litellm.callbacks) + + +def test_remove_optional_pre_call_check_keeps_global_callback_for_another_router(): + router_a = _build_router() + router_b = _build_router() + + router_a.update_settings(optional_pre_call_checks=["prompt_caching"]) + router_b.update_settings(optional_pre_call_checks=["prompt_caching"]) + + router_a.update_settings(optional_pre_call_checks=[]) + + assert not any(type(callback) is PromptCachingDeploymentCheck for callback in (router_a.optional_callbacks or [])) + assert any(type(callback) is PromptCachingDeploymentCheck for callback in (router_b.optional_callbacks or [])) + assert any(type(callback) is PromptCachingDeploymentCheck for callback in litellm.callbacks) + + router_b.update_settings(optional_pre_call_checks=[]) + + assert not any(type(callback) is PromptCachingDeploymentCheck for callback in (router_b.optional_callbacks or [])) + assert not any(type(callback) is PromptCachingDeploymentCheck for callback in litellm.callbacks) + + +def test_remove_optional_pre_call_check_keeps_global_callback_when_second_router_clears_first(): + router_a = _build_router() + router_b = _build_router() + + router_a.update_settings(optional_pre_call_checks=["prompt_caching"]) + router_b.update_settings(optional_pre_call_checks=["prompt_caching"]) + + router_b.update_settings(optional_pre_call_checks=[]) + + assert any(type(callback) is PromptCachingDeploymentCheck for callback in (router_a.optional_callbacks or [])) + assert not any(type(callback) is PromptCachingDeploymentCheck for callback in (router_b.optional_callbacks or [])) + assert any(type(callback) is PromptCachingDeploymentCheck for callback in litellm.callbacks) + + router_a.update_settings(optional_pre_call_checks=[]) + + assert not any(type(callback) is PromptCachingDeploymentCheck for callback in litellm.callbacks) def test_update_settings_replaces_toggleable_pre_call_checks(): From 974b331a4d09e2883d6fe84bb87ce57cc49ab5bb Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 06:15:52 +0000 Subject: [PATCH 09/12] fix: accept persistable router settings Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 1 + litellm/proxy/proxy_server.py | 32 +++++++++----- .../proxy/proxy_server/test_routes_config.py | 43 +++++++++++++++++++ 3 files changed, 65 insertions(+), 11 deletions(-) diff --git a/litellm/constants.py b/litellm/constants.py index 5b44e8b5f51..b6cb6b32187 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -31,6 +31,7 @@ RUNTIME_UPDATABLE_ROUTER_SETTINGS: Final[frozenset[str]] = frozenset( "optional_pre_call_checks", } ) +ROUTER_SETTINGS_MANAGED_OUTSIDE_CONFIG: Final[frozenset[str]] = frozenset({"model_list", "search_tools"}) DEFAULT_BATCH_SIZE: Final = int(os.getenv("DEFAULT_BATCH_SIZE", 512)) DEFAULT_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_FLUSH_INTERVAL_SECONDS", 5)) DEFAULT_S3_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_S3_FLUSH_INTERVAL_SECONDS", 10)) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index dbeb8486539..ae5e08ab566 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -254,6 +254,7 @@ from litellm.constants import ( PROXY_BUDGET_RESCHEDULER_MAX_TIME, PROXY_BUDGET_RESCHEDULER_MIN_TIME, PROXY_CONFIG_RELOAD_INTERVAL_SECONDS, + ROUTER_SETTINGS_MANAGED_OUTSIDE_CONFIG, USER_SPEND_ALERTS_JOB_ID, WEEKLY_SPEND_REPORT_JOB_ID, ) @@ -5711,13 +5712,9 @@ class ProxyConfig: router_settings: Final = config.get("router_settings", None) if router_settings and isinstance(router_settings, dict): - # model list and search_tools already set - exclude_args: Final = { - "model_list", - "search_tools", - } - - available_args: Final = [x for x in litellm.Router.get_valid_args() if x not in exclude_args] + available_args: Final = [ + x for x in litellm.Router.get_valid_args() if x not in ROUTER_SETTINGS_MANAGED_OUTSIDE_CONFIG + ] for k, v in router_settings.items(): if k in available_args: @@ -16229,14 +16226,17 @@ async def update_config( ) raw_router_settings: Final = request_body.get("router_settings") if isinstance(raw_router_settings, dict): - unsupported_router_settings: Final = sorted(set(raw_router_settings) - RUNTIME_UPDATABLE_ROUTER_SETTINGS) + supported_router_settings: Final = RUNTIME_UPDATABLE_ROUTER_SETTINGS | ( + frozenset(litellm.Router.get_valid_args()) - ROUTER_SETTINGS_MANAGED_OUTSIDE_CONFIG + ) + unsupported_router_settings: Final = sorted(set(raw_router_settings) - supported_router_settings) if unsupported_router_settings: raise HTTPException( status_code=400, detail={ "error": ( f"Unsupported router settings: {', '.join(unsupported_router_settings)} " - "are not runtime-updatable router settings" + "are not valid router settings" ) }, ) @@ -16342,10 +16342,20 @@ async def update_config( ) # router_settings: merge existing + request, request wins. - if config_info.router_settings is not None: + if isinstance(raw_router_settings, dict): existing = await _read_section("router_settings") before_router_settings: Final = copy.deepcopy(existing) - updates = config_info.router_settings.dict(exclude_none=True) + typed_router_settings: Final = ( + config_info.router_settings.dict(exclude_none=True) + if config_info.router_settings is not None + else {} + ) + raw_router_settings_without_none: Final = { + key: value + for key, value in raw_router_settings.items() + if key not in typed_router_settings and value is not None + } + updates: Final = {**typed_router_settings, **raw_router_settings_without_none} new_router_settings: Final = {**existing, **updates} await _upsert_section("router_settings", new_router_settings) asyncio.create_task( diff --git a/tests/test_litellm/proxy/proxy_server/test_routes_config.py b/tests/test_litellm/proxy/proxy_server/test_routes_config.py index 4b0954c350a..6166513d229 100644 --- a/tests/test_litellm/proxy/proxy_server/test_routes_config.py +++ b/tests/test_litellm/proxy/proxy_server/test_routes_config.py @@ -81,6 +81,49 @@ def test_config_update_persists_optional_pre_call_checks(client, auth_as, mock_p assert persisted["optional_pre_call_checks"] == ["prompt_caching"] +def test_config_update_persists_model_group_affinity_config(client, auth_as, mock_prisma, monkeypatch): + from litellm.proxy import proxy_server as ps + from litellm.proxy._types import LitellmUserRoles + + table = _install_litellm_config(mock_prisma) + monkeypatch.setattr(ps, "prisma_client", mock_prisma) + fake_proxy_config = MagicMock() + fake_proxy_config.add_deployment = AsyncMock() + monkeypatch.setattr(ps, "proxy_config", fake_proxy_config) + + model_group_affinity_config = {"gpt-4": ["session_affinity"]} + with auth_as(LitellmUserRoles.PROXY_ADMIN): + response = client.post( + "/config/update", + json={"router_settings": {"model_group_affinity_config": model_group_affinity_config}}, + ) + + assert response.status_code == 200 + persisted = json.loads(table.upsert.call_args.kwargs["data"]["create"]["param_value"]) + assert persisted["model_group_affinity_config"] == model_group_affinity_config + + +def test_config_update_persists_disable_cooldowns(client, auth_as, mock_prisma, monkeypatch): + from litellm.proxy import proxy_server as ps + from litellm.proxy._types import LitellmUserRoles + + table = _install_litellm_config(mock_prisma) + monkeypatch.setattr(ps, "prisma_client", mock_prisma) + fake_proxy_config = MagicMock() + fake_proxy_config.add_deployment = AsyncMock() + monkeypatch.setattr(ps, "proxy_config", fake_proxy_config) + + with auth_as(LitellmUserRoles.PROXY_ADMIN): + response = client.post( + "/config/update", + json={"router_settings": {"disable_cooldowns": True}}, + ) + + assert response.status_code == 200 + persisted = json.loads(table.upsert.call_args.kwargs["data"]["create"]["param_value"]) + assert persisted["disable_cooldowns"] is True + + def test_config_update_rejects_unknown_router_setting(client, auth_as, mock_prisma, monkeypatch): from litellm.proxy import proxy_server as ps from litellm.proxy._types import LitellmUserRoles From 95e09db661471d8fbdaa1a93f79bf8e49ecdd0eb Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 06:56:27 +0000 Subject: [PATCH 10/12] style: apply ruff format to router settings merge Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/proxy/proxy_server.py | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index ae5e08ab566..1bdf2ba9987 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -16346,9 +16346,7 @@ async def update_config( existing = await _read_section("router_settings") before_router_settings: Final = copy.deepcopy(existing) typed_router_settings: Final = ( - config_info.router_settings.dict(exclude_none=True) - if config_info.router_settings is not None - else {} + config_info.router_settings.dict(exclude_none=True) if config_info.router_settings is not None else {} ) raw_router_settings_without_none: Final = { key: value From aba9644297e4e709232775031cb03f028dd1cbd6 Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 07:10:04 +0000 Subject: [PATCH 11/12] fix: avoid router settings update name collision Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/proxy/proxy_server.py | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py index 1bdf2ba9987..6632b209905 100644 --- a/litellm/proxy/proxy_server.py +++ b/litellm/proxy/proxy_server.py @@ -16353,8 +16353,8 @@ async def update_config( for key, value in raw_router_settings.items() if key not in typed_router_settings and value is not None } - updates: Final = {**typed_router_settings, **raw_router_settings_without_none} - new_router_settings: Final = {**existing, **updates} + router_settings_updates: Final = {**typed_router_settings, **raw_router_settings_without_none} + new_router_settings: Final = {**existing, **router_settings_updates} await _upsert_section("router_settings", new_router_settings) asyncio.create_task( create_config_audit_log( From 385957e830c6b5edae9455dae8361527908cd4bb Mon Sep 17 00:00:00 2001 From: mateo Date: Wed, 2 Sep 2026 07:49:25 +0000 Subject: [PATCH 12/12] fix: reject constructor-managed router settings Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/constants.py | 11 +++++- .../proxy/proxy_server/test_routes_config.py | 36 +++++++++++++++++++ 2 files changed, 46 insertions(+), 1 deletion(-) diff --git a/litellm/constants.py b/litellm/constants.py index b6cb6b32187..1c1939bd350 100644 --- a/litellm/constants.py +++ b/litellm/constants.py @@ -31,7 +31,16 @@ RUNTIME_UPDATABLE_ROUTER_SETTINGS: Final[frozenset[str]] = frozenset( "optional_pre_call_checks", } ) -ROUTER_SETTINGS_MANAGED_OUTSIDE_CONFIG: Final[frozenset[str]] = frozenset({"model_list", "search_tools"}) +ROUTER_SETTINGS_MANAGED_OUTSIDE_CONFIG: Final[frozenset[str]] = frozenset( + { + "model_list", + "search_tools", + "assistants_config", + "router_general_settings", + "ignore_invalid_deployments", + "fallback_access_check", + } +) DEFAULT_BATCH_SIZE: Final = int(os.getenv("DEFAULT_BATCH_SIZE", 512)) DEFAULT_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_FLUSH_INTERVAL_SECONDS", 5)) DEFAULT_S3_FLUSH_INTERVAL_SECONDS: Final = int(os.getenv("DEFAULT_S3_FLUSH_INTERVAL_SECONDS", 10)) diff --git a/tests/test_litellm/proxy/proxy_server/test_routes_config.py b/tests/test_litellm/proxy/proxy_server/test_routes_config.py index 6166513d229..dcb63b8ca82 100644 --- a/tests/test_litellm/proxy/proxy_server/test_routes_config.py +++ b/tests/test_litellm/proxy/proxy_server/test_routes_config.py @@ -124,6 +124,42 @@ def test_config_update_persists_disable_cooldowns(client, auth_as, mock_prisma, assert persisted["disable_cooldowns"] is True +def test_config_update_rejects_assistants_config(client, auth_as, mock_prisma, monkeypatch): + from litellm.proxy import proxy_server as ps + from litellm.proxy._types import LitellmUserRoles + + table = _install_litellm_config(mock_prisma) + monkeypatch.setattr(ps, "prisma_client", mock_prisma) + + with auth_as(LitellmUserRoles.PROXY_ADMIN): + response = client.post( + "/config/update", + json={"router_settings": {"assistants_config": {"enabled": True}}}, + ) + + assert response.status_code == 400 + assert "assistants_config" in response.json()["error"]["message"] + table.upsert.assert_not_called() + + +def test_config_update_rejects_router_general_settings(client, auth_as, mock_prisma, monkeypatch): + from litellm.proxy import proxy_server as ps + from litellm.proxy._types import LitellmUserRoles + + table = _install_litellm_config(mock_prisma) + monkeypatch.setattr(ps, "prisma_client", mock_prisma) + + with auth_as(LitellmUserRoles.PROXY_ADMIN): + response = client.post( + "/config/update", + json={"router_settings": {"router_general_settings": {"async_only_mode": True}}}, + ) + + assert response.status_code == 400 + assert "router_general_settings" in response.json()["error"]["message"] + table.upsert.assert_not_called() + + def test_config_update_rejects_unknown_router_setting(client, auth_as, mock_prisma, monkeypatch): from litellm.proxy import proxy_server as ps from litellm.proxy._types import LitellmUserRoles