From a57483d1c80032945ef2180707acfd71b6cc2548 Mon Sep 17 00:00:00 2001 From: yucheng Date: Thu, 17 Sep 2026 05:24:05 +0000 Subject: [PATCH 1/3] fix(cost): price Azure PTU spillover requests at standard token rates Azure PTU deployments carry zeroed per-token pricing because the reservation is billed flat by the hour. When Azure spills a request onto pay-as-you-go capacity it returns x-ms-is-spilled-over: true, and that traffic was still priced at zero. The response cost calculator now detects the spillover header on the result's hidden params or the logged provider response headers and skips the zeroed custom pricing only for genuine PTU deployments while the feature flag is on. Azure sync streaming now also records response headers on the logging object, matching the async paths. Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/litellm_core_utils/litellm_logging.py | 34 +++- litellm/litellm_core_utils/ptu_pricing.py | 20 +++ litellm/llms/azure/azure.py | 1 + .../test_litellm_logging.py | 152 ++++++++++++++++++ .../litellm_core_utils/test_ptu_pricing.py | 37 ++++- tests/test_litellm/llms/azure/test_azure.py | 54 +++++++ 6 files changed, 293 insertions(+), 5 deletions(-) create mode 100644 tests/test_litellm/llms/azure/test_azure.py diff --git a/litellm/litellm_core_utils/litellm_logging.py b/litellm/litellm_core_utils/litellm_logging.py index 40621a2f68d..bbae3021677 100644 --- a/litellm/litellm_core_utils/litellm_logging.py +++ b/litellm/litellm_core_utils/litellm_logging.py @@ -90,6 +90,7 @@ from litellm.litellm_core_utils.logging_utils import ( truncate_base64_in_messages_async, ) from litellm.litellm_core_utils.model_param_helper import ModelParamHelper +from litellm.litellm_core_utils.ptu_pricing import is_spilled_over_ptu_request from litellm.litellm_core_utils.redact_messages import ( redact_message_input_output_from_custom_logger, redact_message_input_output_from_logging, @@ -1746,8 +1747,14 @@ class Logging(LiteLLMLoggingBaseClass): if transformed_result is not None: result = transformed_result + result_hidden_params: Final = getattr(result, "_hidden_params", None) or MappingProxyType({}) + result_additional_headers: Final = ( + result_hidden_params.get("additional_headers") + if isinstance(result_hidden_params, dict) + else getattr(result_hidden_params, "additional_headers", None) + ) if isinstance(result, (BaseModel, HttpxBinaryResponseContent)) and hasattr(result, "_hidden_params"): - hidden_params: Final = getattr(result, "_hidden_params", {}) + hidden_params: Final = result_hidden_params if ( "response_cost" in hidden_params and hidden_params["response_cost"] is not None ): # use cost if already calculated @@ -1762,8 +1769,17 @@ class Logging(LiteLLMLoggingBaseClass): router_model_id = self.get_router_model_id() ## RESPONSE COST ## - custom_pricing: Final = use_custom_pricing_for_model( - litellm_params=(self.litellm_params if hasattr(self, "litellm_params") else None) + spilled_over: Final = is_spilled_over_ptu_request( + model_info=_deployment_model_info(self.litellm_params if hasattr(self, "litellm_params") else None), + response_headers=self.model_call_details.get("response_headers"), + additional_headers=result_additional_headers, + ) + custom_pricing: Final = ( + False + if spilled_over + else use_custom_pricing_for_model( + litellm_params=(self.litellm_params if hasattr(self, "litellm_params") else None) + ) ) prompt = self._prompt_for_cost_calculation() @@ -5257,6 +5273,18 @@ def _get_custom_logger_settings_from_proxy_server(callback_name: str) -> dict: return {} +def _deployment_model_info(litellm_params: dict | None) -> Mapping[str, object]: + """The router-stamped deployment model_info from whichever metadata field carries it.""" + if litellm_params is None: + return MappingProxyType({}) + for metadata_key in ("metadata", "litellm_metadata"): + if not isinstance(metadata := litellm_params.get(metadata_key), Mapping): + continue + if model_info := metadata.get("model_info"): + return model_info + return MappingProxyType({}) + + def use_custom_pricing_for_model(litellm_params: dict | None) -> bool: """ Check if the model uses custom pricing diff --git a/litellm/litellm_core_utils/ptu_pricing.py b/litellm/litellm_core_utils/ptu_pricing.py index f545ba4aa3b..2cf86c30e9c 100644 --- a/litellm/litellm_core_utils/ptu_pricing.py +++ b/litellm/litellm_core_utils/ptu_pricing.py @@ -17,6 +17,7 @@ from litellm.types.router import ModelInfo from litellm.types.utils import CustomPricingLiteLLMParams, MirroredPricingParams PTU_COST_ATTRIBUTION_ENV_VAR: Final = "LITELLM_ENABLE_PTU_COST_ATTRIBUTION" +AZURE_SPILLOVER_HEADER: Final = "x-ms-is-spilled-over" def is_ptu_cost_attribution_enabled() -> bool: @@ -235,3 +236,22 @@ def zeroed_ptu_pricing( ), } ) + + +def is_spilled_over_ptu_request( + model_info: Mapping[str, object], + response_headers: Mapping[str, object] | None, + additional_headers: Mapping[str, object] | None, +) -> bool: + """Whether Azure served this request from pay-as-you-go capacity, so the zeroed PTU rates must not apply.""" + if ptu_terms(model_info) is None: + return False + if not is_ptu_cost_attribution_enabled(): + return False + for headers, key in ( + (response_headers, AZURE_SPILLOVER_HEADER), + (additional_headers, f"llm_provider-{AZURE_SPILLOVER_HEADER}"), + ): + if headers is not None and str(headers.get(key)).lower() == "true": + return True + return False diff --git a/litellm/llms/azure/azure.py b/litellm/llms/azure/azure.py index 587165e6991..3cb17259b93 100644 --- a/litellm/llms/azure/azure.py +++ b/litellm/llms/azure/azure.py @@ -561,6 +561,7 @@ class AzureChatCompletion(BaseAzureLLM, BaseLLM): headers, response = self.make_sync_azure_openai_chat_completion_request( azure_client=azure_client, data=data, timeout=timeout ) + logging_obj.model_call_details["response_headers"] = headers streamwrapper: Final = CustomStreamWrapper( completion_stream=response, model=model, diff --git a/tests/test_litellm/litellm_core_utils/test_litellm_logging.py b/tests/test_litellm/litellm_core_utils/test_litellm_logging.py index aaf44b8e918..e937142e046 100644 --- a/tests/test_litellm/litellm_core_utils/test_litellm_logging.py +++ b/tests/test_litellm/litellm_core_utils/test_litellm_logging.py @@ -7229,3 +7229,155 @@ def test_add_dynamic_callback_registers_once_per_list_without_touching_the_calle assert logging_obj.dynamic_async_failure_callbacks == [callback] assert LitellmLogging._with_dynamic_callback(None, callback) == [callback] assert LitellmLogging._with_dynamic_callback((callback,), callback) == [callback] + + +class TestAzurePTUSpilloverCost: + """Azure PTU deployments price tokens at zero because the reservation is billed flat. + + A request Azure spills onto pay-as-you-go capacity must bill per token instead, so + the zeroed custom pricing has to be skipped when the provider reports spillover. + """ + + ROUTER_MODEL_ID: Final = "ptu-spill-router-model-id" + SERVED_MODEL: Final = "azure/spill-served-model-ptu" + PTU_MODEL_INFO: Final = { + "id": ROUTER_MODEL_ID, + "team_id": "team-1", + "ptu_count": 100, + "cost_per_ptu_per_hour": 1.0, + "ptu_effective_from": "2026-01-01", + "input_cost_per_token": 0.0, + "output_cost_per_token": 0.0, + } + EXPECTED_SPILL_COST: Final = 100 * 2e-6 + 50 * 8e-6 + + @staticmethod + def _register_models() -> None: + litellm.register_model( + model_cost={ + TestAzurePTUSpilloverCost.ROUTER_MODEL_ID: { + "input_cost_per_token": 0.0, + "output_cost_per_token": 0.0, + "litellm_provider": "azure", + "mode": "chat", + }, + TestAzurePTUSpilloverCost.SERVED_MODEL: { + "input_cost_per_token": 2e-6, + "output_cost_per_token": 8e-6, + "litellm_provider": "azure", + "mode": "chat", + }, + } + ) + + @staticmethod + def _unregister_models() -> None: + litellm.model_cost.pop(TestAzurePTUSpilloverCost.ROUTER_MODEL_ID, None) + litellm.model_cost.pop(TestAzurePTUSpilloverCost.SERVED_MODEL, None) + + def _logging_obj(self, model_info: dict, *, flag: str, litellm_rate: float, monkeypatch) -> LitellmLogging: + monkeypatch.setenv("LITELLM_ENABLE_PTU_COST_ATTRIBUTION", flag) + obj = LitellmLogging( + model=self.SERVED_MODEL, + messages=[{"role": "user", "content": "Hi"}], + stream=False, + call_type="completion", + start_time=time.time(), + litellm_call_id="ptu-spill-1", + function_id="f", + ) + obj.update_environment_variables( + model=self.SERVED_MODEL, + user="", + optional_params={}, + litellm_params={ + "api_base": "", + "metadata": {"model_info": model_info}, + "input_cost_per_token": litellm_rate, + "output_cost_per_token": litellm_rate, + }, + custom_llm_provider="azure", + ) + return obj + + @staticmethod + def _response() -> ModelResponse: + from litellm.types.utils import Usage + + return ModelResponse( + id="chatcmpl-spill-1", + created=1234567890, + model="spill-served-model-ptu", + choices=[ + { + "index": 0, + "message": {"role": "assistant", "content": "ok"}, + "finish_reason": "stop", + } + ], + usage=Usage(prompt_tokens=100, completion_tokens=50, total_tokens=150), + ) + + def test_spillover_via_response_additional_headers_bills_per_token(self, monkeypatch) -> None: + self._register_models() + try: + obj = self._logging_obj(dict(self.PTU_MODEL_INFO), flag="True", litellm_rate=0.0, monkeypatch=monkeypatch) + response = self._response() + response._hidden_params["additional_headers"] = {"llm_provider-x-ms-is-spilled-over": "true"} + + assert obj._response_cost_calculator(result=response) == pytest.approx(self.EXPECTED_SPILL_COST) + finally: + self._unregister_models() + + def test_spillover_via_streaming_response_headers_bills_per_token(self, monkeypatch) -> None: + self._register_models() + try: + obj = self._logging_obj(dict(self.PTU_MODEL_INFO), flag="True", litellm_rate=0.0, monkeypatch=monkeypatch) + obj.model_call_details["response_headers"] = { + "x-ms-is-spilled-over": "true", + "x-ms-spillover-from-deployment": "ptu-dep", + } + + assert obj._response_cost_calculator(result=self._response()) == pytest.approx(self.EXPECTED_SPILL_COST) + finally: + self._unregister_models() + + def test_non_spilled_ptu_request_stays_zero_priced(self, monkeypatch) -> None: + self._register_models() + try: + obj = self._logging_obj(dict(self.PTU_MODEL_INFO), flag="True", litellm_rate=0.0, monkeypatch=monkeypatch) + + assert obj._response_cost_calculator(result=self._response()) == 0.0 + finally: + self._unregister_models() + + def test_spillover_header_without_the_flag_stays_zero_priced(self, monkeypatch) -> None: + self._register_models() + try: + obj = self._logging_obj(dict(self.PTU_MODEL_INFO), flag="", litellm_rate=0.0, monkeypatch=monkeypatch) + response = self._response() + response._hidden_params["additional_headers"] = {"llm_provider-x-ms-is-spilled-over": "true"} + + assert obj._response_cost_calculator(result=response) == 0.0 + finally: + self._unregister_models() + + def test_spillover_header_does_not_touch_non_ptu_custom_pricing(self, monkeypatch) -> None: + self._register_models() + custom_model_id: Final = "non-ptu-custom-router-model-id" + litellm.model_cost[custom_model_id] = { + "input_cost_per_token": 1e-6, + "output_cost_per_token": 1e-6, + "litellm_provider": "azure", + "mode": "chat", + } + try: + model_info: Final = {"id": custom_model_id, "input_cost_per_token": 1e-6} + obj = self._logging_obj(model_info, flag="True", litellm_rate=1e-6, monkeypatch=monkeypatch) + response = self._response() + response._hidden_params["additional_headers"] = {"llm_provider-x-ms-is-spilled-over": "true"} + + assert obj._response_cost_calculator(result=response) == pytest.approx(150 * 1e-6) + finally: + litellm.model_cost.pop(custom_model_id, None) + self._unregister_models() diff --git a/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py b/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py index b8fb372d537..464c56d5132 100644 --- a/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py +++ b/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py @@ -7,13 +7,14 @@ from unittest.mock import patch import pytest from litellm.litellm_core_utils.ptu_pricing import ( - ptu_config_error, - ptu_identity_error, CUSTOM_PRICING_FIELDS, PTU_EMPTIED_PRICING_FIELDS, PTU_ZEROED_PRICING_FIELDS, PTU_ZEROED_TABLE_FIELDS, SEARCH_CONTEXT_SIZES, + is_spilled_over_ptu_request, + ptu_config_error, + ptu_identity_error, ptu_terms, zeroed_ptu_pricing, ) @@ -294,3 +295,35 @@ def test_an_empty_id_is_no_id(): assert error is not None assert error.startswith("model_info.id is required") + + +def test_the_spillover_header_marks_the_request_as_pay_as_you_go(): + with patch.dict(os.environ, {"LITELLM_ENABLE_PTU_COST_ATTRIBUTION": "True"}, clear=False): + assert ( + is_spilled_over_ptu_request( + model_info=_VALID, + response_headers={"x-ms-is-spilled-over": "True"}, + additional_headers=None, + ) + is True + ) + + +def test_no_spillover_marker_keeps_the_zeroed_ptu_rates(): + with patch.dict(os.environ, {"LITELLM_ENABLE_PTU_COST_ATTRIBUTION": "True"}, clear=False): + assert ( + is_spilled_over_ptu_request( + model_info=_VALID, + response_headers={"x-ms-is-spilled-over": "false"}, + additional_headers=None, + ) + is False + ) + assert ( + is_spilled_over_ptu_request( + model_info=_VALID, + response_headers=None, + additional_headers={"llm_provider-x-ms-is-spilled-over": "absent"}, + ) + is False + ) diff --git a/tests/test_litellm/llms/azure/test_azure.py b/tests/test_litellm/llms/azure/test_azure.py new file mode 100644 index 00000000000..6b6832f623c --- /dev/null +++ b/tests/test_litellm/llms/azure/test_azure.py @@ -0,0 +1,54 @@ +"""Tests for litellm/llms/azure/azure.py AzureChatCompletion handler behaviour.""" + +import time +from typing import Final + +from openai import AzureOpenAI + +from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj +from litellm.llms.azure.azure import AzureChatCompletion + + +class _FakeRawResponse: + headers: Final = {"x-ms-is-spilled-over": "true"} + + def parse(self): + return iter(()) + + +class _FakeRawCompletions: + def create(self, **kwargs): + return _FakeRawResponse() + + +def test_sync_streaming_stamps_response_headers_on_the_logging_obj() -> None: + """Sync streaming must mirror async_streaming and record the provider response + headers on model_call_details, or downstream consumers (spillover-aware cost + calculation) cannot see them.""" + client = AzureOpenAI(api_key="fake", api_version="2024-02-01", azure_endpoint="https://fake.openai.azure.com") + client.chat.completions.with_raw_response = _FakeRawCompletions() + + logging_obj = LiteLLMLoggingObj( + model="azure/gpt-4o-spill-test", + messages=[{"role": "user", "content": "Hi"}], + stream=True, + call_type="completion", + start_time=time.time(), + litellm_call_id="spill-sync-1", + function_id="f", + ) + + AzureChatCompletion().streaming( + logging_obj=logging_obj, + api_base="https://fake.openai.azure.com", + api_key="fake", + api_version="2024-02-01", + dynamic_params=False, + data={"messages": [{"role": "user", "content": "Hi"}], "stream": True}, + model="gpt-4o-spill-test", + timeout=30.0, + max_retries=0, + client=client, + ) + + assert logging_obj.model_call_details["response_headers"] == {"x-ms-is-spilled-over": "true"} From 7b855bd53f50f1a070abef1701d942fae503ac74 Mon Sep 17 00:00:00 2001 From: yucheng Date: Thu, 17 Sep 2026 05:48:51 +0000 Subject: [PATCH 2/3] feat(spend-logs): record Azure spillover source deployment in spend log metadata SpendLogsMetadata gains a typed azure_spillover key so a request Azure served off pay-as-you-go capacity is visible in spend tracking, stamped from the provider response headers or the processed llm_provider- headers on the standard logging payload. The header parsing moves into a shared azure_spillover() helper that is_spilled_over_ptu_request() now wraps. Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- litellm/litellm_core_utils/ptu_pricing.py | 26 ++++++--- litellm/proxy/_types.py | 2 + .../spend_tracking/spend_tracking_utils.py | 19 ++++++- litellm/types/utils.py | 6 ++ .../litellm_core_utils/test_ptu_pricing.py | 29 ++++++++++ .../test_spend_tracking_utils.py | 55 +++++++++++++++++++ 6 files changed, 129 insertions(+), 8 deletions(-) diff --git a/litellm/litellm_core_utils/ptu_pricing.py b/litellm/litellm_core_utils/ptu_pricing.py index 2cf86c30e9c..80f7a822b96 100644 --- a/litellm/litellm_core_utils/ptu_pricing.py +++ b/litellm/litellm_core_utils/ptu_pricing.py @@ -14,10 +14,11 @@ from typing import Final from litellm.secret_managers.main import get_secret_bool from litellm.types.router import ModelInfo -from litellm.types.utils import CustomPricingLiteLLMParams, MirroredPricingParams +from litellm.types.utils import AzureSpillover, CustomPricingLiteLLMParams, MirroredPricingParams PTU_COST_ATTRIBUTION_ENV_VAR: Final = "LITELLM_ENABLE_PTU_COST_ATTRIBUTION" AZURE_SPILLOVER_HEADER: Final = "x-ms-is-spilled-over" +AZURE_SPILLOVER_FROM_HEADER: Final = "x-ms-spillover-from-deployment" def is_ptu_cost_attribution_enabled() -> bool: @@ -248,10 +249,21 @@ def is_spilled_over_ptu_request( return False if not is_ptu_cost_attribution_enabled(): return False - for headers, key in ( - (response_headers, AZURE_SPILLOVER_HEADER), - (additional_headers, f"llm_provider-{AZURE_SPILLOVER_HEADER}"), + return azure_spillover(response_headers, additional_headers) is not None + + +def azure_spillover( + response_headers: Mapping[str, object] | None, + additional_headers: Mapping[str, object] | None, +) -> AzureSpillover | None: + """The spillover Azure reports in the response headers, else None.""" + for headers, prefix in ( + (response_headers, ""), + (additional_headers, "llm_provider-"), ): - if headers is not None and str(headers.get(key)).lower() == "true": - return True - return False + if headers is None or str(headers.get(f"{prefix}{AZURE_SPILLOVER_HEADER}")).lower() != "true": + continue + return AzureSpillover( + from_deployment=str(v) if (v := headers.get(f"{prefix}{AZURE_SPILLOVER_FROM_HEADER}")) is not None else None + ) + return None diff --git a/litellm/proxy/_types.py b/litellm/proxy/_types.py index 680c63393e8..957f79d79d4 100644 --- a/litellm/proxy/_types.py +++ b/litellm/proxy/_types.py @@ -51,6 +51,7 @@ from litellm.types.router import RouterErrors, UpdateRouterConfig from litellm.types.router_weights import validate_router_settings_dict from litellm.types.secret_managers.main import KeyManagementSystem from litellm.types.utils import ( + AzureSpillover, CallTypes, CostBreakdown, EmbeddingResponse, @@ -3895,6 +3896,7 @@ class SpendLogsMetadata(TypedDict): autorouter_savings: ReadOnly[float | None] # stamped by the logging payload; None = not auto-routed litellm_gateway_injected_cache: ReadOnly[str | None] router_metadata: ReadOnly[SpendLogsRouterMetadata | None] # None = deployment not flagged internal_router_model + azure_spillover: ReadOnly[AzureSpillover | None] # None = Azure did not report spillover class SpendLogsPayload(TypedDict): diff --git a/litellm/proxy/spend_tracking/spend_tracking_utils.py b/litellm/proxy/spend_tracking/spend_tracking_utils.py index 52900c33745..f1a54841e3a 100644 --- a/litellm/proxy/spend_tracking/spend_tracking_utils.py +++ b/litellm/proxy/spend_tracking/spend_tracking_utils.py @@ -39,6 +39,7 @@ from litellm.litellm_core_utils.litellm_logging import ( is_valid_sha256_hash, request_model_access_groups_from_litellm_params, ) +from litellm.litellm_core_utils.ptu_pricing import azure_spillover from litellm.litellm_core_utils.safe_json_dumps import safe_dumps, strip_null_bytes from litellm.proxy._types import SpendLogsMetadata, SpendLogsPayload, SpendLogsRouterMetadata from litellm.proxy.route_llm_request import ProxyModelNotFoundError @@ -47,6 +48,7 @@ from litellm.proxy.utils import PrismaClient, hash_token from litellm.types.router import DeploymentTypedDict, LiteLLM_Params from litellm.types.utils import ( PROMPT_CARRYING_GUARDRAIL_FIELDS, + AzureSpillover, CallTypes, CostBreakdown, LlmProviders, @@ -133,6 +135,9 @@ def _get_router_metadata_for_spend_log( ) +_STAMPED_METADATA_KEYS: Final = frozenset(("router_metadata", "azure_spillover")) + + def _get_spend_logs_metadata( metadata: dict | None, applied_guardrails: list[str] | None = None, @@ -150,6 +155,7 @@ def _get_spend_logs_metadata( litellm_call_id: str | None = None, autorouter_savings: float | None = None, router_metadata: SpendLogsRouterMetadata | None = None, + azure_spillover: AzureSpillover | None = None, ) -> SpendLogsMetadata: if metadata is None: return SpendLogsMetadata( @@ -191,6 +197,7 @@ def _get_spend_logs_metadata( litellm_gateway_injected_cache=None, litellm_call_id=litellm_call_id, router_metadata=router_metadata, + azure_spillover=azure_spillover, ) verbose_proxy_logger.debug( "getting payload for SpendLogs, available keys in metadata: " + str(list(metadata.keys())) @@ -198,8 +205,9 @@ def _get_spend_logs_metadata( # Filter the metadata dictionary to include only the specified keys clean_metadata: Final = SpendLogsMetadata( - **{key: metadata.get(key) for key in SpendLogsMetadata.__annotations__ if key != "router_metadata"}, + **{key: metadata.get(key) for key in SpendLogsMetadata.__annotations__ if key not in _STAMPED_METADATA_KEYS}, router_metadata=router_metadata, + azure_spillover=azure_spillover, ) _raw_key: Final = clean_metadata.get("user_api_key") _trusted_hash: Final = metadata.get("user_api_key_hash") @@ -570,6 +578,15 @@ def get_logging_payload( selected_provider=custom_llm_provider, router_correlation_id=litellm_call_id, ), + azure_spillover=azure_spillover( + response_headers=kwargs.get("response_headers") + if isinstance(kwargs.get("response_headers"), Mapping) + else None, + additional_headers=standard_logging_payload["hidden_params"].get("additional_headers") + if standard_logging_payload is not None + and isinstance(standard_logging_payload.get("hidden_params"), Mapping) + else None, + ), ) special_usage_fields: Final = ["completion_tokens", "prompt_tokens", "total_tokens"] diff --git a/litellm/types/utils.py b/litellm/types/utils.py index aaa16fd2d44..eee9288b9bb 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -3075,6 +3075,12 @@ class StandardLoggingMetadata(StandardLoggingUserAPIKeyMetadata): team_id: str | None +class AzureSpillover(TypedDict): + """Spillover Azure reports in its response headers for a request it served from pay-as-you-go capacity.""" + + from_deployment: ReadOnly[str | None] + + class StandardLoggingAdditionalHeaders(TypedDict, total=False): x_ratelimit_limit_requests: int x_ratelimit_limit_tokens: int diff --git a/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py b/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py index 464c56d5132..1689da2696f 100644 --- a/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py +++ b/tests/test_litellm/litellm_core_utils/test_ptu_pricing.py @@ -12,6 +12,7 @@ from litellm.litellm_core_utils.ptu_pricing import ( PTU_ZEROED_PRICING_FIELDS, PTU_ZEROED_TABLE_FIELDS, SEARCH_CONTEXT_SIZES, + azure_spillover, is_spilled_over_ptu_request, ptu_config_error, ptu_identity_error, @@ -327,3 +328,31 @@ def test_no_spillover_marker_keeps_the_zeroed_ptu_rates(): ) is False ) + + +def test_azure_spillover_carries_the_source_deployment_from_raw_headers(): + assert azure_spillover( + response_headers={ + "x-ms-is-spilled-over": "true", + "x-ms-spillover-from-deployment": "my-ptu", + }, + additional_headers=None, + ) == {"from_deployment": "my-ptu"} + + +def test_azure_spillover_from_processed_headers_has_no_source_when_absent(): + assert azure_spillover( + response_headers=None, + additional_headers={"llm_provider-x-ms-is-spilled-over": "true"}, + ) == {"from_deployment": None} + + +def test_no_spillover_marker_returns_none(): + assert ( + azure_spillover( + response_headers={"x-ms-is-spilled-over": "false"}, + additional_headers=None, + ) + is None + ) + assert azure_spillover(response_headers=None, additional_headers=None) is None diff --git a/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py b/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py index 1072e970094..2c86de40a8d 100644 --- a/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py +++ b/tests/test_litellm/proxy/spend_tracking/test_spend_tracking_utils.py @@ -4829,3 +4829,58 @@ def test_spend_log_request_id_is_the_response_id_a_bridged_messages_caller_recei ) == "resp_01Lit6806Bridged" ) + + +def test_azure_spillover_stamped_from_response_headers(): + """Raw provider response headers on the logging kwargs mark the request as spilled.""" + kwargs: Final = { + **_routed_call_kwargs({"id": "mi-1"}), + "response_headers": { + "x-ms-is-spilled-over": "true", + "x-ms-spillover-from-deployment": "my-ptu", + }, + } + payload = get_logging_payload( + kwargs=kwargs, + response_obj=litellm.ModelResponse(id="chatcmpl-spill-raw", choices=[], usage=litellm.Usage()), + start_time=datetime.datetime.now(timezone.utc), + end_time=datetime.datetime.now(timezone.utc), + ) + metadata = json.loads(payload["metadata"]) + assert metadata["azure_spillover"] == {"from_deployment": "my-ptu"} + + +def test_azure_spillover_stamped_from_standard_logging_additional_headers(): + """Streaming requests carry the processed llm_provider- headers on the standard payload.""" + kwargs: Final = { + **_routed_call_kwargs({"id": "mi-1"}), + "standard_logging_object": { + "hidden_params": { + "additional_headers": { + "llm_provider-x-ms-is-spilled-over": "true", + "llm_provider-x-ms-spillover-from-deployment": "my-ptu", + } + }, + "metadata": {}, + "model_map_information": None, + }, + } + payload = get_logging_payload( + kwargs=kwargs, + response_obj=litellm.ModelResponse(id="chatcmpl-spill-sl", choices=[], usage=litellm.Usage()), + start_time=datetime.datetime.now(timezone.utc), + end_time=datetime.datetime.now(timezone.utc), + ) + metadata = json.loads(payload["metadata"]) + assert metadata["azure_spillover"] == {"from_deployment": "my-ptu"} + + +def test_azure_spillover_absent_without_spillover_headers(): + payload = get_logging_payload( + kwargs=_routed_call_kwargs({"id": "mi-1"}), + response_obj=litellm.ModelResponse(id="chatcmpl-no-spill", choices=[], usage=litellm.Usage()), + start_time=datetime.datetime.now(timezone.utc), + end_time=datetime.datetime.now(timezone.utc), + ) + metadata = json.loads(payload["metadata"]) + assert metadata["azure_spillover"] is None From 3406913ca0e386a40ce512da45f8ce6e5d268d68 Mon Sep 17 00:00:00 2001 From: yucheng Date: Thu, 17 Sep 2026 05:58:19 +0000 Subject: [PATCH 3/3] test(spend-logs): expect azure_spillover in spend log metadata golden Co-Authored-By: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> --- .../proxy/spend_tracking/test_spend_management_endpoints.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/tests/test_litellm/proxy/spend_tracking/test_spend_management_endpoints.py b/tests/test_litellm/proxy/spend_tracking/test_spend_management_endpoints.py index 772c5f674d5..8d15fb094d5 100644 --- a/tests/test_litellm/proxy/spend_tracking/test_spend_management_endpoints.py +++ b/tests/test_litellm/proxy/spend_tracking/test_spend_management_endpoints.py @@ -3745,7 +3745,7 @@ class TestSpendLogsPayload: "model": "gpt-4o", "user": "", "team_id": "", - "metadata": '{"applied_guardrails": [], "attempted_fallbacks": null, "original_model_group": null, "batch_models": null, "batch_successful_requests": null, "batch_failed_requests": null, "mcp_tool_call_metadata": null, "vector_store_request_metadata": null, "routing_decision": null, "internal_call_origin": null, "guardrail_information": null, "compression_savings": null, "litellm_gateway_injected_cache": null, "router_metadata": null, "usage_object": {"completion_tokens": 20, "prompt_tokens": 10, "total_tokens": 30, "completion_tokens_details": null, "prompt_tokens_details": null}, "model_map_information": {"model_map_key": "gpt-4o", "model_map_value": {"key": "gpt-4o", "max_tokens": 16384, "max_input_tokens": 128000, "max_output_tokens": 16384, "input_cost_per_token": 2.5e-06, "cache_creation_input_token_cost": null, "cache_read_input_token_cost": 1.25e-06, "input_cost_per_character": null, "input_cost_per_token_above_128k_tokens": null, "input_cost_per_token_above_200k_tokens": null, "input_cost_per_query": null, "input_cost_per_second": null, "input_cost_per_audio_token": null, "input_cost_per_token_batches": 1.25e-06, "output_cost_per_token_batches": 5e-06, "output_cost_per_token": 1e-05, "output_cost_per_audio_token": null, "output_cost_per_character": null, "output_cost_per_token_above_128k_tokens": null, "output_cost_per_character_above_128k_tokens": null, "output_cost_per_token_above_200k_tokens": null, "output_cost_per_second": null, "output_cost_per_reasoning_token": null, "output_cost_per_image": null, "output_vector_size": null, "litellm_provider": "openai", "mode": "chat", "supports_system_messages": true, "supports_response_schema": true, "supports_vision": true, "supports_function_calling": true, "supports_tool_choice": true, "supports_assistant_prefill": false, "supports_prompt_caching": true, "supports_audio_input": false, "supports_audio_output": false, "supports_pdf_input": false, "supports_embedding_image_input": false, "supports_native_streaming": null, "supports_web_search": true, "supports_reasoning": false, "search_context_cost_per_query": {"search_context_size_low": 0.03, "search_context_size_medium": 0.035, "search_context_size_high": 0.05}, "tpm": null, "rpm": null, "supported_openai_params": ["frequency_penalty", "logit_bias", "logprobs", "top_logprobs", "max_tokens", "max_completion_tokens", "modalities", "prediction", "n", "presence_penalty", "seed", "stop", "stream", "stream_options", "temperature", "top_p", "tools", "tool_choice", "function_call", "functions", "max_retries", "extra_headers", "parallel_tool_calls", "audio", "response_format", "user"]}}, "additional_usage_values": {"completion_tokens_details": null, "prompt_tokens_details": null}}', + "metadata": '{"applied_guardrails": [], "attempted_fallbacks": null, "original_model_group": null, "batch_models": null, "batch_successful_requests": null, "batch_failed_requests": null, "mcp_tool_call_metadata": null, "vector_store_request_metadata": null, "routing_decision": null, "internal_call_origin": null, "guardrail_information": null, "compression_savings": null, "litellm_gateway_injected_cache": null, "router_metadata": null, "azure_spillover": null, "usage_object": {"completion_tokens": 20, "prompt_tokens": 10, "total_tokens": 30, "completion_tokens_details": null, "prompt_tokens_details": null}, "model_map_information": {"model_map_key": "gpt-4o", "model_map_value": {"key": "gpt-4o", "max_tokens": 16384, "max_input_tokens": 128000, "max_output_tokens": 16384, "input_cost_per_token": 2.5e-06, "cache_creation_input_token_cost": null, "cache_read_input_token_cost": 1.25e-06, "input_cost_per_character": null, "input_cost_per_token_above_128k_tokens": null, "input_cost_per_token_above_200k_tokens": null, "input_cost_per_query": null, "input_cost_per_second": null, "input_cost_per_audio_token": null, "input_cost_per_token_batches": 1.25e-06, "output_cost_per_token_batches": 5e-06, "output_cost_per_token": 1e-05, "output_cost_per_audio_token": null, "output_cost_per_character": null, "output_cost_per_token_above_128k_tokens": null, "output_cost_per_character_above_128k_tokens": null, "output_cost_per_token_above_200k_tokens": null, "output_cost_per_second": null, "output_cost_per_reasoning_token": null, "output_cost_per_image": null, "output_vector_size": null, "litellm_provider": "openai", "mode": "chat", "supports_system_messages": true, "supports_response_schema": true, "supports_vision": true, "supports_function_calling": true, "supports_tool_choice": true, "supports_assistant_prefill": false, "supports_prompt_caching": true, "supports_audio_input": false, "supports_audio_output": false, "supports_pdf_input": false, "supports_embedding_image_input": false, "supports_native_streaming": null, "supports_web_search": true, "supports_reasoning": false, "search_context_cost_per_query": {"search_context_size_low": 0.03, "search_context_size_medium": 0.035, "search_context_size_high": 0.05}, "tpm": null, "rpm": null, "supported_openai_params": ["frequency_penalty", "logit_bias", "logprobs", "top_logprobs", "max_tokens", "max_completion_tokens", "modalities", "prediction", "n", "presence_penalty", "seed", "stop", "stream", "stream_options", "temperature", "top_p", "tools", "tool_choice", "function_call", "functions", "max_retries", "extra_headers", "parallel_tool_calls", "audio", "response_format", "user"]}}, "additional_usage_values": {"completion_tokens_details": null, "prompt_tokens_details": null}}', "cache_key": "Cache OFF", "spend": 0.00022500000000000002, "total_tokens": 30, @@ -3841,7 +3841,7 @@ class TestSpendLogsPayload: "model": "claude-4-sonnet-20250514", "user": "", "team_id": "", - "metadata": '{"applied_guardrails": [], "attempted_fallbacks": null, "original_model_group": null, "batch_models": null, "batch_successful_requests": null, "batch_failed_requests": null, "mcp_tool_call_metadata": null, "vector_store_request_metadata": null, "routing_decision": null, "internal_call_origin": null, "guardrail_information": null, "compression_savings": null, "litellm_gateway_injected_cache": null, "router_metadata": null, "usage_object": {"completion_tokens": 503, "prompt_tokens": 2095, "total_tokens": 2598, "completion_tokens_details": null, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}, "model_map_information": {"model_map_key": "claude-4-sonnet-20250514", "model_map_value": {"key": "claude-4-sonnet-20250514", "max_tokens": 128000, "max_input_tokens": 200000, "max_output_tokens": 128000, "input_cost_per_token": 3e-06, "cache_creation_input_token_cost": 3.75e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_character": null, "input_cost_per_token_above_128k_tokens": null, "input_cost_per_token_above_200k_tokens": null, "input_cost_per_query": null, "input_cost_per_second": null, "input_cost_per_audio_token": null, "input_cost_per_token_batches": null, "output_cost_per_token_batches": null, "output_cost_per_token": 1.5e-05, "output_cost_per_audio_token": null, "output_cost_per_character": null, "output_cost_per_token_above_128k_tokens": null, "output_cost_per_character_above_128k_tokens": null, "output_cost_per_token_above_200k_tokens": null, "output_cost_per_second": null, "output_cost_per_image": null, "output_vector_size": null, "litellm_provider": "anthropic", "mode": "chat", "supports_system_messages": null, "supports_response_schema": true, "supports_vision": true, "supports_function_calling": true, "supports_tool_choice": true, "supports_assistant_prefill": true, "supports_prompt_caching": true, "supports_audio_input": false, "supports_audio_output": false, "supports_pdf_input": true, "supports_embedding_image_input": false, "supports_native_streaming": null, "supports_web_search": false, "supports_reasoning": true, "search_context_cost_per_query": null, "tpm": null, "rpm": null, "supported_openai_params": ["stream", "stop", "temperature", "top_p", "max_tokens", "max_completion_tokens", "tools", "tool_choice", "extra_headers", "parallel_tool_calls", "response_format", "user", "reasoning_effort", "thinking"]}}, "additional_usage_values": {"completion_tokens_details": {"accepted_prediction_tokens": null, "audio_tokens": null, "reasoning_tokens": null, "rejected_prediction_tokens": null, "text_tokens": 503, "image_tokens": null}, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0, "text_tokens": null, "image_tokens": null}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}}', + "metadata": '{"applied_guardrails": [], "attempted_fallbacks": null, "original_model_group": null, "batch_models": null, "batch_successful_requests": null, "batch_failed_requests": null, "mcp_tool_call_metadata": null, "vector_store_request_metadata": null, "routing_decision": null, "internal_call_origin": null, "guardrail_information": null, "compression_savings": null, "litellm_gateway_injected_cache": null, "router_metadata": null, "azure_spillover": null, "usage_object": {"completion_tokens": 503, "prompt_tokens": 2095, "total_tokens": 2598, "completion_tokens_details": null, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}, "model_map_information": {"model_map_key": "claude-4-sonnet-20250514", "model_map_value": {"key": "claude-4-sonnet-20250514", "max_tokens": 128000, "max_input_tokens": 200000, "max_output_tokens": 128000, "input_cost_per_token": 3e-06, "cache_creation_input_token_cost": 3.75e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_character": null, "input_cost_per_token_above_128k_tokens": null, "input_cost_per_token_above_200k_tokens": null, "input_cost_per_query": null, "input_cost_per_second": null, "input_cost_per_audio_token": null, "input_cost_per_token_batches": null, "output_cost_per_token_batches": null, "output_cost_per_token": 1.5e-05, "output_cost_per_audio_token": null, "output_cost_per_character": null, "output_cost_per_token_above_128k_tokens": null, "output_cost_per_character_above_128k_tokens": null, "output_cost_per_token_above_200k_tokens": null, "output_cost_per_second": null, "output_cost_per_image": null, "output_vector_size": null, "litellm_provider": "anthropic", "mode": "chat", "supports_system_messages": null, "supports_response_schema": true, "supports_vision": true, "supports_function_calling": true, "supports_tool_choice": true, "supports_assistant_prefill": true, "supports_prompt_caching": true, "supports_audio_input": false, "supports_audio_output": false, "supports_pdf_input": true, "supports_embedding_image_input": false, "supports_native_streaming": null, "supports_web_search": false, "supports_reasoning": true, "search_context_cost_per_query": null, "tpm": null, "rpm": null, "supported_openai_params": ["stream", "stop", "temperature", "top_p", "max_tokens", "max_completion_tokens", "tools", "tool_choice", "extra_headers", "parallel_tool_calls", "response_format", "user", "reasoning_effort", "thinking"]}}, "additional_usage_values": {"completion_tokens_details": {"accepted_prediction_tokens": null, "audio_tokens": null, "reasoning_tokens": null, "rejected_prediction_tokens": null, "text_tokens": 503, "image_tokens": null}, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0, "text_tokens": null, "image_tokens": null}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}}', "cache_key": "Cache OFF", "spend": 0.01383, "total_tokens": 2598, @@ -3935,7 +3935,7 @@ class TestSpendLogsPayload: "model": "claude-4-sonnet-20250514", "user": "", "team_id": "", - "metadata": '{"applied_guardrails": [], "attempted_fallbacks": 0, "original_model_group": "my-anthropic-model-group", "batch_models": null, "batch_successful_requests": null, "batch_failed_requests": null, "mcp_tool_call_metadata": null, "vector_store_request_metadata": null, "routing_decision": null, "internal_call_origin": null, "guardrail_information": null, "compression_savings": null, "litellm_gateway_injected_cache": null, "router_metadata": null, "usage_object": {"completion_tokens": 503, "prompt_tokens": 2095, "total_tokens": 2598, "completion_tokens_details": null, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}, "model_map_information": {"model_map_key": "claude-4-sonnet-20250514", "model_map_value": {"key": "claude-4-sonnet-20250514", "max_tokens": 128000, "max_input_tokens": 200000, "max_output_tokens": 128000, "input_cost_per_token": 3e-06, "cache_creation_input_token_cost": 3.75e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_character": null, "input_cost_per_token_above_128k_tokens": null, "input_cost_per_token_above_200k_tokens": null, "input_cost_per_query": null, "input_cost_per_second": null, "input_cost_per_audio_token": null, "input_cost_per_token_batches": null, "output_cost_per_token_batches": null, "output_cost_per_token": 1.5e-05, "output_cost_per_audio_token": null, "output_cost_per_character": null, "output_cost_per_token_above_128k_tokens": null, "output_cost_per_character_above_128k_tokens": null, "output_cost_per_token_above_200k_tokens": null, "output_cost_per_second": null, "output_cost_per_image": null, "output_vector_size": null, "litellm_provider": "anthropic", "mode": "chat", "supports_system_messages": null, "supports_response_schema": true, "supports_vision": true, "supports_function_calling": true, "supports_tool_choice": true, "supports_assistant_prefill": true, "supports_prompt_caching": true, "supports_audio_input": false, "supports_audio_output": false, "supports_pdf_input": true, "supports_embedding_image_input": false, "supports_native_streaming": null, "supports_web_search": false, "supports_reasoning": true, "search_context_cost_per_query": null, "tpm": null, "rpm": null, "supported_openai_params": ["stream", "stop", "temperature", "top_p", "max_tokens", "max_completion_tokens", "tools", "tool_choice", "extra_headers", "parallel_tool_calls", "response_format", "user", "reasoning_effort", "thinking"]}}, "additional_usage_values": {"completion_tokens_details": {"accepted_prediction_tokens": null, "audio_tokens": null, "reasoning_tokens": null, "rejected_prediction_tokens": null, "text_tokens": 503, "image_tokens": null}, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0, "text_tokens": null, "image_tokens": null}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}}', + "metadata": '{"applied_guardrails": [], "attempted_fallbacks": 0, "original_model_group": "my-anthropic-model-group", "batch_models": null, "batch_successful_requests": null, "batch_failed_requests": null, "mcp_tool_call_metadata": null, "vector_store_request_metadata": null, "routing_decision": null, "internal_call_origin": null, "guardrail_information": null, "compression_savings": null, "litellm_gateway_injected_cache": null, "router_metadata": null, "azure_spillover": null, "usage_object": {"completion_tokens": 503, "prompt_tokens": 2095, "total_tokens": 2598, "completion_tokens_details": null, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}, "model_map_information": {"model_map_key": "claude-4-sonnet-20250514", "model_map_value": {"key": "claude-4-sonnet-20250514", "max_tokens": 128000, "max_input_tokens": 200000, "max_output_tokens": 128000, "input_cost_per_token": 3e-06, "cache_creation_input_token_cost": 3.75e-06, "cache_read_input_token_cost": 3e-07, "input_cost_per_character": null, "input_cost_per_token_above_128k_tokens": null, "input_cost_per_token_above_200k_tokens": null, "input_cost_per_query": null, "input_cost_per_second": null, "input_cost_per_audio_token": null, "input_cost_per_token_batches": null, "output_cost_per_token_batches": null, "output_cost_per_token": 1.5e-05, "output_cost_per_audio_token": null, "output_cost_per_character": null, "output_cost_per_token_above_128k_tokens": null, "output_cost_per_character_above_128k_tokens": null, "output_cost_per_token_above_200k_tokens": null, "output_cost_per_second": null, "output_cost_per_image": null, "output_vector_size": null, "litellm_provider": "anthropic", "mode": "chat", "supports_system_messages": null, "supports_response_schema": true, "supports_vision": true, "supports_function_calling": true, "supports_tool_choice": true, "supports_assistant_prefill": true, "supports_prompt_caching": true, "supports_audio_input": false, "supports_audio_output": false, "supports_pdf_input": true, "supports_embedding_image_input": false, "supports_native_streaming": null, "supports_web_search": false, "supports_reasoning": true, "search_context_cost_per_query": null, "tpm": null, "rpm": null, "supported_openai_params": ["stream", "stop", "temperature", "top_p", "max_tokens", "max_completion_tokens", "tools", "tool_choice", "extra_headers", "parallel_tool_calls", "response_format", "user", "reasoning_effort", "thinking"]}}, "additional_usage_values": {"completion_tokens_details": {"accepted_prediction_tokens": null, "audio_tokens": null, "reasoning_tokens": null, "rejected_prediction_tokens": null, "text_tokens": 503, "image_tokens": null}, "prompt_tokens_details": {"audio_tokens": null, "cached_tokens": 0, "text_tokens": null, "image_tokens": null}, "cache_creation_input_tokens": 0, "cache_read_input_tokens": 0}}', "cache_key": "Cache OFF", "spend": 0.01383, "total_tokens": 2598,