From 65eae963a7da34e9d4b714d4ce0b485168efaa21 Mon Sep 17 00:00:00 2001 From: Kunal Nayyar Date: Tue, 11 Aug 2026 13:03:55 +0530 Subject: [PATCH 1/3] feat(proxy): opt-in enforce rpm/tpm when adding a model Add general_settings toggle 'enforce_rpm_tpm_on_model_add' (default false). When true, /model/new rejects a model whose rpm or tpm is missing or not a positive value, so the Admin UI Add Model form surfaces a 400 validation error instead of silently storing an unbounded model (or one with a zero/negative limit that would exclude it from routing). --- .../model_management_endpoints.py | 37 +++++++++++++++++++ .../test_model_management_endpoints.py | 33 +++++++++++++++++ .../molecules/notifications_manager.tsx | 1 + 3 files changed, 71 insertions(+) diff --git a/litellm/proxy/management_endpoints/model_management_endpoints.py b/litellm/proxy/management_endpoints/model_management_endpoints.py index 8a52b0d1abb..d9030f42f69 100644 --- a/litellm/proxy/management_endpoints/model_management_endpoints.py +++ b/litellm/proxy/management_endpoints/model_management_endpoints.py @@ -239,6 +239,38 @@ def _raise_on_strategy_router_write_violation( ) +ENFORCE_RPM_TPM_ON_MODEL_ADD_SETTING: Final = "enforce_rpm_tpm_on_model_add" +_REQUIRED_RATE_LIMIT_FIELDS: Final = ("rpm", "tpm") + + +def _raise_if_rate_limits_required_but_missing(*, litellm_params: GenericLiteLLMParams, enforced: bool) -> None: + """Require both rpm and tpm (each a positive value) when the operator opts in via config.yaml. + + Off by default, so deployments keep adding models without limits. When + ``enforce_rpm_tpm_on_model_add: true`` is set under general_settings, a model added + without both rpm and tpm set to a positive value is rejected rather than stored + unbounded (or effectively excluded from routing by a zero/negative limit). + """ + if not enforced: + return + missing: Final = tuple( + field + for field in _REQUIRED_RATE_LIMIT_FIELDS + if (value := getattr(litellm_params, field)) is None or value <= 0 + ) + if not missing: + return + raise ProxyException( + message=( + f"{' and '.join(missing)} must be set to a positive value when " + f"'{ENFORCE_RPM_TPM_ON_MODEL_ADD_SETTING}' is enabled in general_settings" + ), + type=ProxyErrorTypes.validation_error.value, + code=status.HTTP_400_BAD_REQUEST, + param=f"litellm_params.{missing[0]}", + ) + + _PTU_MODEL_INFO_FIELDS: Final = ("ptu_count", "cost_per_ptu_per_hour", "ptu_effective_from", "ptu_effective_to") @@ -1566,6 +1598,11 @@ async def add_new_model( existing_params=None, ) + _raise_if_rate_limits_required_but_missing( + litellm_params=model_params.litellm_params, + enforced=bool(general_settings.get(ENFORCE_RPM_TPM_ON_MODEL_ADD_SETTING, False)), + ) + model_response: LiteLLM_ProxyModelTable | None = None # update DB incoming_model_info: Final = model_params.model_info.model_dump(exclude_none=True) diff --git a/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py index 454849d6430..cb87dfadcfa 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py @@ -23,6 +23,7 @@ from litellm.proxy._types import ( from litellm.proxy.management_endpoints.model_management_endpoints import ( ModelManagementAuthChecks, _get_team_deployments, + _raise_if_rate_limits_required_but_missing, clear_cache, delete_team_models, ) @@ -3825,3 +3826,35 @@ class TestAutoRouterClassifierDefaultPrompt: for empty in (None, "", "{}"): response = await get_auto_router_classifier_default_prompt(context_window_size=5, tier_labels=empty) assert response.system_prompt == classification_system_prompt(5) + + +class TestEnforceRpmTpmOnModelAdd: + def test_passes_when_disabled_even_without_limits(self): + _raise_if_rate_limits_required_but_missing( + litellm_params=LiteLLM_Params(model="azure/gpt-5.2"), + enforced=False, + ) + + def test_passes_when_enabled_and_both_set(self): + _raise_if_rate_limits_required_but_missing( + litellm_params=LiteLLM_Params(model="azure/gpt-5.2", rpm=10, tpm=1000), + enforced=True, + ) + + @pytest.mark.parametrize( + "params, expected_missing", + [ + (LiteLLM_Params(model="azure/gpt-5.2"), "rpm and tpm"), + (LiteLLM_Params(model="azure/gpt-5.2", rpm=10), "tpm"), + (LiteLLM_Params(model="azure/gpt-5.2", tpm=1000), "rpm"), + (LiteLLM_Params(model="azure/gpt-5.2", rpm=0, tpm=1000), "rpm"), + (LiteLLM_Params(model="azure/gpt-5.2", rpm=10, tpm=-1), "tpm"), + ], + ) + def test_raises_when_enabled_and_missing(self, params, expected_missing): + from litellm.proxy._types import ProxyException + + with pytest.raises(ProxyException) as exc_info: + _raise_if_rate_limits_required_but_missing(litellm_params=params, enforced=True) + assert expected_missing in str(exc_info.value.message) + assert exc_info.value.code == "400" diff --git a/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx b/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx index 59b048b412c..31daee0b23b 100644 --- a/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx +++ b/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx @@ -104,6 +104,7 @@ const VALIDATION_MATCH = [ "invalid file type", "invalid field", "invalid date format", + "must be set when", ]; const NOT_FOUND_MATCH = [ From 526fc9eab192a6854f30c75aa574daf0d8d1f992 Mon Sep 17 00:00:00 2001 From: Kunal Nayyar Date: Tue, 11 Aug 2026 13:03:55 +0530 Subject: [PATCH 2/3] fix(ui): title validation errors correctly instead of Rate Limit Exceeded The /model/new endpoint returns a 400 validation error (type: validation_error) when 'rpm and tpm must be set to a positive value when enforce_rpm_tpm_on_model_add is enabled in general_settings' but the frontend's titleFor() keyword matcher mistitled it as 'Rate Limit Exceeded' because the message contains 'rpm'/'tpm' substrings, which matched the generic rate-limit keyword check before the more specific validation check could catch it. Add "'enforce_rpm_tpm_on_model_add' is enabled" to VALIDATION_MATCH so this message is classified as a Validation Error, matching the actual HTTP 400 validation_error the backend already returns. A narrow match on the setting name (rather than the generic "must be set when") avoids overriding the status-based classification of unrelated 401s, e.g. the PKCE 'GENERIC_CLIENT_ID must be set when PKCE is enabled' error. --- .../src/components/molecules/notifications_manager.tsx | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx b/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx index 31daee0b23b..aa3552da50a 100644 --- a/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx +++ b/ui/litellm-dashboard/src/components/molecules/notifications_manager.tsx @@ -104,7 +104,7 @@ const VALIDATION_MATCH = [ "invalid file type", "invalid field", "invalid date format", - "must be set when", + "'enforce_rpm_tpm_on_model_add' is enabled", ]; const NOT_FOUND_MATCH = [ From 968c96143f2e608c5f44ca1650f71ab5f9c11fe5 Mon Sep 17 00:00:00 2001 From: ryan-crabbe-berri Date: Wed, 26 Aug 2026 13:27:02 -0700 Subject: [PATCH 3/3] test(proxy): assert the rpm/tpm guard returns None on the passing paths Satisfies the TQ001 zero-assert gate --- .../test_model_management_endpoints.py | 18 ++++++++++++------ 1 file changed, 12 insertions(+), 6 deletions(-) diff --git a/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py b/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py index 79dfcbc75b2..dd7b36dd909 100644 --- a/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py +++ b/tests/test_litellm/proxy/management_endpoints/test_model_management_endpoints.py @@ -4261,15 +4261,21 @@ class TestAutoRouterClassifierDefaultPrompt: class TestEnforceRpmTpmOnModelAdd: def test_passes_when_disabled_even_without_limits(self): - _raise_if_rate_limits_required_but_missing( - litellm_params=LiteLLM_Params(model="azure/gpt-5.2"), - enforced=False, + assert ( + _raise_if_rate_limits_required_but_missing( + litellm_params=LiteLLM_Params(model="azure/gpt-5.2"), + enforced=False, + ) + is None ) def test_passes_when_enabled_and_both_set(self): - _raise_if_rate_limits_required_but_missing( - litellm_params=LiteLLM_Params(model="azure/gpt-5.2", rpm=10, tpm=1000), - enforced=True, + assert ( + _raise_if_rate_limits_required_but_missing( + litellm_params=LiteLLM_Params(model="azure/gpt-5.2", rpm=10, tpm=1000), + enforced=True, + ) + is None ) @pytest.mark.parametrize(