diff --git a/.github/workflows/test-unit.yml b/.github/workflows/test-unit.yml index faa02c8f1a9..18df865c988 100644 --- a/.github/workflows/test-unit.yml +++ b/.github/workflows/test-unit.yml @@ -140,6 +140,7 @@ jobs: tests/unit/batch_completion tests/unit/decisions tests/unit/litellm_core_utils + tests/unit/systemone python-version: "3.12" workers: 4 reruns: 0 diff --git a/litellm/__init__.py b/litellm/__init__.py index 554b37f11d4..a611b90bc37 100644 --- a/litellm/__init__.py +++ b/litellm/__init__.py @@ -1490,6 +1490,7 @@ from .rust_bridge import rust from .rag.main import * from .sandbox.main import * from .decisions.main import * +from .systemone.main import * from .tool_loop import ToolLoopMaxRoundsExceeded, arun_tool_loop, run_tool_loop from .search.main import * from .realtime_api.main import ( diff --git a/litellm/cost_calculator.py b/litellm/cost_calculator.py index da420ab0578..45d98043144 100644 --- a/litellm/cost_calculator.py +++ b/litellm/cost_calculator.py @@ -2026,6 +2026,8 @@ def response_cost_calculator( "asearch", "decisions", "adecisions", + "systemone", + "asystemone", ], optional_params: dict, cache_hit: bool | None = None, diff --git a/litellm/decisions/call.py b/litellm/decisions/call.py new file mode 100644 index 00000000000..1ad2f006736 --- /dev/null +++ b/litellm/decisions/call.py @@ -0,0 +1,347 @@ +from collections.abc import Mapping, Sequence +from dataclasses import dataclass, field, replace +from typing import Final, TypeAlias + +import httpx +from pydantic import ConfigDict, TypeAdapter, ValidationError +from typing_extensions import assert_never + +import litellm +from litellm.litellm_core_utils.core_helpers import RESPONSE_COST_HEADER, normalize_drop_params +from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj +from litellm.llms.base_llm.chat.transformation import BaseLLMException +from litellm.llms.base_llm.decisions.transformation import ( + BaseDecisionsConfig, + ir_to_systemone_response, + systemone_request_to_ir, +) +from litellm.llms.custom_httpx.llm_http_handler import BaseLLMHTTPHandler +from litellm.llms.openai.decisions.transformation import ir_to_openai_response, openai_request_to_ir +from litellm.types.decisions import ( + DecisionQuestion, + DecisionsIRRequest, + DecisionsIRResponse, + DecisionsJSON, + DecisionsRequestBody, + DecisionsResponse, + OpenAIDecisionInput, + OpenAIDecisionQuestion, + OpenAIDecisionRequestBody, + OpenAIDecisionResponse, + UnsupportedDecisionsRequest, +) +from litellm.types.utils import LlmProviders +from litellm.utils import ProviderConfigManager + +SystemOneQuestions: TypeAlias = Mapping[str, DecisionQuestion | Mapping[str, object]] +OpenAIDecisionQuestions: TypeAlias = Sequence[OpenAIDecisionQuestion | Mapping[str, object]] +DecisionsRequestFormat: TypeAlias = DecisionsRequestBody | OpenAIDecisionRequestBody + +_SYSTEMONE_REQUEST_ADAPTER: Final[TypeAdapter[DecisionsRequestBody]] = TypeAdapter(DecisionsRequestBody) +_OPENAI_REQUEST_ADAPTER: Final[TypeAdapter[OpenAIDecisionRequestBody]] = TypeAdapter(OpenAIDecisionRequestBody) +_SAFETY_IDENTIFIER_ADAPTER: Final[TypeAdapter[str | None]] = TypeAdapter( + str | None, config=ConfigDict(title="safety_identifier") +) +_HANDLER: Final = BaseLLMHTTPHandler() + + +@dataclass(frozen=True, slots=True) +class SystemOneRequestFields: + state: DecisionsJSON | None + questions: SystemOneQuestions | None + + def validate(self, safety_identifier: str | None) -> DecisionsRequestBody: + return _SYSTEMONE_REQUEST_ADAPTER.validate_python({"state": self.state, "questions": self.questions}) + + +@dataclass(frozen=True, slots=True) +class OpenAIDecisionRequestFields: + input: OpenAIDecisionInput | None + questions: OpenAIDecisionQuestions | None + + def validate(self, safety_identifier: str | None) -> OpenAIDecisionRequestBody: + return _OPENAI_REQUEST_ADAPTER.validate_python( + {"input": self.input, "questions": self.questions, "safety_identifier": safety_identifier} + ) + + +RequestFields: TypeAlias = SystemOneRequestFields | OpenAIDecisionRequestFields + + +@dataclass(frozen=True, slots=True, repr=False) +class DecisionsCall: + model: str + requested_model: str + custom_llm_provider: str + provider_config: BaseDecisionsConfig + request: DecisionsRequestFormat + ir_request: DecisionsIRRequest + body: Mapping[str, object] = field(repr=False) + api_base: str + api_key: str | None = field(repr=False) + logging_obj: LiteLLMLoggingObj | None + headers: Mapping[str, str] + timeout: float | httpx.Timeout | None + litellm_params: Mapping[str, object] = field(repr=False) + + +def supported_decisions_providers() -> tuple[str, ...]: + return tuple( + provider.value + for provider in LlmProviders + if ProviderConfigManager.get_provider_decisions_config(model="", provider=provider) is not None + ) + + +def _provider_config(model: str, custom_llm_provider: str) -> BaseDecisionsConfig: + provider: Final = next((member for member in LlmProviders if member.value == custom_llm_provider), None) + provider_config: Final = ( + None if provider is None else ProviderConfigManager.get_provider_decisions_config(model, provider) + ) + if provider_config is None: + supported: Final = ", ".join(supported_decisions_providers()) + raise litellm.BadRequestError( + message=f"Unknown Decisions provider '{custom_llm_provider}'. Supported providers: {supported}", + model=model, + llm_provider=custom_llm_provider, + ) + return provider_config + + +def _ir_request(request: DecisionsRequestFormat, safety_identifier: str | None) -> DecisionsIRRequest: + match request: + case DecisionsRequestBody(): + return replace(systemone_request_to_ir(request), safety_identifier=safety_identifier) + case OpenAIDecisionRequestBody(): + return openai_request_to_ir(request) + case _: + assert_never(request) + + +def _drops_params(kwargs: Mapping[str, object]) -> bool: + return litellm.drop_params is True or normalize_drop_params(kwargs.get("drop_params")) is True + + +def _request_safety_identifier(safety_identifier: object, kwargs: Mapping[str, object]) -> str | None: + if isinstance(safety_identifier, str) or not _drops_params(kwargs): + return _SAFETY_IDENTIFIER_ADAPTER.validate_python(safety_identifier) + return None + + +@dataclass(frozen=True, slots=True) +class _UnsupportedSafetyIdentifier: + pass + + +def _provider_ir_request( + request: DecisionsRequestFormat, + *, + safety_identifier: str | None, + provider_config: BaseDecisionsConfig, + kwargs: Mapping[str, object], +) -> DecisionsIRRequest | _UnsupportedSafetyIdentifier: + ir_request: Final = _ir_request(request, safety_identifier) + if ir_request.safety_identifier is None or provider_config.supports_safety_identifier: + return ir_request + if _drops_params(kwargs): + return replace(ir_request, safety_identifier=None) + return _UnsupportedSafetyIdentifier() + + +def prepare_call( + *, + model: str, + fields: RequestFields, + safety_identifier: str | None, + api_key: str | None, + api_base: str | None, + timeout: float | httpx.Timeout | None, + custom_llm_provider: str | None, + extra_headers: Mapping[str, str] | None, + kwargs: Mapping[str, object], +) -> DecisionsCall: + upstream_model, provider, dynamic_api_key, dynamic_api_base = litellm.get_llm_provider( + model=model, + custom_llm_provider=custom_llm_provider, + api_base=api_base, + api_key=api_key, + ) + provider_config: Final = _provider_config(upstream_model, provider) + if not upstream_model: + raise litellm.BadRequestError( + message="A model name is required for the Decisions API", + model=model, + llm_provider=provider, + ) + try: + canonical_model: Final = provider_config.canonical_model(upstream_model) + except ValueError as error: + raise litellm.BadRequestError(message=str(error), model=model, llm_provider=provider) from error + try: + request_safety_identifier: Final = _request_safety_identifier(safety_identifier, kwargs) + request: Final = fields.validate(request_safety_identifier) + except ValidationError as error: + raise litellm.BadRequestError( + message=f"Invalid Decisions request: {error}", + model=model, + llm_provider=provider, + ) from error + + resolved_api_base: Final = provider_config.resolve_api_base(dynamic_api_base or api_base) + if resolved_api_base is None: + raise litellm.BadRequestError( + message=provider_config.missing_api_base_message(provider), + model=model, + llm_provider=provider, + ) + resolved_api_key: Final = provider_config.resolve_api_key(dynamic_api_key or api_key) + if resolved_api_key is None and provider_config.api_key_required: + raise litellm.AuthenticationError( + message=f"Missing API key for Decisions provider '{provider}'", + model=model, + llm_provider=provider, + ) + + ir_request: Final = _provider_ir_request( + request, + safety_identifier=request_safety_identifier, + provider_config=provider_config, + kwargs=kwargs, + ) + if isinstance(ir_request, _UnsupportedSafetyIdentifier): + raise litellm.UnsupportedParamsError( + message=( + f"{provider} does not support parameters: ['safety_identifier'], for model={model}. " + "To drop these, set `litellm.drop_params=True` or for proxy:\n\n" + "`litellm_settings:\n drop_params: true`\n" + ), + model=model, + llm_provider=provider, + ) + body: Final = provider_config.transform_decisions_request( + model=canonical_model, request=ir_request, custom_llm_provider=provider + ) + if isinstance(body, UnsupportedDecisionsRequest): + raise litellm.BadRequestError( + message=f"Decisions provider '{provider}' cannot serve this request: {body.reason}", + model=model, + llm_provider=provider, + ) + + logging_obj: Final = kwargs.get("litellm_logging_obj") + if isinstance(logging_obj, LiteLLMLoggingObj): + logging_obj.update_from_kwargs( + kwargs=dict(kwargs), + model=canonical_model, + litellm_params={ + "litellm_call_id": kwargs.get("litellm_call_id"), + "api_base": provider_config.get_complete_url(resolved_api_base, canonical_model), + }, + custom_llm_provider=provider, + ) + return DecisionsCall( + model=canonical_model, + requested_model=model, + custom_llm_provider=provider, + provider_config=provider_config, + request=request, + ir_request=ir_request, + body=body, + api_base=resolved_api_base, + api_key=resolved_api_key, + logging_obj=logging_obj if isinstance(logging_obj, LiteLLMLoggingObj) else None, + headers=extra_headers or {}, + timeout=timeout, + litellm_params=kwargs, + ) + + +def _format_response(response: DecisionsIRResponse, call: DecisionsCall) -> DecisionsResponse | OpenAIDecisionResponse: + formatted: Final = _formatted_response(response, call) + provider_cost: Final = call.provider_config.provider_reported_cost(response) + formatted.set_hidden_params( + { + "model": f"{call.custom_llm_provider}/{call.model}", + "custom_llm_provider": call.custom_llm_provider, + "provider_response_model": f"{call.custom_llm_provider}/{call.model}", + **({"additional_headers": {RESPONSE_COST_HEADER: provider_cost}} if provider_cost is not None else {}), + } + ) + return formatted + + +def _formatted_response( + response: DecisionsIRResponse, call: DecisionsCall +) -> DecisionsResponse | OpenAIDecisionResponse: + match call.request: + case DecisionsRequestBody(): + return ir_to_systemone_response(response, call.ir_request) + case OpenAIDecisionRequestBody(): + return ir_to_openai_response(response, call.ir_request, call.requested_model) + case _: + assert_never(call.request) + + +def _map_upstream_exception(error: Exception, call: DecisionsCall) -> Exception: + if isinstance(error, BaseLLMException) and error.status_code_is_synthesized: + provider_label: Final = f"{call.custom_llm_provider[0].upper()}{call.custom_llm_provider[1:]}Exception" + return litellm.APIConnectionError( + message=f"{provider_label} - {error.message}", + llm_provider=call.custom_llm_provider, + model=f"{call.custom_llm_provider}/{call.model}", + ) + return litellm.exception_type( + model=f"{call.custom_llm_provider}/{call.model}", + custom_llm_provider=call.custom_llm_provider, + original_exception=error, + ) + + +async def asend(call: DecisionsCall) -> DecisionsResponse | OpenAIDecisionResponse: + try: + response: Final = await _HANDLER.adecisions( + model=call.model, + custom_llm_provider=call.custom_llm_provider, + logging_obj=call.logging_obj, + provider_config=call.provider_config, + request=call.ir_request, + body=call.body, + api_base=call.api_base, + api_key=call.api_key, + headers=call.headers, + timeout=call.timeout, + litellm_params=call.litellm_params, + ) + except Exception as error: + raise _map_upstream_exception(error, call) from error + return _format_response(response, call) + + +def send(call: DecisionsCall) -> DecisionsResponse | OpenAIDecisionResponse: + try: + response: Final = _HANDLER.decisions( + model=call.model, + custom_llm_provider=call.custom_llm_provider, + logging_obj=call.logging_obj, + provider_config=call.provider_config, + request=call.ir_request, + body=call.body, + api_base=call.api_base, + api_key=call.api_key, + headers=call.headers, + timeout=call.timeout, + litellm_params=call.litellm_params, + ) + except Exception as error: + raise _map_upstream_exception(error, call) from error + return _format_response(response, call) + + +def reject_other_format(*, model: str, kwargs: Mapping[str, object], field_name: str, other_function: str) -> None: + if field_name not in kwargs: + return + raise litellm.BadRequestError( + message=f"Requests with '{field_name}' go through litellm.{other_function}", + model=model, + llm_provider="", + ) diff --git a/litellm/decisions/main.py b/litellm/decisions/main.py index 5fd5a133eb2..8b13d04770a 100644 --- a/litellm/decisions/main.py +++ b/litellm/decisions/main.py @@ -1,151 +1,56 @@ -from collections.abc import Mapping, Sequence -from dataclasses import dataclass, field, replace -from typing import Final, TypeAlias +import warnings +from collections.abc import Mapping +from typing import Final import httpx -from pydantic import ConfigDict, TypeAdapter, ValidationError -from typing_extensions import assert_never +from pydantic import TypeAdapter -import litellm -from litellm.litellm_core_utils.core_helpers import RESPONSE_COST_HEADER, normalize_drop_params -from litellm.litellm_core_utils.litellm_logging import Logging as LiteLLMLoggingObj -from litellm.llms.base_llm.chat.transformation import BaseLLMException -from litellm.llms.base_llm.decisions.transformation import ( - BaseDecisionsConfig, - ir_to_systemone_response, - systemone_request_to_ir, +from litellm.decisions.call import ( + DecisionsCall, + OpenAIDecisionQuestions, + OpenAIDecisionRequestFields, + RequestFields, + SystemOneQuestions, + SystemOneRequestFields, + asend, + prepare_call, + send, ) -from litellm.llms.custom_httpx.llm_http_handler import BaseLLMHTTPHandler -from litellm.llms.openai.decisions.transformation import ir_to_openai_response, openai_request_to_ir -from litellm.types.decisions import ( - DecisionQuestion, - DecisionsIRRequest, - DecisionsIRResponse, - DecisionsJSON, - DecisionsRequestBody, - DecisionsResponse, - OpenAIDecisionInput, - OpenAIDecisionQuestion, - OpenAIDecisionRequestBody, - OpenAIDecisionResponse, - UnsupportedDecisionsRequest, +from litellm.types.decisions import DecisionsJSON, DecisionsResponse, OpenAIDecisionInput, OpenAIDecisionResponse +from litellm.utils import client + +_STATE_ADAPTER: Final[TypeAdapter[DecisionsJSON | None]] = TypeAdapter(DecisionsJSON | None) +_SYSTEMONE_QUESTIONS_ADAPTER: Final[TypeAdapter[SystemOneQuestions | None]] = TypeAdapter(SystemOneQuestions | None) +_OPENAI_QUESTIONS_ADAPTER: Final[TypeAdapter[OpenAIDecisionQuestions | None]] = TypeAdapter( + OpenAIDecisionQuestions | None ) -from litellm.types.utils import LlmProviders -from litellm.utils import ProviderConfigManager, client - -DecisionsQuestions: TypeAlias = ( - Mapping[str, DecisionQuestion | Mapping[str, object]] | Sequence[OpenAIDecisionQuestion | Mapping[str, object]] -) -DecisionsRequestFormat: TypeAlias = DecisionsRequestBody | OpenAIDecisionRequestBody - -_SYSTEMONE_REQUEST_ADAPTER: Final[TypeAdapter[DecisionsRequestBody]] = TypeAdapter(DecisionsRequestBody) -_OPENAI_REQUEST_ADAPTER: Final[TypeAdapter[OpenAIDecisionRequestBody]] = TypeAdapter(OpenAIDecisionRequestBody) -_SAFETY_IDENTIFIER_ADAPTER: Final[TypeAdapter[str | None]] = TypeAdapter( - str | None, config=ConfigDict(title="safety_identifier") -) -_HANDLER: Final = BaseLLMHTTPHandler() -@dataclass(frozen=True, slots=True, repr=False) -class _DecisionsCall: - model: str - requested_model: str - custom_llm_provider: str - provider_config: BaseDecisionsConfig - request: DecisionsRequestFormat - ir_request: DecisionsIRRequest - body: Mapping[str, object] = field(repr=False) - api_base: str - api_key: str | None = field(repr=False) - logging_obj: LiteLLMLoggingObj | None - headers: Mapping[str, str] - timeout: float | httpx.Timeout | None - litellm_params: Mapping[str, object] = field(repr=False) - - -def supported_decisions_providers() -> tuple[str, ...]: - return tuple( - provider.value - for provider in LlmProviders - if ProviderConfigManager.get_provider_decisions_config(model="", provider=provider) is not None - ) - - -def _provider_config(model: str, custom_llm_provider: str) -> BaseDecisionsConfig: - provider: Final = next((member for member in LlmProviders if member.value == custom_llm_provider), None) - provider_config: Final = ( - None if provider is None else ProviderConfigManager.get_provider_decisions_config(model, provider) - ) - if provider_config is None: - supported: Final = ", ".join(supported_decisions_providers()) - raise litellm.BadRequestError( - message=f"Unknown Decisions provider '{custom_llm_provider}'. Supported providers: {supported}", - model=model, - llm_provider=custom_llm_provider, - ) - return provider_config - - -def _validate_request( - *, - state: DecisionsJSON | None, - questions: DecisionsQuestions | None, +def _request_fields( decision_input: OpenAIDecisionInput | None, - safety_identifier: str | None, -) -> DecisionsRequestFormat: - if decision_input is None: - return _SYSTEMONE_REQUEST_ADAPTER.validate_python({"state": state, "questions": questions}) - return _OPENAI_REQUEST_ADAPTER.validate_python( - {"input": decision_input, "questions": questions, "safety_identifier": safety_identifier} + questions: OpenAIDecisionQuestions | SystemOneQuestions | None, + kwargs: Mapping[str, object], +) -> RequestFields: + if "state" not in kwargs: + return OpenAIDecisionRequestFields( + input=decision_input, questions=_OPENAI_QUESTIONS_ADAPTER.validate_python(questions) + ) + warnings.warn( + "litellm.decisions(state=...) is deprecated, call litellm.systemone for the System One request format", + UserWarning, + stacklevel=5, + ) + return SystemOneRequestFields( + state=_STATE_ADAPTER.validate_python(kwargs["state"]), + questions=_SYSTEMONE_QUESTIONS_ADAPTER.validate_python(questions), ) -def _ir_request(request: DecisionsRequestFormat, safety_identifier: str | None) -> DecisionsIRRequest: - match request: - case DecisionsRequestBody(): - return replace(systemone_request_to_ir(request), safety_identifier=safety_identifier) - case OpenAIDecisionRequestBody(): - return openai_request_to_ir(request) - case _: - assert_never(request) - - -def _drops_params(kwargs: Mapping[str, object]) -> bool: - return litellm.drop_params is True or normalize_drop_params(kwargs.get("drop_params")) is True - - -def _request_safety_identifier(safety_identifier: object, kwargs: Mapping[str, object]) -> str | None: - if isinstance(safety_identifier, str) or not _drops_params(kwargs): - return _SAFETY_IDENTIFIER_ADAPTER.validate_python(safety_identifier) - return None - - -@dataclass(frozen=True, slots=True) -class _UnsupportedSafetyIdentifier: - pass - - -def _provider_ir_request( - request: DecisionsRequestFormat, - *, - safety_identifier: str | None, - provider_config: BaseDecisionsConfig, - kwargs: Mapping[str, object], -) -> DecisionsIRRequest | _UnsupportedSafetyIdentifier: - ir_request: Final = _ir_request(request, safety_identifier) - if ir_request.safety_identifier is None or provider_config.supports_safety_identifier: - return ir_request - if _drops_params(kwargs): - return replace(ir_request, safety_identifier=None) - return _UnsupportedSafetyIdentifier() - - -def _prepare_call( +def _prepare_decisions_call( *, model: str, - state: DecisionsJSON | None, - questions: DecisionsQuestions | None, decision_input: OpenAIDecisionInput | None, + questions: OpenAIDecisionQuestions | SystemOneQuestions | None, safety_identifier: str | None, api_key: str | None, api_base: str | None, @@ -153,174 +58,37 @@ def _prepare_call( custom_llm_provider: str | None, extra_headers: Mapping[str, str] | None, kwargs: Mapping[str, object], -) -> _DecisionsCall: - upstream_model, provider, dynamic_api_key, dynamic_api_base = litellm.get_llm_provider( +) -> DecisionsCall: + return prepare_call( model=model, - custom_llm_provider=custom_llm_provider, - api_base=api_base, + fields=_request_fields(decision_input, questions, kwargs), + safety_identifier=safety_identifier, api_key=api_key, - ) - provider_config: Final = _provider_config(upstream_model, provider) - if not upstream_model: - raise litellm.BadRequestError( - message="A model name is required for the Decisions API", - model=model, - llm_provider=provider, - ) - if state is not None and decision_input is not None: - raise litellm.BadRequestError( - message="Pass either state (System One format) or input (OpenAI format) to the Decisions API, not both", - model=model, - llm_provider=provider, - ) - try: - canonical_model: Final = provider_config.canonical_model(upstream_model) - except ValueError as error: - raise litellm.BadRequestError(message=str(error), model=model, llm_provider=provider) from error - try: - request_safety_identifier: Final = _request_safety_identifier(safety_identifier, kwargs) - request: Final = _validate_request( - state=state, - questions=questions, - decision_input=decision_input, - safety_identifier=request_safety_identifier, - ) - except ValidationError as error: - raise litellm.BadRequestError( - message=f"Invalid Decisions request: {error}", - model=model, - llm_provider=provider, - ) from error - - resolved_api_base: Final = provider_config.resolve_api_base(dynamic_api_base or api_base) - if resolved_api_base is None: - raise litellm.BadRequestError( - message=provider_config.missing_api_base_message(provider), - model=model, - llm_provider=provider, - ) - resolved_api_key: Final = provider_config.resolve_api_key(dynamic_api_key or api_key) - if resolved_api_key is None and provider_config.api_key_required: - raise litellm.AuthenticationError( - message=f"Missing API key for Decisions provider '{provider}'", - model=model, - llm_provider=provider, - ) - - ir_request: Final = _provider_ir_request( - request, - safety_identifier=request_safety_identifier, - provider_config=provider_config, - kwargs=kwargs, - ) - if isinstance(ir_request, _UnsupportedSafetyIdentifier): - raise litellm.UnsupportedParamsError( - message=( - f"{provider} does not support parameters: ['safety_identifier'], for model={model}. " - "To drop these, set `litellm.drop_params=True` or for proxy:\n\n" - "`litellm_settings:\n drop_params: true`\n" - ), - model=model, - llm_provider=provider, - ) - body: Final = provider_config.transform_decisions_request( - model=canonical_model, request=ir_request, custom_llm_provider=provider - ) - if isinstance(body, UnsupportedDecisionsRequest): - raise litellm.BadRequestError( - message=f"Decisions provider '{provider}' cannot serve this request: {body.reason}", - model=model, - llm_provider=provider, - ) - - logging_obj: Final = kwargs.get("litellm_logging_obj") - if isinstance(logging_obj, LiteLLMLoggingObj): - logging_obj.update_from_kwargs( - kwargs=dict(kwargs), - model=canonical_model, - litellm_params={ - "litellm_call_id": kwargs.get("litellm_call_id"), - "api_base": provider_config.get_complete_url(resolved_api_base, canonical_model), - }, - custom_llm_provider=provider, - ) - return _DecisionsCall( - model=canonical_model, - requested_model=model, - custom_llm_provider=provider, - provider_config=provider_config, - request=request, - ir_request=ir_request, - body=body, - api_base=resolved_api_base, - api_key=resolved_api_key, - logging_obj=logging_obj if isinstance(logging_obj, LiteLLMLoggingObj) else None, - headers=extra_headers or {}, + api_base=api_base, timeout=timeout, - litellm_params=kwargs, - ) - - -def _format_response(response: DecisionsIRResponse, call: _DecisionsCall) -> DecisionsResponse | OpenAIDecisionResponse: - formatted: Final = _formatted_response(response, call) - provider_cost: Final = call.provider_config.provider_reported_cost(response) - formatted.set_hidden_params( - { - "model": f"{call.custom_llm_provider}/{call.model}", - "custom_llm_provider": call.custom_llm_provider, - "provider_response_model": f"{call.custom_llm_provider}/{call.model}", - **({"additional_headers": {RESPONSE_COST_HEADER: provider_cost}} if provider_cost is not None else {}), - } - ) - return formatted - - -def _formatted_response( - response: DecisionsIRResponse, call: _DecisionsCall -) -> DecisionsResponse | OpenAIDecisionResponse: - match call.request: - case DecisionsRequestBody(): - return ir_to_systemone_response(response, call.ir_request) - case OpenAIDecisionRequestBody(): - return ir_to_openai_response(response, call.ir_request, call.requested_model) - case _: - assert_never(call.request) - - -def _map_upstream_exception(error: Exception, call: _DecisionsCall) -> Exception: - if isinstance(error, BaseLLMException) and error.status_code_is_synthesized: - provider_label: Final = f"{call.custom_llm_provider[0].upper()}{call.custom_llm_provider[1:]}Exception" - return litellm.APIConnectionError( - message=f"{provider_label} - {error.message}", - llm_provider=call.custom_llm_provider, - model=f"{call.custom_llm_provider}/{call.model}", - ) - return litellm.exception_type( - model=f"{call.custom_llm_provider}/{call.model}", - custom_llm_provider=call.custom_llm_provider, - original_exception=error, + custom_llm_provider=custom_llm_provider, + extra_headers=extra_headers, + kwargs={key: value for key, value in kwargs.items() if key != "state"}, ) @client async def adecisions( model: str, - state: DecisionsJSON | None = None, - questions: DecisionsQuestions | None = None, + input: OpenAIDecisionInput | None = None, + questions: OpenAIDecisionQuestions | SystemOneQuestions | None = None, + safety_identifier: str | None = None, api_key: str | None = None, api_base: str | None = None, timeout: float | httpx.Timeout | None = None, custom_llm_provider: str | None = None, extra_headers: Mapping[str, str] | None = None, - input: OpenAIDecisionInput | None = None, - safety_identifier: str | None = None, **kwargs: object, -) -> DecisionsResponse | OpenAIDecisionResponse: - call: Final = _prepare_call( +) -> OpenAIDecisionResponse | DecisionsResponse: + call: Final = _prepare_decisions_call( model=model, - state=state, - questions=questions, decision_input=input, + questions=questions, safety_identifier=safety_identifier, api_key=api_key, api_base=api_base, @@ -329,44 +97,26 @@ async def adecisions( extra_headers=extra_headers, kwargs=kwargs, ) - try: - response: Final = await _HANDLER.adecisions( - model=call.model, - custom_llm_provider=call.custom_llm_provider, - logging_obj=call.logging_obj, - provider_config=call.provider_config, - request=call.ir_request, - body=call.body, - api_base=call.api_base, - api_key=call.api_key, - headers=call.headers, - timeout=call.timeout, - litellm_params=call.litellm_params, - ) - except Exception as error: - raise _map_upstream_exception(error, call) from error - return _format_response(response, call) + return await asend(call) @client def decisions( model: str, - state: DecisionsJSON | None = None, - questions: DecisionsQuestions | None = None, + input: OpenAIDecisionInput | None = None, + questions: OpenAIDecisionQuestions | SystemOneQuestions | None = None, + safety_identifier: str | None = None, api_key: str | None = None, api_base: str | None = None, timeout: float | httpx.Timeout | None = None, custom_llm_provider: str | None = None, extra_headers: Mapping[str, str] | None = None, - input: OpenAIDecisionInput | None = None, - safety_identifier: str | None = None, **kwargs: object, -) -> DecisionsResponse | OpenAIDecisionResponse: - call: Final = _prepare_call( +) -> OpenAIDecisionResponse | DecisionsResponse: + call: Final = _prepare_decisions_call( model=model, - state=state, - questions=questions, decision_input=input, + questions=questions, safety_identifier=safety_identifier, api_key=api_key, api_base=api_base, @@ -375,23 +125,7 @@ def decisions( extra_headers=extra_headers, kwargs=kwargs, ) - try: - response: Final = _HANDLER.decisions( - model=call.model, - custom_llm_provider=call.custom_llm_provider, - logging_obj=call.logging_obj, - provider_config=call.provider_config, - request=call.ir_request, - body=call.body, - api_base=call.api_base, - api_key=call.api_key, - headers=call.headers, - timeout=call.timeout, - litellm_params=call.litellm_params, - ) - except Exception as error: - raise _map_upstream_exception(error, call) from error - return _format_response(response, call) + return send(call) __all__ = ["adecisions", "decisions"] diff --git a/litellm/litellm_core_utils/health_check_helpers.py b/litellm/litellm_core_utils/health_check_helpers.py index a035f86ea56..3cbc2d770cc 100644 --- a/litellm/litellm_core_utils/health_check_helpers.py +++ b/litellm/litellm_core_utils/health_check_helpers.py @@ -311,7 +311,7 @@ class HealthCheckHelpers: **_filter_model_params(model_params=model_params), document=_ocr_health_check_document(model=model, custom_llm_provider=custom_llm_provider), ), - "evaluation": lambda: litellm.adecisions( + "evaluation": lambda: litellm.asystemone( **DECISIONS_CALL_PARAMS.validate_python( { "state": prompt or "health check", diff --git a/litellm/llms/vertex_ai/common_utils.py b/litellm/llms/vertex_ai/common_utils.py index 6b759065706..7c96b190480 100644 --- a/litellm/llms/vertex_ai/common_utils.py +++ b/litellm/llms/vertex_ai/common_utils.py @@ -1,6 +1,7 @@ import re from collections.abc import Mapping from copy import deepcopy +from datetime import datetime, timezone from enum import Enum from functools import lru_cache from types import MappingProxyType @@ -1000,18 +1001,18 @@ def strip_field(schema, field_name: str): strip_field(items, field_name) +_VERTEX_DATETIME_FRACTION: Final = re.compile(r"\.\d+(?=Z$)", re.IGNORECASE) + + def convert_vertex_datetime_to_openai_datetime(vertex_datetime: str) -> int: """ - Converts a Vertex AI datetime string to an OpenAI datetime integer + Converts a Vertex AI / GCS RFC 3339 UTC timestamp to Unix epoch seconds vertex_datetime: str = "2024-12-04T21:53:12.120184Z" - returns: int = 1722729192 + returns: int = 1733349192 """ - from datetime import datetime - - # Parse the ISO format string to datetime object - dt: Final = datetime.strptime(vertex_datetime, "%Y-%m-%dT%H:%M:%S.%fZ") - # Convert to Unix timestamp (seconds since epoch) + whole_seconds: Final = _VERTEX_DATETIME_FRACTION.sub("", vertex_datetime) + dt: Final = datetime.strptime(whole_seconds, "%Y-%m-%dT%H:%M:%SZ").replace(tzinfo=timezone.utc) return int(dt.timestamp()) diff --git a/litellm/proxy/common_request_processing.py b/litellm/proxy/common_request_processing.py index ed9ac258e16..6b87960fcc0 100644 --- a/litellm/proxy/common_request_processing.py +++ b/litellm/proxy/common_request_processing.py @@ -184,6 +184,7 @@ ProxyRouteType: TypeAlias = Literal[ "aocr", "asearch", "adecisions", + "asystemone", "avideo_generation", "avideo_list", "avideo_status", @@ -2074,6 +2075,7 @@ class ProxyBaseLLMRequestProcessing: "aocr", "asearch", "adecisions", + "asystemone", "avideo_generation", "avideo_list", "avideo_status", diff --git a/litellm/proxy/decisions_endpoints/endpoints.py b/litellm/proxy/decisions_endpoints/endpoints.py index c8d4768dad0..dbb7b69bd8b 100644 --- a/litellm/proxy/decisions_endpoints/endpoints.py +++ b/litellm/proxy/decisions_endpoints/endpoints.py @@ -1,6 +1,6 @@ from collections.abc import Mapping from types import MappingProxyType -from typing import Annotated, Final +from typing import Annotated, Final, Literal from fastapi import APIRouter, Depends, Request, Response from fastapi.responses import ORJSONResponse # pyright: ignore[reportDeprecated] # required endpoint contract @@ -109,6 +109,7 @@ async def _process_decisions( fastapi_response: Response, user_api_key_dict: UserAPIKeyAuth, body_adapter: TypeAdapter[DecisionsRequestBody] | TypeAdapter[OpenAIDecisionRequestBody], + route_type: Literal["adecisions", "asystemone"], ) -> object: from litellm.proxy.proxy_server import ( general_settings as proxy_general_settings, @@ -146,7 +147,7 @@ async def _process_decisions( request=request, fastapi_response=fastapi_response, user_api_key_dict=user_api_key_dict, - route_type="adecisions", + route_type=route_type, proxy_logging_obj=proxy_logging_obj, llm_router=llm_router, general_settings=general_settings, @@ -208,6 +209,7 @@ async def systemone( fastapi_response=fastapi_response, user_api_key_dict=user_api_key_dict, body_adapter=_DECISIONS_REQUEST_BODY_ADAPTER, + route_type="asystemone", ) @@ -250,4 +252,5 @@ async def decisions( fastapi_response=fastapi_response, user_api_key_dict=user_api_key_dict, body_adapter=_OPENAI_DECISION_REQUEST_BODY_ADAPTER, + route_type="adecisions", ) diff --git a/litellm/proxy/guardrails/guardrail_endpoints.py b/litellm/proxy/guardrails/guardrail_endpoints.py index 14ee449790d..94a9d7c2bdf 100644 --- a/litellm/proxy/guardrails/guardrail_endpoints.py +++ b/litellm/proxy/guardrails/guardrail_endpoints.py @@ -1496,7 +1496,7 @@ async def get_guardrail_ui_settings() -> GuardrailUIAddGuardrailSettings: - PII entity categories for UI organization - Content filter settings (patterns and categories) """ - from litellm.decisions.main import supported_decisions_providers + from litellm.decisions.call import supported_decisions_providers from litellm.proxy.guardrails.guardrail_hooks.litellm_content_filter.patterns import ( PATTERN_CATEGORIES, get_available_content_categories, diff --git a/litellm/proxy/route_llm_request.py b/litellm/proxy/route_llm_request.py index 2a9312dc0e3..d8b8a68c8f9 100644 --- a/litellm/proxy/route_llm_request.py +++ b/litellm/proxy/route_llm_request.py @@ -94,6 +94,7 @@ ROUTE_ENDPOINT_MAPPING: Final = { "aocr": "/ocr", "asearch": "/search", "adecisions": "/decisions", + "asystemone": "/systemone", "avideo_generation": "/videos", "avideo_list": "/videos", "avideo_status": "/videos/{video_id}", @@ -519,6 +520,7 @@ RouteType = Literal[ "aocr", "asearch", "adecisions", + "asystemone", "avideo_generation", "avideo_list", "avideo_status", diff --git a/litellm/router.py b/litellm/router.py index f173cdb37b4..fd16c289b0d 100644 --- a/litellm/router.py +++ b/litellm/router.py @@ -2233,6 +2233,11 @@ class Router: self.adecisions = self.factory_function(adecisions, call_type="adecisions") self.decisions = self.factory_function(decisions, call_type="decisions") + from litellm.systemone import asystemone, systemone + + self.asystemone = self.factory_function(asystemone, call_type="asystemone") # pyright: ignore[reportUnknownMemberType] # factory_function takes a bare Callable + self.systemone = self.factory_function(systemone, call_type="systemone") # pyright: ignore[reportUnknownMemberType] # factory_function takes a bare Callable + def _initialize_video_endpoints(self): """Initialize video endpoints.""" from litellm.videos import ( @@ -6996,6 +7001,8 @@ class Router: "search", "adecisions", "decisions", + "asystemone", + "systemone", "aadapter_generate_content", "avideo_generation", "video_generation", @@ -7070,6 +7077,7 @@ class Router: "ocr", "search", "decisions", + "systemone", "video_generation", "video_list", "video_status", @@ -7238,6 +7246,7 @@ class Router: "aocr", "ocr", "adecisions", + "asystemone", "avideo_generation", "avideo_list", "avideo_status", diff --git a/litellm/systemone/__init__.py b/litellm/systemone/__init__.py new file mode 100644 index 00000000000..eb4971a16e4 --- /dev/null +++ b/litellm/systemone/__init__.py @@ -0,0 +1,3 @@ +from litellm.systemone.main import asystemone, systemone + +__all__ = ["asystemone", "systemone"] diff --git a/litellm/systemone/main.py b/litellm/systemone/main.py new file mode 100644 index 00000000000..754b69142f9 --- /dev/null +++ b/litellm/systemone/main.py @@ -0,0 +1,107 @@ +from collections.abc import Mapping +from typing import Final + +import httpx + +from litellm.decisions.call import ( + DecisionsCall, + SystemOneQuestions, + SystemOneRequestFields, + asend, + prepare_call, + reject_other_format, + send, +) +from litellm.types.decisions import DecisionsJSON, DecisionsResponse +from litellm.utils import client # pyright: ignore[reportUnknownVariableType] # client is an untyped decorator + + +def _prepare_systemone_call( + *, + model: str, + state: DecisionsJSON | None, + questions: SystemOneQuestions | None, + safety_identifier: str | None, + api_key: str | None, + api_base: str | None, + timeout: float | httpx.Timeout | None, + custom_llm_provider: str | None, + extra_headers: Mapping[str, str] | None, + kwargs: Mapping[str, object], +) -> DecisionsCall: + reject_other_format(model=model, kwargs=kwargs, field_name="input", other_function="decisions") + return prepare_call( + model=model, + fields=SystemOneRequestFields(state=state, questions=questions), + safety_identifier=safety_identifier, + api_key=api_key, + api_base=api_base, + timeout=timeout, + custom_llm_provider=custom_llm_provider, + extra_headers=extra_headers, + kwargs=kwargs, + ) + + +def _systemone_response(response: object) -> DecisionsResponse: + assert isinstance(response, DecisionsResponse) + return response + + +@client +async def asystemone( + model: str, + state: DecisionsJSON | None = None, + questions: SystemOneQuestions | None = None, + api_key: str | None = None, + api_base: str | None = None, + timeout: float | httpx.Timeout | None = None, + custom_llm_provider: str | None = None, + extra_headers: Mapping[str, str] | None = None, + safety_identifier: str | None = None, + **kwargs: object, # kwargs-ok: litellm_params forwarded by @client +) -> DecisionsResponse: + call: Final = _prepare_systemone_call( + model=model, + state=state, + questions=questions, + safety_identifier=safety_identifier, + api_key=api_key, + api_base=api_base, + timeout=timeout, + custom_llm_provider=custom_llm_provider, + extra_headers=extra_headers, + kwargs=kwargs, + ) + return _systemone_response(await asend(call)) + + +@client +def systemone( + model: str, + state: DecisionsJSON | None = None, + questions: SystemOneQuestions | None = None, + api_key: str | None = None, + api_base: str | None = None, + timeout: float | httpx.Timeout | None = None, + custom_llm_provider: str | None = None, + extra_headers: Mapping[str, str] | None = None, + safety_identifier: str | None = None, + **kwargs: object, # kwargs-ok: litellm_params forwarded by @client +) -> DecisionsResponse: + call: Final = _prepare_systemone_call( + model=model, + state=state, + questions=questions, + safety_identifier=safety_identifier, + api_key=api_key, + api_base=api_base, + timeout=timeout, + custom_llm_provider=custom_llm_provider, + extra_headers=extra_headers, + kwargs=kwargs, + ) + return _systemone_response(send(call)) + + +__all__ = ["asystemone", "systemone"] diff --git a/litellm/types/utils.py b/litellm/types/utils.py index 13fd86b74a7..6f89ac614e8 100644 --- a/litellm/types/utils.py +++ b/litellm/types/utils.py @@ -489,6 +489,8 @@ class CallTypes(str, Enum): asearch = "asearch" decisions = "decisions" adecisions = "adecisions" + systemone = "systemone" + asystemone = "asystemone" arealtime = "_arealtime" aresponses_websocket = "_aresponses_websocket" create_batch = "create_batch" @@ -677,6 +679,8 @@ CallTypesLiteral = Literal[ "asearch", "decisions", "adecisions", + "systemone", + "asystemone", "_arealtime", "_aresponses_websocket", "create_batch", @@ -788,8 +792,8 @@ API_ROUTE_TO_CALL_TYPES: Final[Mapping[str, Sequence[CallTypes]]] = { "/v1/search": [CallTypes.asearch, CallTypes.search], "/decisions": [CallTypes.adecisions, CallTypes.decisions], "/v1/decisions": [CallTypes.adecisions, CallTypes.decisions], - "/systemone": [CallTypes.adecisions, CallTypes.decisions], - "/v1/systemone": [CallTypes.adecisions, CallTypes.decisions], + "/systemone": [CallTypes.asystemone, CallTypes.systemone], + "/v1/systemone": [CallTypes.asystemone, CallTypes.systemone], # Batches "/batches": [CallTypes.acreate_batch, CallTypes.create_batch], "/v1/batches": [CallTypes.acreate_batch, CallTypes.create_batch], diff --git a/litellm/utils.py b/litellm/utils.py index 6774771dbf9..44dc75f45d7 100644 --- a/litellm/utils.py +++ b/litellm/utils.py @@ -974,6 +974,12 @@ def _ocr_document_summary(document: object) -> str: # Runs once per call to check if the user wants to send their data anywhere - PostHog/Sentry/Slack/etc. + +_DECISION_CALL_TYPES: Final[frozenset[str]] = frozenset( + (CallTypes.decisions.value, CallTypes.adecisions.value, CallTypes.systemone.value, CallTypes.asystemone.value) +) + + def function_setup( original_function: str, rules_obj: Rules, @@ -1220,7 +1226,7 @@ def function_setup( if isinstance(search_query, list) else search_query ) - elif call_type in (CallTypes.decisions.value, CallTypes.adecisions.value): + elif call_type in _DECISION_CALL_TYPES: decisions_state: Final = args[1] if len(args) > 1 else kwargs.get("state") or kwargs.get("input") or "" messages = decisions_state if isinstance(decisions_state, str) else json.dumps(decisions_state, default=str) elif call_type in (CallTypes.image_edit.value, CallTypes.aimage_edit.value): diff --git a/tests/integration/cost_calculation/cost_tracking_cases.json b/tests/integration/cost_calculation/cost_tracking_cases.json index 81beff68c7c..9e22212249c 100644 --- a/tests/integration/cost_calculation/cost_tracking_cases.json +++ b/tests/integration/cost_calculation/cost_tracking_cases.json @@ -642,20 +642,20 @@ "gpt-realtime-mini-2025-12-15": { "litellm_provider": "openai", "mode": "realtime", - "input_cost_per_token": 6.0e-07, + "input_cost_per_token": 6e-07, "output_cost_per_token": 2.4e-06, - "input_cost_per_audio_token": 1.0e-05, - "cache_read_input_token_cost": 6.0e-08, - "cache_read_input_audio_token_cost": 3.0e-07, - "output_cost_per_audio_token": 2.0e-05 + "input_cost_per_audio_token": 1e-05, + "cache_read_input_token_cost": 6e-08, + "cache_read_input_audio_token_cost": 3e-07, + "output_cost_per_audio_token": 2e-05 }, "gpt-realtime-2.1": { "litellm_provider": "openai", "mode": "realtime", - "input_cost_per_token": 4.0e-06, + "input_cost_per_token": 4e-06, "input_cost_per_audio_token": 3.2e-05, - "cache_read_input_token_cost": 4.0e-07, - "cache_read_input_audio_token_cost": 4.0e-07, + "cache_read_input_token_cost": 4e-07, + "cache_read_input_audio_token_cost": 4e-07, "output_cost_per_token": 2.4e-05, "output_cost_per_audio_token": 6.4e-05 }, @@ -769,6 +769,52 @@ "search_context_size_medium": 0.014 }, "web_search_billing_unit": "per_query" + }, + "xai/grok-4.7": { + "litellm_provider": "xai", + "mode": "chat", + "input_cost_per_token": 2e-06, + "output_cost_per_token": 6e-06, + "cache_read_input_token_cost": 5e-07, + "max_tokens": 500000, + "max_input_tokens": 500000, + "max_output_tokens": 500000, + "supports_function_calling": true + }, + "us.anthropic.claude-opus-5-5": { + "litellm_provider": "bedrock_converse", + "mode": "chat", + "input_cost_per_token": 4.4e-06, + "output_cost_per_token": 2.2e-05, + "cache_read_input_token_cost": 2.2e-07, + "cache_creation_input_token_cost": 5.5e-06, + "max_tokens": 128000, + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "supports_function_calling": true + }, + "claude-opus-5-5": { + "litellm_provider": "anthropic", + "mode": "chat", + "input_cost_per_token": 4e-06, + "output_cost_per_token": 2e-05, + "cache_read_input_token_cost": 2e-07, + "cache_creation_input_token_cost": 5e-06, + "max_tokens": 128000, + "max_input_tokens": 1000000, + "max_output_tokens": 128000, + "supports_function_calling": true + }, + "gpt-6.1-sol": { + "litellm_provider": "openai", + "mode": "chat", + "input_cost_per_token": 2e-06, + "output_cost_per_token": 1e-05, + "cache_read_input_token_cost": 1e-07, + "max_tokens": 128000, + "max_input_tokens": 922000, + "max_output_tokens": 128000, + "supports_function_calling": true } }, "cases": [ @@ -29687,7 +29733,13 @@ } } }, - "expected": {"spend": 0.0025118, "input_cost": 0.0020792, "output_cost": 0.0004326, "prompt_tokens": 1840, "completion_tokens": 412} + "expected": { + "spend": 0.0025118, + "input_cost": 0.0020792, + "output_cost": 0.0004326, + "prompt_tokens": 1840, + "completion_tokens": 412 + } }, { "name": "deepseek-deepseek-v4-chat-prompt_cache_hit", @@ -32102,6 +32154,190 @@ "prompt_tokens": 100, "completion_tokens": 100 } + }, + { + "name": "xai-grok-4.7-input_text", + "covers": "quota_management.spend_tracking.cost_matrix.logs_cost", + "model": "xai/grok-4.7", + "request": { + "model": "$MODEL", + "messages": [ + { + "role": "user", + "content": "Say hello." + } + ], + "stream": false, + "allowed_openai_params": [] + }, + "response": { + "content_type": "application/json", + "body": { + "id": "chatcmpl-$REQUEST_ID", + "object": "chat.completion", + "model": "grok-4.7", + "choices": [ + { + "index": 0, + "message": { + "role": "assistant", + "content": "Hello." + }, + "finish_reason": "stop" + } + ], + "usage": { + "prompt_tokens": 1840, + "completion_tokens": 412, + "total_tokens": 2252 + } + } + }, + "expected": { + "spend": 0.006152, + "input_cost": 0.00368, + "output_cost": 0.002472, + "prompt_tokens": 1840, + "completion_tokens": 412 + } + }, + { + "name": "us.anthropic.claude-opus-5-5-input_text", + "covers": "quota_management.spend_tracking.cost_matrix.logs_cost", + "model": "us.anthropic.claude-opus-5-5", + "request": { + "model": "$MODEL", + "messages": [ + { + "role": "user", + "content": "Say hello." + } + ], + "stream": false, + "allowed_openai_params": [] + }, + "response": { + "content_type": "application/json", + "body": { + "output": { + "message": { + "role": "assistant", + "content": [ + { + "text": "Hello." + } + ] + } + }, + "stopReason": "end_turn", + "usage": { + "inputTokens": 1840, + "outputTokens": 412, + "totalTokens": 2252 + }, + "metrics": { + "latencyMs": 42 + } + } + }, + "expected": { + "spend": 0.01716, + "input_cost": 0.008096, + "output_cost": 0.009064, + "prompt_tokens": 1840, + "completion_tokens": 412 + } + }, + { + "name": "claude-opus-5-5-input_text", + "covers": "quota_management.spend_tracking.cost_matrix.logs_cost", + "model": "claude-opus-5-5", + "request": { + "model": "$MODEL", + "messages": [ + { + "role": "user", + "content": "Say hello." + } + ], + "stream": false, + "allowed_openai_params": [] + }, + "response": { + "content_type": "application/json", + "body": { + "id": "msg_$REQUEST_ID", + "type": "message", + "role": "assistant", + "model": "claude-opus-5-5", + "content": [ + { + "type": "text", + "text": "Hello." + } + ], + "stop_reason": "end_turn", + "stop_sequence": null, + "usage": { + "input_tokens": 1840, + "output_tokens": 412 + } + } + }, + "expected": { + "spend": 0.0156, + "input_cost": 0.00736, + "output_cost": 0.00824, + "prompt_tokens": 1840, + "completion_tokens": 412 + } + }, + { + "name": "gpt-6.1-sol-input_text", + "covers": "quota_management.spend_tracking.cost_matrix.logs_cost", + "model": "gpt-6.1-sol", + "request": { + "model": "$MODEL", + "messages": [ + { + "role": "user", + "content": "Say hello." + } + ], + "stream": false, + "allowed_openai_params": [] + }, + "response": { + "content_type": "application/json", + "body": { + "id": "chatcmpl-$REQUEST_ID", + "object": "chat.completion", + "created": 1, + "model": "gpt-6.1-sol", + "choices": [ + { + "index": 0, + "message": { + "role": "assistant", + "content": "Hello." + }, + "finish_reason": "stop" + } + ], + "usage": { + "prompt_tokens": 1840, + "completion_tokens": 412, + "total_tokens": 2252 + } + } + }, + "expected": { + "spend": 0.0078, + "input_cost": 0.00368, + "output_cost": 0.00412, + "prompt_tokens": 1840, + "completion_tokens": 412 + } } ], "batch_cases": [ @@ -32149,8 +32385,8 @@ ], "expected": { "spend": 0.000126, - "input_cost": 0.000056, - "output_cost": 0.00007, + "input_cost": 5.6e-05, + "output_cost": 7e-05, "prompt_tokens": 100, "completion_tokens": 10, "cost_header": false diff --git a/tests/integration/cost_calculation/stream_parity/__init__.py b/tests/integration/cost_calculation/stream_parity/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/tests/integration/cost_calculation/stream_parity/bases/__init__.py b/tests/integration/cost_calculation/stream_parity/bases/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/tests/integration/cost_calculation/stream_parity/bases/anthropic.py b/tests/integration/cost_calculation/stream_parity/bases/anthropic.py new file mode 100644 index 00000000000..347e6b17b4f --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/bases/anthropic.py @@ -0,0 +1,123 @@ +from typing import Final + +from integration.cost_calculation.cost_tracking_case import ( + CostTrackingTestCase, + Deployment, + ExactExpected, + JsonResponse, + SseResponse, +) +from integration.cost_calculation.stream_parity.case import COVERS, MODEL, REQUEST_ID, StreamParityTestCase, sse_frames + +CLAUDE_OPUS_5_5_MESSAGES: Final = CostTrackingTestCase( + name="claude-opus-5-5-messages-parity", + covers=COVERS, + model="claude-opus-5-5", + endpoint="/v1/messages", + deployment=Deployment(input_cost_per_token=0.001, output_cost_per_token=0.002), + request={ + "model": MODEL, + "max_tokens": 64, + "messages": [{"role": "user", "content": "Say hello."}], + "cache": {"no-cache": True}, + }, + response=JsonResponse( + content_type="application/json", + body={ + "id": f"msg_{REQUEST_ID}", + "type": "message", + "role": "assistant", + "model": "claude-opus-5-5", + "content": [{"type": "text", "text": "Hello."}], + "stop_reason": "end_turn", + "stop_sequence": None, + "usage": {"input_tokens": 30, "output_tokens": 40}, + }, + ), + expected=ExactExpected(spend=0.11, input_cost=0.03, output_cost=0.08, prompt_tokens=30, completion_tokens=40), +) + +CLAUDE_OPUS_5_5_STREAM_RESPONSE: Final = SseResponse( + content_type="text/event-stream", + frames=sse_frames( + { + "type": "message_start", + "message": { + "id": f"msg_{REQUEST_ID}", + "type": "message", + "role": "assistant", + "model": "claude-opus-5-5", + "content": [], + "stop_reason": None, + "stop_sequence": None, + "usage": {"input_tokens": 30, "output_tokens": 1}, + }, + }, + {"type": "content_block_start", "index": 0, "content_block": {"type": "text", "text": ""}}, + {"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": "Hello."}}, + {"type": "content_block_stop", "index": 0}, + { + "type": "message_delta", + "delta": {"stop_reason": "end_turn", "stop_sequence": None}, + "usage": {"output_tokens": 40}, + }, + {"type": "message_stop"}, + ), +) + +CLAUDE_OPUS_5_5_MESSAGES_PARITY: Final = StreamParityTestCase( + plain=CLAUDE_OPUS_5_5_MESSAGES, + streamed=CLAUDE_OPUS_5_5_MESSAGES.model_copy( + update={ + "name": "claude-opus-5-5-messages-parity-stream", + "request": {**CLAUDE_OPUS_5_5_MESSAGES.request, "stream": True}, + "response": CLAUDE_OPUS_5_5_STREAM_RESPONSE, + } + ), +) + +CLAUDE_OPUS_5_5_CHAT_COMPLETIONS: Final = CLAUDE_OPUS_5_5_MESSAGES.model_copy( + update={ + "name": "claude-opus-5-5-chat-parity", + "endpoint": "/v1/chat/completions", + "request": { + "model": MODEL, + "messages": [{"role": "user", "content": "Say hello."}], + "cache": {"no-cache": True}, + }, + } +) + +CLAUDE_OPUS_5_5_CHAT_COMPLETIONS_PARITY: Final = StreamParityTestCase( + plain=CLAUDE_OPUS_5_5_CHAT_COMPLETIONS, + streamed=CLAUDE_OPUS_5_5_CHAT_COMPLETIONS.model_copy( + update={ + "name": "claude-opus-5-5-chat-parity-stream", + "request": { + **CLAUDE_OPUS_5_5_CHAT_COMPLETIONS.request, + "stream": True, + "stream_options": {"include_usage": True}, + }, + "response": CLAUDE_OPUS_5_5_STREAM_RESPONSE, + } + ), +) + +CLAUDE_OPUS_5_5_RESPONSES: Final = CLAUDE_OPUS_5_5_MESSAGES.model_copy( + update={ + "name": "claude-opus-5-5-responses-parity", + "endpoint": "/v1/responses", + "request": {"model": MODEL, "input": "Say hello.", "cache": {"no-cache": True}}, + } +) + +CLAUDE_OPUS_5_5_RESPONSES_PARITY: Final = StreamParityTestCase( + plain=CLAUDE_OPUS_5_5_RESPONSES, + streamed=CLAUDE_OPUS_5_5_RESPONSES.model_copy( + update={ + "name": "claude-opus-5-5-responses-parity-stream", + "request": {**CLAUDE_OPUS_5_5_RESPONSES.request, "stream": True}, + "response": CLAUDE_OPUS_5_5_STREAM_RESPONSE, + } + ), +) diff --git a/tests/integration/cost_calculation/stream_parity/bases/bedrock.py b/tests/integration/cost_calculation/stream_parity/bases/bedrock.py new file mode 100644 index 00000000000..501cb5e8cab --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/bases/bedrock.py @@ -0,0 +1,62 @@ +"""/v1/messages on a Bedrock Converse deployment: the messages -> converse bridge.""" + +from typing import Final + +from integration.cost_calculation.cost_tracking_case import ( + CostTrackingTestCase, + Deployment, + EventStreamEvent, + EventStreamResponse, + ExactExpected, + JsonResponse, +) +from integration.cost_calculation.stream_parity.case import COVERS, MODEL, StreamParityTestCase + +_USAGE: Final = {"inputTokens": 30, "outputTokens": 40, "totalTokens": 70} + +CLAUDE_OPUS_5_5_CONVERSE_MESSAGES: Final = CostTrackingTestCase( + name="bedrock-converse-claude-opus-5-5-messages-parity", + covers=COVERS, + model="us.anthropic.claude-opus-5-5", + endpoint="/v1/messages", + deployment=Deployment(input_cost_per_token=0.001, output_cost_per_token=0.002), + request={ + "model": MODEL, + "max_tokens": 64, + "messages": [{"role": "user", "content": "Say hello."}], + "cache": {"no-cache": True}, + }, + response=JsonResponse( + content_type="application/json", + body={ + "output": {"message": {"role": "assistant", "content": [{"text": "Hello."}]}}, + "stopReason": "end_turn", + "usage": _USAGE, + "metrics": {"latencyMs": 1}, + }, + ), + expected=ExactExpected(spend=0.11, input_cost=0.03, output_cost=0.08, prompt_tokens=30, completion_tokens=40), +) + +CLAUDE_OPUS_5_5_CONVERSE_MESSAGES_PARITY: Final = StreamParityTestCase( + plain=CLAUDE_OPUS_5_5_CONVERSE_MESSAGES, + streamed=CLAUDE_OPUS_5_5_CONVERSE_MESSAGES.model_copy( + update={ + "name": "bedrock-converse-claude-opus-5-5-messages-parity-stream", + "request": {**CLAUDE_OPUS_5_5_CONVERSE_MESSAGES.request, "stream": True}, + "response": EventStreamResponse( + content_type="application/vnd.amazon.eventstream", + framing="converse", + events=( + EventStreamEvent(event_type="messageStart", payload={"role": "assistant"}), + EventStreamEvent( + event_type="contentBlockDelta", payload={"delta": {"text": "Hello."}, "contentBlockIndex": 0} + ), + EventStreamEvent(event_type="contentBlockStop", payload={"contentBlockIndex": 0}), + EventStreamEvent(event_type="messageStop", payload={"stopReason": "end_turn"}), + EventStreamEvent(event_type="metadata", payload={"usage": _USAGE, "metrics": {"latencyMs": 1}}), + ), + ), + } + ), +) diff --git a/tests/integration/cost_calculation/stream_parity/bases/openai.py b/tests/integration/cost_calculation/stream_parity/bases/openai.py new file mode 100644 index 00000000000..ef94094e9f0 --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/bases/openai.py @@ -0,0 +1,91 @@ +from typing import Final + +from integration.cost_calculation.cost_tracking_case import ( + CostTrackingTestCase, + Deployment, + ExactExpected, + JsonResponse, + SseResponse, +) +from integration.cost_calculation.stream_parity.bases.openai_responses import ( + GPT_6_1_SOL_RESPONSES, + GPT_6_1_SOL_RESPONSES_PARITY_STREAM_RESPONSE, +) +from integration.cost_calculation.stream_parity.case import COVERS, MODEL, REQUEST_ID, StreamParityTestCase, sse_frames + +_USAGE: Final = {"prompt_tokens": 30, "completion_tokens": 40, "total_tokens": 70} +_CHUNK: Final = { + "id": f"chatcmpl-{REQUEST_ID}", + "object": "chat.completion.chunk", + "created": 1, + "model": "gpt-6.1-sol", +} + +GPT_6_1_SOL_CHAT_COMPLETIONS: Final = CostTrackingTestCase( + name="gpt-6.1-sol-chat-parity", + covers=COVERS, + model="gpt-6.1-sol", + endpoint="/v1/chat/completions", + deployment=Deployment(input_cost_per_token=0.001, output_cost_per_token=0.002), + request={"model": MODEL, "messages": [{"role": "user", "content": "Say hello."}], "cache": {"no-cache": True}}, + response=JsonResponse( + content_type="application/json", + body={ + "id": f"chatcmpl-{REQUEST_ID}", + "object": "chat.completion", + "created": 1, + "model": "gpt-6.1-sol", + "choices": [{"index": 0, "message": {"role": "assistant", "content": "Hello."}, "finish_reason": "stop"}], + "usage": _USAGE, + }, + ), + expected=ExactExpected(spend=0.11, input_cost=0.03, output_cost=0.08, prompt_tokens=30, completion_tokens=40), +) + +GPT_6_1_SOL_CHAT_COMPLETIONS_PARITY: Final = StreamParityTestCase( + plain=GPT_6_1_SOL_CHAT_COMPLETIONS, + streamed=GPT_6_1_SOL_CHAT_COMPLETIONS.model_copy( + update={ + "name": "gpt-6.1-sol-chat-parity-stream", + "request": { + **GPT_6_1_SOL_CHAT_COMPLETIONS.request, + "stream": True, + "stream_options": {"include_usage": True}, + }, + "response": SseResponse( + content_type="text/event-stream", + frames=sse_frames( + {**_CHUNK, "choices": [{"index": 0, "delta": {"role": "assistant", "content": "Hello."}}]}, + {**_CHUNK, "choices": [{"index": 0, "delta": {}, "finish_reason": "stop"}]}, + {**_CHUNK, "choices": [], "usage": _USAGE}, + done=True, + ), + ), + } + ), +) + +GPT_6_1_SOL_MESSAGES: Final = GPT_6_1_SOL_RESPONSES.model_copy( + update={ + "name": "gpt-6.1-sol-messages-parity", + "model": "gpt-6.1-sol", + "endpoint": "/v1/messages", + "request": { + "model": MODEL, + "max_tokens": 64, + "messages": [{"role": "user", "content": "Say hello."}], + "cache": {"no-cache": True}, + }, + } +) + +GPT_6_1_SOL_MESSAGES_PARITY: Final = StreamParityTestCase( + plain=GPT_6_1_SOL_MESSAGES, + streamed=GPT_6_1_SOL_MESSAGES.model_copy( + update={ + "name": "gpt-6.1-sol-messages-parity-stream", + "request": {**GPT_6_1_SOL_MESSAGES.request, "stream": True}, + "response": GPT_6_1_SOL_RESPONSES_PARITY_STREAM_RESPONSE, + } + ), +) diff --git a/tests/integration/cost_calculation/stream_parity/bases/openai_responses.py b/tests/integration/cost_calculation/stream_parity/bases/openai_responses.py new file mode 100644 index 00000000000..93a47ed74bf --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/bases/openai_responses.py @@ -0,0 +1,115 @@ +from typing import Final + +from integration.cost_calculation.cost_tracking_case import ( + CostTrackingTestCase, + Deployment, + ExactExpected, + JsonResponse, + SseResponse, +) +from integration.cost_calculation.stream_parity.case import COVERS, MODEL, REQUEST_ID, StreamParityTestCase, sse_frames + +_ITEM: Final = { + "id": "msg_1", + "type": "message", + "role": "assistant", + "status": "completed", + "content": [{"type": "output_text", "text": "Hello.", "annotations": []}], +} +_RESPONSE: Final = { + "id": f"resp_{REQUEST_ID}", + "object": "response", + "created_at": 1, + "status": "completed", + "model": "gpt-6.1-sol", + "output": [_ITEM], + "usage": {"input_tokens": 30, "output_tokens": 40, "total_tokens": 70}, +} + +GPT_6_1_SOL_RESPONSES: Final = CostTrackingTestCase( + name="gpt-6.1-sol-responses-parity", + covers=COVERS, + model="gpt-6.1-sol", + endpoint="/v1/responses", + deployment=Deployment(input_cost_per_token=0.001, output_cost_per_token=0.002), + request={"model": MODEL, "input": "Say hello.", "cache": {"no-cache": True}}, + response=JsonResponse(content_type="application/json", body=_RESPONSE), + expected=ExactExpected(spend=0.11, input_cost=0.03, output_cost=0.08, prompt_tokens=30, completion_tokens=40), +) + +GPT_6_1_SOL_RESPONSES_PARITY_STREAM_RESPONSE: Final = SseResponse( + content_type="text/event-stream", + frames=sse_frames( + { + "type": "response.created", + "sequence_number": 0, + "response": {**_RESPONSE, "status": "in_progress", "output": []}, + }, + { + "type": "response.output_item.added", + "sequence_number": 1, + "output_index": 0, + "item": {**_ITEM, "status": "in_progress", "content": []}, + }, + { + "type": "response.output_text.delta", + "sequence_number": 2, + "item_id": "msg_1", + "output_index": 0, + "content_index": 0, + "delta": "Hello.", + }, + {"type": "response.output_item.done", "sequence_number": 3, "output_index": 0, "item": _ITEM}, + {"type": "response.completed", "sequence_number": 4, "response": _RESPONSE}, + ), +) + +GPT_6_1_SOL_RESPONSES_PARITY: Final = StreamParityTestCase( + plain=GPT_6_1_SOL_RESPONSES, + streamed=GPT_6_1_SOL_RESPONSES.model_copy( + update={ + "name": "gpt-6.1-sol-responses-parity-stream", + "request": {**GPT_6_1_SOL_RESPONSES.request, "stream": True}, + "response": GPT_6_1_SOL_RESPONSES_PARITY_STREAM_RESPONSE, + } + ), +) + +_PRO_RESPONSE: Final = {**_RESPONSE, "model": "gpt-5.5-pro"} +GPT_5_5_PRO_STREAM_RESPONSE: Final = GPT_6_1_SOL_RESPONSES_PARITY_STREAM_RESPONSE.model_copy( + update={ + "frames": tuple( + frame.replace('"model": "gpt-6.1-sol"', '"model": "gpt-5.5-pro"') + for frame in GPT_6_1_SOL_RESPONSES_PARITY_STREAM_RESPONSE.frames + ) + } +) + +GPT_5_5_PRO_CHAT_COMPLETIONS: Final = GPT_6_1_SOL_RESPONSES.model_copy( + update={ + "name": "gpt-5.5-pro-chat-parity", + "model": "gpt-5.5-pro", + "response": JsonResponse(content_type="application/json", body=_PRO_RESPONSE), + "endpoint": "/v1/chat/completions", + "request": { + "model": MODEL, + "messages": [{"role": "user", "content": "Say hello."}], + "cache": {"no-cache": True}, + }, + } +) + +GPT_5_5_PRO_CHAT_COMPLETIONS_PARITY: Final = StreamParityTestCase( + plain=GPT_5_5_PRO_CHAT_COMPLETIONS, + streamed=GPT_5_5_PRO_CHAT_COMPLETIONS.model_copy( + update={ + "name": "gpt-5.5-pro-chat-parity-stream", + "request": { + **GPT_5_5_PRO_CHAT_COMPLETIONS.request, + "stream": True, + "stream_options": {"include_usage": True}, + }, + "response": GPT_5_5_PRO_STREAM_RESPONSE, + } + ), +) diff --git a/tests/integration/cost_calculation/stream_parity/bases/xai.py b/tests/integration/cost_calculation/stream_parity/bases/xai.py new file mode 100644 index 00000000000..2ab0a5a5c4a --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/bases/xai.py @@ -0,0 +1,65 @@ +"""/v1/messages on a chat-completions provider: the messages -> chat/completions bridge.""" + +from typing import Final + +from integration.cost_calculation.cost_tracking_case import ( + CostTrackingTestCase, + Deployment, + ExactExpected, + JsonResponse, + SseResponse, +) +from integration.cost_calculation.stream_parity.case import COVERS, MODEL, REQUEST_ID, StreamParityTestCase, sse_frames + +_USAGE: Final = {"prompt_tokens": 30, "completion_tokens": 40, "total_tokens": 70} +_CHUNK: Final = { + "id": f"chatcmpl-{REQUEST_ID}", + "object": "chat.completion.chunk", + "created": 1, + "model": "grok-4.7", +} + +GROK_4_7_MESSAGES: Final = CostTrackingTestCase( + name="grok-4.7-messages-parity", + covers=COVERS, + model="xai/grok-4.7", + endpoint="/v1/messages", + deployment=Deployment(input_cost_per_token=0.001, output_cost_per_token=0.002), + request={ + "model": MODEL, + "max_tokens": 64, + "messages": [{"role": "user", "content": "Say hello."}], + "cache": {"no-cache": True}, + }, + response=JsonResponse( + content_type="application/json", + body={ + "id": f"chatcmpl-{REQUEST_ID}", + "object": "chat.completion", + "created": 1, + "model": "grok-4.7", + "choices": [{"index": 0, "message": {"role": "assistant", "content": "Hello."}, "finish_reason": "stop"}], + "usage": _USAGE, + }, + ), + expected=ExactExpected(spend=0.11, input_cost=0.03, output_cost=0.08, prompt_tokens=30, completion_tokens=40), +) + +GROK_4_7_MESSAGES_PARITY: Final = StreamParityTestCase( + plain=GROK_4_7_MESSAGES, + streamed=GROK_4_7_MESSAGES.model_copy( + update={ + "name": "grok-4.7-messages-parity-stream", + "request": {**GROK_4_7_MESSAGES.request, "stream": True}, + "response": SseResponse( + content_type="text/event-stream", + frames=sse_frames( + {**_CHUNK, "choices": [{"index": 0, "delta": {"role": "assistant", "content": "Hello."}}]}, + {**_CHUNK, "choices": [{"index": 0, "delta": {}, "finish_reason": "stop"}]}, + {**_CHUNK, "choices": [], "usage": _USAGE}, + done=True, + ), + ), + } + ), +) diff --git a/tests/integration/cost_calculation/stream_parity/case.py b/tests/integration/cost_calculation/stream_parity/case.py new file mode 100644 index 00000000000..1eb7c598939 --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/case.py @@ -0,0 +1,44 @@ +import json +from collections.abc import Mapping +from dataclasses import dataclass +from typing import Final + +from integration.cost_calculation.cost_tracking_case import CostTrackingTestCase +from pydantic import JsonValue + +COVERS: Final = "quota_management.spend_tracking.scripted_wire.logs_cost" +REQUEST_ID: Final = "$REQUEST_ID" +MODEL: Final = "$MODEL" + + +def sse_frames(*events: Mapping[str, JsonValue], done: bool = False) -> tuple[str, ...]: + frames: Final = tuple( + f"event: {event['type']}\ndata: {json.dumps(event)}" if "type" in event else f"data: {json.dumps(event)}" + for event in events + ) + return (*frames, "data: [DONE]") if done else frames + + +@dataclass(frozen=True, slots=True, kw_only=True) +class StreamParityTestCase: + """The same request as two CostTrackingTestCases, `plain` answered with JSON and `streamed` answered with SSE, + that must both bill the one `expected` row.""" + + plain: CostTrackingTestCase + streamed: CostTrackingTestCase + + def __post_init__(self) -> None: + if self.plain.expected != self.streamed.expected: + raise ValueError(f"{self.id}: plain and streamed cases expect different rows") + if self.streamed.request.get("stream") is not True or self.plain.request.get("stream") is True: + raise ValueError(f"{self.id}: streamed case must set stream: true and the plain case must not") + if (self.plain.model, self.plain.endpoint, self.plain.deployment) != ( + self.streamed.model, + self.streamed.endpoint, + self.streamed.deployment, + ): + raise ValueError(f"{self.id}: plain and streamed cases must share model, endpoint and deployment") + + @property + def id(self) -> str: + return self.plain.name diff --git a/tests/integration/cost_calculation/stream_parity/runner.py b/tests/integration/cost_calculation/stream_parity/runner.py new file mode 100644 index 00000000000..17ce7058c14 --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/runner.py @@ -0,0 +1,70 @@ +from collections import Counter +from hashlib import sha256 +from typing import Final + +from integration._support.client import Gateway, Scenario +from integration.cost_calculation.assertions import assert_exact +from integration.cost_calculation.conftest import ( + poll_rows, + poll_rows_where, + read_rows_now, + register_scenario_deployment, +) +from integration.cost_calculation.cost_tracking_case import ExactExpected +from integration.cost_calculation.stream_parity.bases import openai_responses +from integration.cost_calculation.stream_parity.case import StreamParityTestCase + +FLUSH_CASE: Final = openai_responses.GPT_6_1_SOL_RESPONSES + + +def flush_spend_logs(scenario: Scenario, gateway: Gateway, key: str) -> str: + """A /v1/responses row wakes the spend-log writer at once, and the queue is FIFO, so every row queued before it lands.""" + flush: Final = register_scenario_deployment(scenario, FLUSH_CASE, "flush", key) + response: Final = gateway.request( + "POST", FLUSH_CASE.endpoint, {**FLUSH_CASE.request, "model": flush.model_name}, key=key + ) + assert response.status_code == 200, ("flush", response.text[:400]) + poll_rows_where(key, 1, lambda row: row.model_id == flush.identity) + return flush.identity + + +def assert_stream_parity(case: StreamParityTestCase, gateway: Gateway) -> None: + expected: Final = case.plain.expected + assert isinstance(expected, ExactExpected), case.id + with gateway.scenario() as scenario: + key: Final = scenario.key() + deployments: Final = { + leg.name: register_scenario_deployment(scenario, leg, sha256(leg.name.encode()).hexdigest()[:12], key) + for leg in (case.plain, case.streamed) + } + responses: Final = { + leg.name: gateway.request( + "POST", leg.endpoint, {**leg.request, "model": deployments[leg.name].model_name}, key=key + ) + for leg in (case.plain, case.streamed) + } + for leg in (case.plain, case.streamed): + assert responses[leg.name].status_code == 200, (leg.name, responses[leg.name].text[:400]) + + poll_rows(key, 2) + flush_identity: Final = flush_spend_logs(scenario, gateway, key) + rows: Final = read_rows_now(key) + rows_per_deployment: Final = Counter(row.model_id for row in rows) + expected_rows: Final = {deployments[leg.name].identity: 1 for leg in (case.plain, case.streamed)} + assert rows_per_deployment == {**expected_rows, flush_identity: 1}, ( + f"{case.id}: expected exactly one SpendLogs row per leg plus the flush row, got {dict(rows_per_deployment)}", + rows, + ) + billed: Final = { + leg.name: next(row for row in rows if row.model_id == deployments[leg.name].identity) + for leg in (case.plain, case.streamed) + } + for leg in (case.plain, case.streamed): + assert_exact(leg.name, leg.response.content_type, expected, billed[leg.name], responses[leg.name]) + plain: Final = billed[case.plain.name] + streamed: Final = billed[case.streamed.name] + assert (streamed.spend, streamed.prompt_tokens, streamed.completion_tokens) == ( + plain.spend, + plain.prompt_tokens, + plain.completion_tokens, + ), (plain, streamed) diff --git a/tests/integration/cost_calculation/stream_parity/test_basic.py b/tests/integration/cost_calculation/stream_parity/test_basic.py new file mode 100644 index 00000000000..00089f9bb26 --- /dev/null +++ b/tests/integration/cost_calculation/stream_parity/test_basic.py @@ -0,0 +1,40 @@ +"""A streamed request bills exactly what the same request bills without streaming. + +Streaming reconstructs the response, its usage and its model name from chunks after the fact, in a separate code +path from the non-streamed one (LIT-6241, LIT-6872, LIT-7632, LIT-7729, LIT-9044, LIT-9065). Each case is a plain and +a streamed CostTrackingTestCase for one request, sent through the same key, and both LiteLLM_SpendLogs rows must +bill the case's expected row. + +Three cells are the native routes (chat, /v1/messages, /v1/responses); the rest are the bridges that translate one +API into another before the provider call, each with its own stream reassembly +""" + +from typing import Final + +import pytest +from integration._support.client import Gateway +from integration.cost_calculation.stream_parity.bases import anthropic, bedrock, openai, openai_responses, xai +from integration.cost_calculation.stream_parity.case import StreamParityTestCase +from integration.cost_calculation.stream_parity.runner import assert_stream_parity + +NATIVE: Final = ( + openai.GPT_6_1_SOL_CHAT_COMPLETIONS_PARITY, + anthropic.CLAUDE_OPUS_5_5_MESSAGES_PARITY, + openai_responses.GPT_6_1_SOL_RESPONSES_PARITY, +) +BRIDGES: Final = ( + xai.GROK_4_7_MESSAGES_PARITY, + openai.GPT_6_1_SOL_MESSAGES_PARITY, + bedrock.CLAUDE_OPUS_5_5_CONVERSE_MESSAGES_PARITY, + openai_responses.GPT_5_5_PRO_CHAT_COMPLETIONS_PARITY, + anthropic.CLAUDE_OPUS_5_5_RESPONSES_PARITY, +) +CASES: Final = (*NATIVE, *BRIDGES) + + +@pytest.mark.parametrize("case", CASES, ids=[case.id for case in CASES]) +@pytest.mark.timeout(180) +def test_streamed_request_bills_the_same_row_as_the_non_streamed_request( + case: StreamParityTestCase, gateway: Gateway +) -> None: + assert_stream_parity(case, gateway) diff --git a/tests/integration/providers/test_databricks_decisions_wire.py b/tests/integration/providers/test_databricks_decisions_wire.py index 93dc2ea04fa..9b9f54e815d 100644 --- a/tests/integration/providers/test_databricks_decisions_wire.py +++ b/tests/integration/providers/test_databricks_decisions_wire.py @@ -179,7 +179,7 @@ def _assert_refused_before_any_upstream_call(gateway: Gateway, model: str, messa assert message in response.text, response.text assert _upstream_calls(gateway, handle) == [] row: Final = _spend_row(response.headers["x-litellm-call-id"]) - assert (row["status"], row["call_type"], _number(row["spend"])) == ("failure", "adecisions", 0.0), row + assert (row["status"], row["call_type"], _number(row["spend"])) == ("failure", "asystemone", 0.0), row @pytest.mark.parametrize("name", [name for _, name in _NON_BARE_NAMES], ids=[label for label, _ in _NON_BARE_NAMES]) @@ -251,7 +251,7 @@ def test_an_openai_format_request_at_v1_decisions_reaches_the_serving_endpoint_a row["model_group"], row["api_base"], _number(row["spend"]), - ) == ("success", "adecisions", "databricks", deployment, f"{handle.api_base()}/{_ENDPOINT}/invocations", 0.0), ( + ) == ("success", "asystemone", "databricks", deployment, f"{handle.api_base()}/{_ENDPOINT}/invocations", 0.0), ( row ) @@ -284,7 +284,7 @@ def test_a_wildcard_deployment_sends_each_request_to_the_named_endpoint_and_bill row["model_group"], row["api_base"], _number(row["spend"]), - ) == ("success", "adecisions", "databricks", model, f"{handle.api_base()}/{_ENDPOINT}/invocations", 0.0), ( + ) == ("success", "asystemone", "databricks", model, f"{handle.api_base()}/{_ENDPOINT}/invocations", 0.0), ( row ) @@ -322,7 +322,7 @@ def test_a_five_kilobyte_endpoint_name_reaches_the_upstream_and_its_not_found_an row: Final = _spend_row(response.headers["x-litellm-call-id"]) assert (row["status"], row["call_type"], row["model_group"], _number(row["spend"])) == ( "failure", - "adecisions", + "asystemone", f"{prefix}/{name}", 0.0, ), row @@ -343,7 +343,7 @@ def test_a_databricks_error_keeps_its_status_and_message_and_bills_nothing( row: Final = _spend_row(response.headers["x-litellm-call-id"]) assert (row["status"], row["call_type"], row["model_group"], _number(row["spend"])) == ( "failure", - "adecisions", + "asystemone", model, 0.0, ), row @@ -515,4 +515,4 @@ def test_a_megabyte_endpoint_name_is_refused_in_linear_time_while_liveliness_sta assert _BARE_NAME_RULE in response.text, response.text[-400:] assert _upstream_calls(gateway, handle) == [] row: Final = _spend_row(response.headers["x-litellm-call-id"]) - assert (row["status"], row["call_type"], _number(row["spend"])) == ("failure", "adecisions", 0.0), row + assert (row["status"], row["call_type"], _number(row["spend"])) == ("failure", "asystemone", 0.0), row diff --git a/tests/integration/providers/test_decisions_wire.py b/tests/integration/providers/test_decisions_wire.py index 16c938e3995..2862632d132 100644 --- a/tests/integration/providers/test_decisions_wire.py +++ b/tests/integration/providers/test_decisions_wire.py @@ -278,7 +278,7 @@ def test_each_provider_gets_its_own_path_key_and_body_and_is_billed(gateway: Gat row["api_base"], row["prompt_tokens"], row["completion_tokens"], - ) == ("success", "adecisions", provider.name, model, f"{handle.api_base()}{provider.path}", 367, 3) + ) == ("success", "asystemone", provider.name, model, f"{handle.api_base()}{provider.path}", 367, 3) assert math.isclose(_number(row["spend"]), expected_spend, rel_tol=1e-9), row @@ -339,7 +339,7 @@ async def test_sdk_sync_and_async_clients_send_the_same_request(gateway: Gateway synchronous: Final = litellm.decisions( model=provider.model, state=_STATE, questions=_SDK_QUESTIONS, api_base=handle.api_base(), api_key=_API_KEY ) - asynchronous: Final = await litellm.adecisions( + asynchronous: Final = await litellm.asystemone( model=provider.model, state=_STATE, questions=_SDK_QUESTIONS, api_base=handle.api_base(), api_key=_API_KEY ) for response in (synchronous, asynchronous): @@ -507,7 +507,7 @@ def test_upstream_errors_keep_their_status_and_log_an_unbilled_failure(gateway: row: Final = _spend_row(response.headers["x-litellm-call-id"]) assert (row["status"], row["call_type"], row["model_group"], _number(row["spend"])) == ( "failure", - "adecisions", + "asystemone", model, 0.0, ) @@ -579,7 +579,7 @@ async def test_sdk_openrouter_connection_failures_raise_a_connection_error(gatew model=_OPENROUTER.model, state=_STATE, questions=_SDK_QUESTIONS, api_base=dead_api_base, api_key=_API_KEY ) with pytest.raises(litellm.APIConnectionError): - await litellm.adecisions( + await litellm.asystemone( model=_OPENROUTER.model, state=_STATE, questions=_SDK_QUESTIONS, api_base=dead_api_base, api_key=_API_KEY ) diff --git a/tests/integration/providers/test_strands_decider_agentcore_wire.py b/tests/integration/providers/test_strands_decider_agentcore_wire.py index bf5da453137..493d8c238fa 100644 --- a/tests/integration/providers/test_strands_decider_agentcore_wire.py +++ b/tests/integration/providers/test_strands_decider_agentcore_wire.py @@ -415,7 +415,7 @@ def test_an_arn_deployment_answers_on_every_decisions_route_signed_for_bedrock_a row = _spend_row(response.headers["x-litellm-call-id"]) assert row == { "status": "success", - "call_type": "adecisions", + "call_type": "asystemone", "custom_llm_provider": "strands_decider", "model_group": _SIGNED, "api_base": _invocations_url("audit_signed"), @@ -694,7 +694,7 @@ async def test_sdk_sync_and_async_calls_reach_the_runtime_signed( aws_access_key_id=_ACCESS_KEY, aws_secret_access_key=_SECRET_KEY, ) - asynchronous: Final = await litellm.adecisions( + asynchronous: Final = await litellm.asystemone( model=_MODEL, state=_STATE, questions=_SDK_QUESTIONS, @@ -703,7 +703,7 @@ async def test_sdk_sync_and_async_calls_reach_the_runtime_signed( aws_secret_access_key=_SECRET_KEY, ) monkeypatch.setenv("STRANDS_DECIDER_API_KEY", _ENV_JWT) - bearer: Final = await litellm.adecisions( + bearer: Final = await litellm.asystemone( model=_MODEL, state=_STATE, questions=_SDK_QUESTIONS, api_base=_arn("audit_sdk_env_key") ) finally: diff --git a/tests/integration/providers/test_vertex_timestamps_utc_wire.py b/tests/integration/providers/test_vertex_timestamps_utc_wire.py new file mode 100644 index 00000000000..e5926874377 --- /dev/null +++ b/tests/integration/providers/test_vertex_timestamps_utc_wire.py @@ -0,0 +1,333 @@ +import json +import socket +import threading +import uuid +from collections.abc import Generator, Iterator +from contextlib import contextmanager, suppress +from dataclasses import dataclass +from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer +from pathlib import Path +from typing import Final +from urllib.parse import parse_qs, urlsplit + +import pytest +import yaml +from integration._support.client import JSON_OBJECT, Gateway, Scenario, gateway_from_environment, string_value +from integration._support.process import owned_proxy +from integration._support.tls import server_context, write_self_signed_cert +from integration._support.vertex import service_account_json +from integration._support.wire import Reply, Request, Wire, wire_server +from pydantic import JsonValue + +pytestmark: Final = pytest.mark.timeout(240) + +PROJECT: Final = "scripted-utc-project" +LOCATION: Final = "us-central1" +BUCKET: Final = "scripted-utc-bucket" +GCS_AUTHORITY: Final = "storage.googleapis.com:443" +JOBS_PATH: Final = f"projects/{PROJECT}/locations/{LOCATION}/batchPredictionJobs" +OUTPUT_PREFIX: Final = f"gs://{BUCKET}/litellm-vertex-files/publishers/google/models/gemini-2.5-flash" +VEO_MODEL: Final = "veo-3.1-generate-001" +VEO_OPERATION: Final = f"projects/{PROJECT}/locations/{LOCATION}/publishers/google/models/{VEO_MODEL}/operations/op-1" + +UPLOAD_TIME_CREATED: Final = "2026-10-08T21:15:30.250Z" +UPLOAD_EPOCH: Final = 1791494130 +BATCH_CREATE_TIME: Final = "2026-10-08T21:16:05.123456Z" +WHOLE_SECOND_BATCH_CREATE_TIME: Final = "2026-10-08T21:16:05Z" +WHOLE_SECOND_MODEL: Final = "gemini-2.5-flash-lite" +WHOLE_SECOND_JOB_SUFFIX: Final = "1" +BATCH_EPOCH: Final = 1791494165 +OUTPUT_TIME_CREATED: Final = "2026-10-08T23:41:07.902Z" +OUTPUT_EPOCH: Final = 1791502867 +OUTPUT_BYTES: Final = 2068 +VEO_CREATE_TIME: Final = "2026-10-08T22:00:00.500000Z" +VEO_EPOCH: Final = 1791496800 + +HOST_ZONE: Final = "America/Los_Angeles" + + +def _gcs_object(name: str, size: int, time_created: str) -> bytes: + return json.dumps( + { + "kind": "storage#object", + "id": f"{BUCKET}/{name}/1791494130250000", + "name": name, + "bucket": BUCKET, + "size": str(size), + "contentType": "application/jsonl", + "timeCreated": time_created, + "updated": time_created, + } + ).encode() + + +def _batch_job(job_id: str, state: str, output_directory: str | None) -> bytes: + create_time: Final = ( + WHOLE_SECOND_BATCH_CREATE_TIME if job_id.endswith(WHOLE_SECOND_JOB_SUFFIX) else BATCH_CREATE_TIME + ) + return json.dumps( + { + "name": f"{JOBS_PATH}/{job_id}", + "displayName": "litellm-vertex-batch-scripted", + "model": "publishers/google/models/gemini-2.5-flash", + "state": state, + "outputInfo": {"gcsOutputDirectory": output_directory} if output_directory else None, + "createTime": create_time, + "updateTime": create_time, + } + ).encode() + + +def vertex_peer(request: Request) -> Reply: + path: Final = urlsplit(request.target).path + if request.method == "POST" and path == "/_oauth/token": + return Reply(body=b'{"access_token":"scripted-token","expires_in":3600,"token_type":"Bearer"}') + if request.method == "POST" and path == f"/upload/storage/v1/b/{BUCKET}/o": + name: Final = parse_qs(urlsplit(request.target).query)["name"][0] + return Reply(body=_gcs_object(name, len(request.body), UPLOAD_TIME_CREATED)) + if request.method == "POST" and path.endswith("/batchPredictionJobs"): + suffix: Final = WHOLE_SECOND_JOB_SUFFIX if WHOLE_SECOND_MODEL.encode() in request.body else "0" + return Reply(body=_batch_job(str(uuid.uuid4().int)[:18] + suffix, "JOB_STATE_PENDING", None)) + if request.method == "GET" and f"/{JOBS_PATH}/" in path: + job_id: Final = path.rsplit("/", 1)[-1] + return Reply(body=_batch_job(job_id, "JOB_STATE_SUCCEEDED", f"{OUTPUT_PREFIX}/prediction-model-{job_id}")) + if request.method == "GET" and path.startswith(f"/storage/v1/b/{BUCKET}/o/"): + object_name: Final = path.removeprefix(f"/storage/v1/b/{BUCKET}/o/").replace("%2F", "/") + return Reply(body=_gcs_object(object_name, OUTPUT_BYTES, OUTPUT_TIME_CREATED)) + if request.method == "POST" and path.endswith(f"/{VEO_MODEL}:predictLongRunning"): + return Reply(body=json.dumps({"name": VEO_OPERATION}).encode()) + if request.method == "POST" and path.endswith(f"/{VEO_MODEL}:fetchPredictOperation"): + return Reply( + body=json.dumps( + {"name": VEO_OPERATION, "done": False, "metadata": {"createTime": VEO_CREATE_TIME}} + ).encode() + ) + return Reply(status=404, body=b'{"error":"not scripted"}') + + +def _pipe(source: socket.socket, sink: socket.socket) -> None: + with suppress(OSError): + for chunk in iter(lambda: source.recv(65536), b""): + sink.sendall(chunk) + with suppress(OSError): + sink.shutdown(socket.SHUT_WR) + + +@contextmanager +def gcs_tunnel(destination: Wire) -> Generator[str, None, None]: + """HTTPS_PROXY that sends CONNECT storage.googleapis.com:443 to the TLS peer and refuses every other host.""" + destination_port: Final = urlsplit(destination.url).port + + class Handler(BaseHTTPRequestHandler): + protocol_version = "HTTP/1.1" + timeout = 30 + + def do_CONNECT(self) -> None: + if self.path != GCS_AUTHORITY: + self.send_error(403) + return + self.send_response(200, "Connection Established") + self.end_headers() + with socket.create_connection(("127.0.0.1", destination_port), timeout=10) as upstream: + outbound: Final = threading.Thread(target=_pipe, args=(self.connection, upstream), daemon=True) + outbound.start() + _pipe(upstream, self.connection) + outbound.join(timeout=30) + + def log_message(self, format: str, *args: object) -> None: + pass + + class Server(ThreadingHTTPServer): + daemon_threads = True + + with Server(("127.0.0.1", 0), Handler) as server: + thread: Final = threading.Thread(target=server.serve_forever, kwargs={"poll_interval": 0.05}) + thread.start() + try: + yield f"http://127.0.0.1:{server.server_port}" + finally: + server.shutdown() + thread.join(timeout=6) + server.server_close() + + +@dataclass(frozen=True, slots=True) +class Rig: + gateway: Gateway + api: Wire + gcs: Wire + + +@pytest.fixture(scope="module") +def rig(tmp_path_factory: pytest.TempPathFactory) -> Iterator[Rig]: + directory: Final = tmp_path_factory.mktemp("vertex-utc") + certificate, key = write_self_signed_cert(directory, names=("storage.googleapis.com",)) + with ( + gateway_from_environment() as environment, + wire_server(vertex_peer) as api, + wire_server(vertex_peer, tls=server_context(certificate, key)) as gcs, + gcs_tunnel(gcs) as tunnel, + owned_proxy( + environment, + directory, + { + "TZ": HOST_ZONE, + "HTTPS_PROXY": tunnel, + "https_proxy": tunnel, + "SSL_CERT_FILE": str(certificate), + "NO_PROXY": "127.0.0.1,localhost", + "no_proxy": "127.0.0.1,localhost", + }, + config=_config_allowing(directory, api), + ) as candidate, + ): + yield Rig(candidate, api, gcs) + + +def _config_allowing(directory: Path, api: Wire) -> Path: + shipped: Final = yaml.safe_load(Path("tests/integration/proxy_config.yaml").read_text()) + configuration: Final = { + **shipped, + "litellm_settings": {**shipped["litellm_settings"], "user_url_allowed_hosts": [urlsplit(api.url).netloc]}, + } + path: Final = directory / "vertex-utc.yaml" + path.write_text(yaml.safe_dump(configuration)) + return path + + +def _vertex_model(rig: Rig, scenario: Scenario, model: str) -> str: + return scenario.model( + model=model, + api_key=None, + api_base=rig.api.url, + vertex_project=PROJECT, + vertex_location=LOCATION, + vertex_credentials=service_account_json(PROJECT, rig.api.url), + gcs_bucket_name=BUCKET, + ) + + +def _object(response_content: bytes) -> dict[str, JsonValue]: + return JSON_OBJECT.validate_json(response_content) + + +def test_managed_vertex_batch_and_output_file_created_at_are_utc_epochs(rig: Rig) -> None: + with rig.gateway.scenario() as scenario: + model: Final = _vertex_model(rig, scenario, "vertex_ai/gemini-2.5-flash") + owner_key: Final = scenario.key(user_id=scenario.user(user_role="internal_user"), models=[model]) + line: Final = { + "custom_id": "req-1", + "method": "POST", + "url": "/v1/chat/completions", + "body": {"model": model, "messages": [{"role": "user", "content": "ping"}], "max_tokens": 8}, + } + uploaded: Final = rig.gateway.request_multipart( + "/v1/files", + {"purpose": "batch", "target_model_names": model}, + {"file": ("input.jsonl", (json.dumps(line) + "\n").encode(), "application/jsonl")}, + key=owner_key, + ) + assert uploaded.status_code == 200, uploaded.text + upload_object: Final = _object(uploaded.content) + + created: Final = rig.gateway.request( + "POST", + "/v1/batches", + { + "input_file_id": string_value(upload_object["id"]), + "endpoint": "/v1/chat/completions", + "completion_window": "24h", + }, + key=owner_key, + ) + assert created.status_code == 200, created.text + batch: Final = _object(created.content) + + retrieved: Final = rig.gateway.request("GET", f"/v1/batches/{string_value(batch['id'])}", key=owner_key) + assert retrieved.status_code == 200, retrieved.text + completed: Final = _object(retrieved.content) + assert completed["status"] == "completed", completed + + output_id: Final = string_value(completed["output_file_id"]) + details: Final = rig.gateway.request("GET", f"/v1/files/{output_id}", key=owner_key) + assert details.status_code == 200, details.text + output: Final = _object(details.content) + assert (output["id"], output["bytes"]) == (output_id, OUTPUT_BYTES), output + gcs_reads: Final = tuple( + request for request in rig.gcs.drain() if request.method == "GET" and "/storage/v1/b/" in request.target + ) + assert gcs_reads, "The output file details never came from the GCS object metadata" + + listed: Final = rig.gateway.request("GET", "/v1/files", key=owner_key) + assert listed.status_code == 200, listed.text + listed_files: Final = _object(listed.content)["data"] + assert isinstance(listed_files, list), listed.text + listed_output: Final = next( + ( + JSON_OBJECT.validate_python(file) + for file in listed_files + if isinstance(file, dict) and file.get("id") == output_id + ), + None, + ) + assert listed_output is not None, f"Output file {output_id} is absent from GET /v1/files: {listed.text}" + + assert { + "POST /v1/files": upload_object["created_at"], + "POST /v1/batches": batch["created_at"], + "GET /v1/batches/{id}": completed["created_at"], + "GET /v1/files/{output}": output["created_at"], + "GET /v1/files": listed_output["created_at"], + } == { + "POST /v1/files": UPLOAD_EPOCH, + "POST /v1/batches": BATCH_EPOCH, + "GET /v1/batches/{id}": BATCH_EPOCH, + "GET /v1/files/{output}": OUTPUT_EPOCH, + "GET /v1/files": OUTPUT_EPOCH, + } + + +def test_managed_vertex_batch_with_whole_second_create_time_is_created_and_retrieved(rig: Rig) -> None: + with rig.gateway.scenario() as scenario: + model: Final = _vertex_model(rig, scenario, f"vertex_ai/{WHOLE_SECOND_MODEL}") + owner_key: Final = scenario.key(user_id=scenario.user(user_role="internal_user"), models=[model]) + line: Final = { + "custom_id": "req-1", + "method": "POST", + "url": "/v1/chat/completions", + "body": {"model": model, "messages": [{"role": "user", "content": "ping"}], "max_tokens": 8}, + } + uploaded: Final = rig.gateway.request_multipart( + "/v1/files", + {"purpose": "batch", "target_model_names": model}, + {"file": ("input.jsonl", (json.dumps(line) + "\n").encode(), "application/jsonl")}, + key=owner_key, + ) + assert uploaded.status_code == 200, uploaded.text + created: Final = rig.gateway.request( + "POST", + "/v1/batches", + { + "input_file_id": string_value(_object(uploaded.content)["id"]), + "endpoint": "/v1/chat/completions", + "completion_window": "24h", + }, + key=owner_key, + ) + assert created.status_code == 200, created.text + batch: Final = _object(created.content) + retrieved: Final = rig.gateway.request("GET", f"/v1/batches/{string_value(batch['id'])}", key=owner_key) + assert retrieved.status_code == 200, retrieved.text + assert (batch["created_at"], _object(retrieved.content)["created_at"]) == (BATCH_EPOCH, BATCH_EPOCH) + + +def test_veo_video_status_created_at_is_a_utc_epoch(rig: Rig) -> None: + with rig.gateway.scenario() as scenario: + model: Final = _vertex_model(rig, scenario, f"vertex_ai/{VEO_MODEL}") + created: Final = rig.gateway.request("POST", "/v1/videos", {"model": model, "prompt": "a scripted cat"}) + assert created.status_code == 200, created.text + video_id: Final = string_value(_object(created.content)["id"]) + status: Final = rig.gateway.request("GET", f"/v1/videos/{video_id}") + assert status.status_code == 200, status.text + video: Final = _object(status.content) + assert (video["status"], video["created_at"]) == ("processing", VEO_EPOCH), video diff --git a/tests/unit/decisions/test_main.py b/tests/unit/decisions/test_main.py index ebe0eedea7f..81f5b60a036 100644 --- a/tests/unit/decisions/test_main.py +++ b/tests/unit/decisions/test_main.py @@ -163,7 +163,7 @@ async def test_adecisions_sends_the_provider_wire_contract( ) -> None: route: Final = respx_mock.post(url).respond(json=_RESPONSE) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model=model, state={"source": "unit-test"}, questions=_QUESTIONS, @@ -225,7 +225,7 @@ async def test_router_dispatches_typesafe_decisions_without_api_base( ) upstream: Final = respx_mock.post("https://api.typesafe.ai/v1/systemone").respond(json=_RESPONSE) - response: Final = await router.adecisions( + response: Final = await router.asystemone( model="jev", state="router-test", questions={ @@ -256,7 +256,7 @@ async def test_router_dispatches_typesafe_decisions_without_api_base( def test_decisions_uses_the_same_wire_contract_for_sync_calls(respx_mock: respx.MockRouter) -> None: route: Final = respx_mock.post("https://api.perplexity.ai/v1/decisions").respond(json=_RESPONSE) - response: Final = litellm.decisions( + response: Final = litellm.systemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -276,7 +276,7 @@ def test_openrouter_response_keeps_provider_fields(respx_mock: respx.MockRouter) } respx_mock.post("https://openrouter.ai/api/alpha/decisions").respond(json=payload) - response: Final = litellm.decisions( + response: Final = litellm.systemone( model="openrouter/typesafe/jev-1.13", state="review", questions=_QUESTIONS, @@ -364,7 +364,7 @@ async def test_decisions_cost_is_in_standard_logging_object(respx_mock: respx.Mo litellm.callbacks = [recording_logger] try: - await litellm.adecisions( + await litellm.asystemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -389,7 +389,7 @@ async def test_decisions_cost_is_in_standard_logging_object(respx_mock: respx.Mo @pytest.mark.asyncio async def test_unknown_provider_is_rejected_before_http(respx_mock: respx.MockRouter) -> None: with pytest.raises(litellm.BadRequestError, match="LLM Provider NOT provided"): - await litellm.adecisions( + await litellm.asystemone( model="unknown/jev-1.13", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -402,7 +402,7 @@ async def test_unknown_provider_is_rejected_before_http(respx_mock: respx.MockRo @pytest.mark.asyncio async def test_provider_without_decisions_support_is_rejected_before_http(respx_mock: respx.MockRouter) -> None: with pytest.raises(litellm.BadRequestError, match=r"Unknown Decisions provider 'anthropic'\. Supported providers"): - await litellm.adecisions( + await litellm.asystemone( model="anthropic/claude-sonnet-4-5", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -416,7 +416,7 @@ async def test_provider_without_decisions_support_is_rejected_before_http(respx_ async def test_empty_custom_provider_falls_back_to_the_model_prefix(respx_mock: respx.MockRouter) -> None: route: Final = respx_mock.post("https://api.perplexity.ai/v1/decisions").respond(json=_RESPONSE) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -436,7 +436,7 @@ async def test_upstream_reply_without_answers_is_a_server_error(respx_mock: resp ) with pytest.raises(litellm.InternalServerError, match="unexpected response"): - await litellm.adecisions( + await litellm.asystemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -464,7 +464,7 @@ async def test_router_sends_the_openrouter_deployment_key_when_the_provider_is_a ) upstream: Final = respx_mock.post("https://egress.example/openrouter/alpha/decisions").respond(json=_RESPONSE) - await router.adecisions( + await router.asystemone( model="jev", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -478,7 +478,7 @@ async def test_router_sends_the_openrouter_deployment_key_when_the_provider_is_a @pytest.mark.asyncio async def test_invalid_question_is_rejected_before_http(respx_mock: respx.MockRouter) -> None: with pytest.raises(litellm.BadRequestError, match="Invalid Decisions request"): - await litellm.adecisions( + await litellm.asystemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"sentiment": {"type": "choice"}}, @@ -495,7 +495,7 @@ def test_upstream_bad_request_maps_to_litellm_error(respx_mock: respx.MockRouter ) with pytest.raises(litellm.BadRequestError): - litellm.decisions( + litellm.systemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -509,7 +509,7 @@ def test_unreachable_upstream_maps_to_a_connection_error(respx_mock: respx.MockR ) with pytest.raises(litellm.APIConnectionError, match="PerplexityException - Cannot connect to host"): - litellm.decisions( + litellm.systemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -525,7 +525,7 @@ def test_server_key_is_sent_to_an_explicit_api_base( monkeypatch.delenv("PERPLEXITY_API_KEY", raising=False) route: Final = respx_mock.post("https://egress.example/perplexity/v1/decisions").respond(json=_RESPONSE) - litellm.decisions( + litellm.systemone( model="perplexity/pplx-decider-v1-27b", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -555,7 +555,7 @@ async def test_cloudflare_clef_resolves_model_and_response_envelope( "https://api.cloudflare.com/client/v4/accounts/acct/ai/run/@cf/cloudflare/clef" ).respond(json=response_body) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model=model, state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -585,7 +585,7 @@ async def test_cloudflare_clef_flash_uses_flash_endpoint_and_request_model( "https://api.cloudflare.com/client/v4/accounts/acct/ai/run/@cf/cloudflare/clef-flash" ).respond(json=_RESPONSE) - await litellm.adecisions( + await litellm.asystemone( model="cloudflare/clef-flash", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -607,7 +607,7 @@ async def test_cloudflare_api_base_from_env_uses_workers_ai_run_path( "https://api.cloudflare.com/client/v4/accounts/acct/ai/run/@cf/cloudflare/clef" ).respond(json=_RESPONSE) - await litellm.adecisions( + await litellm.asystemone( model="cloudflare/clef", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -626,7 +626,7 @@ async def test_cloudflare_requires_account_id_or_api_base_before_http( monkeypatch.setenv("CLOUDFLARE_API_KEY", "cloudflare-key") with pytest.raises(litellm.BadRequestError, match="Missing CLOUDFLARE_ACCOUNT_ID - set CLOUDFLARE_ACCOUNT_ID"): - await litellm.adecisions( + await litellm.asystemone( model="cloudflare/clef", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -647,7 +647,7 @@ async def test_cloudflare_clef_cost_uses_the_model_cost_map( json=_RESPONSE ) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model="cloudflare/clef", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -672,7 +672,7 @@ async def test_strands_decider_requires_api_base_before_http( monkeypatch.delenv("STRANDS_DECIDER_API_KEY", raising=False) with pytest.raises(litellm.BadRequestError, match="api_base is required"): - await litellm.adecisions( + await litellm.asystemone( model="strands_decider/strands-decider-2B-hobson-v19", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -690,7 +690,7 @@ async def test_strands_decider_without_key_preserves_response_extras( monkeypatch.delenv("STRANDS_DECIDER_API_KEY", raising=False) route: Final = respx_mock.post("https://strands.example/v1/systemone").respond(json=_STRANDS_RESPONSE) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model="strands_decider/strands-decider-2B-hobson-v19", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -717,7 +717,7 @@ async def test_strands_decider_plain_api_base_answers_while_every_aws_signing_th try: response: Final = await asyncio.wait_for( - litellm.adecisions( + litellm.asystemone( model="strands_decider/strands-decider-2B-hobson-v19", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -741,7 +741,7 @@ async def test_strands_decider_uses_key_from_matching_environment_base( monkeypatch.setenv("STRANDS_DECIDER_API_KEY", "strands-key") route: Final = respx_mock.post("https://strands.example/v1/systemone").respond(json=_STRANDS_RESPONSE) - await litellm.adecisions( + await litellm.asystemone( model="strands_decider/strands-decider-2B-hobson-v19", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -773,7 +773,7 @@ async def test_strands_decider_provider_resolution_and_router_dispatch( ) route: Final = respx_mock.post("https://strands.example/v1/systemone").respond(json=_STRANDS_RESPONSE) - response: Final = await router.adecisions( + response: Final = await router.asystemone( model="strands", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -827,7 +827,7 @@ async def test_strands_decider_runtime_arn_sends_a_sigv4_signed_invoke_agent_run "aws_secret_access_key": _RUNTIME_CREDENTIALS.secret_key, } - response: Final = await litellm.adecisions(**call_args) if use_async else litellm.decisions(**call_args) + response: Final = await litellm.asystemone(**call_args) if use_async else litellm.systemone(**call_args) assert route.call_count == 1 sent: Final = route.calls[0].request @@ -888,7 +888,7 @@ async def test_strands_decider_runtime_arn_with_api_key_sends_the_bearer_token_u monkeypatch.delenv("STRANDS_DECIDER_API_KEY", raising=False) route: Final = respx_mock.post(_RUNTIME_URL).respond(json=_STRANDS_RESPONSE) - await litellm.adecisions( + await litellm.asystemone( model="strands_decider/strands-decider-2B-hobson-v19", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -921,7 +921,7 @@ async def test_strands_decider_runtime_error_envelope_raises_instead_of_answerin respx_mock.post(_RUNTIME_URL).respond(json={"error": {"code": code, "message": "at most 16 questions"}}) with pytest.raises(error_class, match="at most 16 questions"): - await litellm.adecisions( + await litellm.asystemone( model="strands_decider/strands-decider-2B-hobson-v19", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -942,7 +942,7 @@ async def test_strands_decider_plain_api_base_error_envelope_stays_an_unexpected ) with pytest.raises(litellm.InternalServerError, match="returned an unexpected response"): - await litellm.adecisions( + await litellm.asystemone( model="strands_decider/strands-decider-2B-hobson-v19", state="review", questions={"severity": {"type": "score", "criteria": ["none", "low", "high"]}}, @@ -977,7 +977,7 @@ async def test_hosted_vllm_requires_api_base_before_http( monkeypatch.delenv("HOSTED_VLLM_API_KEY", raising=False) with pytest.raises(litellm.BadRequestError, match="api_base is required"): - await litellm.adecisions( + await litellm.asystemone( model="hosted_vllm/Qwen/Qwen3-0.6B", state="review", questions={"is_defect": {"type": "choice", "criteria": {"yes": None, "no": None}}}, @@ -995,7 +995,7 @@ async def test_hosted_vllm_without_key_sends_no_authorization_and_preserves_resp monkeypatch.delenv("HOSTED_VLLM_API_KEY", raising=False) route: Final = respx_mock.post("http://vllm.local:8000/v1/systemone").respond(json=_VLLM_RESPONSE) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model="hosted_vllm/Qwen/Qwen3-0.6B", state="The package arrived broken.", questions={"is_defect": {"type": "choice", "criteria": {"yes": None, "no": None}}}, @@ -1018,7 +1018,7 @@ async def test_hosted_vllm_uses_key_and_base_from_environment( monkeypatch.setenv("HOSTED_VLLM_API_KEY", "vllm-key") route: Final = respx_mock.post("http://vllm.local:8000/v1/systemone").respond(json=_VLLM_RESPONSE) - await litellm.adecisions( + await litellm.asystemone( model="hosted_vllm/Qwen/Qwen3-0.6B", state="The package arrived broken.", questions={"is_defect": {"type": "choice", "criteria": {"yes": None, "no": None}}}, @@ -1037,7 +1037,7 @@ async def test_hosted_vllm_api_base_with_v1_suffix_posts_to_v1_systemone( monkeypatch.delenv("HOSTED_VLLM_API_KEY", raising=False) route: Final = respx_mock.post("http://vllm.local:8000/v1/systemone").respond(json=_VLLM_RESPONSE) - await litellm.adecisions( + await litellm.asystemone( model="hosted_vllm/Qwen/Qwen3-0.6B", state="The package arrived broken.", questions={"is_defect": {"type": "choice", "criteria": {"yes": None, "no": None}}}, @@ -1056,7 +1056,7 @@ async def test_hosted_vllm_choice_question_goes_out_as_a_jev_choice_with_the_ser monkeypatch.delenv("HOSTED_VLLM_API_KEY", raising=False) route: Final = respx_mock.post("http://vllm.local:8000/v1/systemone").respond(json=_VLLM_RESPONSE) - await litellm.adecisions( + await litellm.asystemone( model="hosted_vllm/Qwen/Qwen3-0.6B", state="The package arrived broken.", questions={ @@ -1092,7 +1092,7 @@ async def test_databricks_requires_api_base_before_http( monkeypatch.setenv("DATABRICKS_API_KEY", "dapi-key") with pytest.raises(litellm.BadRequestError, match="DATABRICKS_API_BASE"): - await litellm.adecisions( + await litellm.asystemone( model=f"databricks/{_DATABRICKS_ENDPOINT}", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -1114,7 +1114,7 @@ async def test_databricks_rejects_a_serving_endpoint_name_that_would_rewrite_the monkeypatch.setenv("DATABRICKS_API_KEY", "dapi-key") with pytest.raises(litellm.BadRequestError, match="bare endpoint name"): - await litellm.adecisions( + await litellm.asystemone( model=f"databricks/{endpoint_name}", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -1133,7 +1133,7 @@ async def test_databricks_requires_a_key_before_http( monkeypatch.delenv("DATABRICKS_TOKEN", raising=False) with pytest.raises(litellm.AuthenticationError, match="Missing API key for Decisions provider 'databricks'"): - await litellm.adecisions( + await litellm.asystemone( model=f"databricks/{_DATABRICKS_ENDPOINT}", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -1162,7 +1162,7 @@ async def test_databricks_posts_the_jev_body_to_the_serving_endpoint_invocations "sentiment": {"type": "choice", "criteria": {"positive": None, "negative": "unhappy"}}, } - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model=f"databricks/{_DATABRICKS_ENDPOINT}", state={"ticket": "export hangs"}, questions=questions, @@ -1206,7 +1206,7 @@ async def test_databricks_router_deployment_sends_its_own_key_to_its_own_base( f"https://workspace.example/serving-endpoints/{_DATABRICKS_ENDPOINT}/invocations" ).respond(json=_DATABRICKS_RESPONSE) - response: Final = await router.adecisions( + response: Final = await router.asystemone( model="decider", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -1260,7 +1260,7 @@ async def test_openai_decisions_translate_systemone_to_the_openai_wire_contract_ monkeypatch.delenv("OPENAI_API_BASE", raising=False) route: Final = respx_mock.post(url).respond(json=_OPENAI_RESPONSE) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model="openai/gpt-6-luna", state="The package arrived broken.", questions={ @@ -1345,7 +1345,7 @@ def test_openai_decisions_use_the_same_settings_as_other_openai_calls( monkeypatch.setenv(name, value) route: Final = respx_mock.post(url).respond(json=_OPENAI_RESPONSE) - litellm.decisions( + litellm.systemone( model="openai/gpt-6-luna", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -1408,17 +1408,8 @@ async def test_openai_format_calls_to_a_systemone_provider_get_openai_format_ans }, "cannot serve this request", ), - ( - "openai/gpt-6-luna", - { - "state": "review", - "input": "review", - "questions": [{"type": "predicate", "instructions": "Is this a defect?"}], - }, - "not both", - ), ), - ids=("image_to_systemone_provider", "state_and_input"), + ids=("image_to_systemone_provider",), ) async def test_requests_a_provider_cannot_serve_are_rejected_before_http( respx_mock: respx.MockRouter, @@ -1451,7 +1442,7 @@ async def test_openrouter_decisions_uses_provider_reported_cost_without_cost_map } ) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model="openrouter/typesafe/jev-1.13", state={"source": "unit-test"}, questions=_QUESTIONS, @@ -1466,24 +1457,27 @@ _SAFETY_IDENTIFIER: Final = "end-user-7" _PREDICATE_QUESTIONS: Final[tuple[Mapping[str, object], ...]] = ( {"type": "predicate", "name": "is_defect", "instructions": "Is this a defect?"}, ) -_PREDICATE_REQUESTS: Final[tuple[Mapping[str, object], ...]] = ( - MappingProxyType({"input": "review", "questions": _PREDICATE_QUESTIONS}), - MappingProxyType( - {"state": "review", "questions": {"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}} +_PREDICATE_REQUESTS: Final[tuple[tuple[str, Mapping[str, object]], ...]] = ( + ("decisions", MappingProxyType({"input": "review", "questions": _PREDICATE_QUESTIONS})), + ( + "systemone", + MappingProxyType( + {"state": "review", "questions": {"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}} + ), ), ) -_PREDICATE_REQUEST_IDS: Final = ("input_format", "state_format") +_PREDICATE_REQUEST_IDS: Final = ("decisions", "systemone") -@pytest.mark.parametrize("request_kwargs", _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) +@pytest.mark.parametrize(("api", "request_kwargs"), _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) def test_safety_identifier_is_refused_before_http_when_the_provider_cannot_take_it( - request_kwargs: Mapping[str, object], respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch + api: str, request_kwargs: Mapping[str, object], respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch ) -> None: monkeypatch.setattr(litellm, "drop_params", False) route: Final = respx_mock.post("https://api.perplexity.ai/v1/decisions").respond(json=_RESPONSE) with pytest.raises(litellm.UnsupportedParamsError, match=r"safety_identifier.*drop_params") as caught: - litellm.decisions( + getattr(litellm, api)( model="perplexity/pplx-decider-v1-27b", safety_identifier=_SAFETY_IDENTIFIER, api_key="caller-key", @@ -1524,10 +1518,11 @@ async def test_safety_identifier_is_dropped_from_the_wire_under_drop_params( @pytest.mark.asyncio -@pytest.mark.parametrize("request_kwargs", _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) +@pytest.mark.parametrize(("api", "request_kwargs"), _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) @pytest.mark.parametrize("drop_params", (False, True), ids=("strict", "drop_params")) async def test_safety_identifier_reaches_openai_whether_or_not_params_are_dropped( drop_params: bool, + api: str, request_kwargs: Mapping[str, object], respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch, @@ -1538,7 +1533,7 @@ async def test_safety_identifier_reaches_openai_whether_or_not_params_are_droppe monkeypatch.delenv("OPENAI_API_BASE", raising=False) route: Final = respx_mock.post("https://api.openai.com/v1/decisions").respond(json=_OPENAI_RESPONSE) - await litellm.adecisions( + await getattr(litellm, f"a{api}")( model="openai/gpt-6-luna", safety_identifier=_SAFETY_IDENTIFIER, api_key="caller-key", @@ -1555,16 +1550,16 @@ async def test_safety_identifier_reaches_openai_whether_or_not_params_are_droppe @pytest.mark.asyncio -@pytest.mark.parametrize("request_kwargs", _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) +@pytest.mark.parametrize(("api", "request_kwargs"), _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) async def test_non_string_safety_identifier_is_rejected_before_http( - request_kwargs: Mapping[str, object], respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch + api: str, request_kwargs: Mapping[str, object], respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch ) -> None: monkeypatch.setattr(litellm, "drop_params", False) route: Final = respx_mock.post("https://api.openai.com/v1/decisions").respond(json=_OPENAI_RESPONSE) invalid_safety_identifier: Final[Mapping[str, object]] = MappingProxyType({"safety_identifier": 7}) with pytest.raises(litellm.BadRequestError, match=r"(?s)Invalid Decisions request.*safety_identifier"): - await litellm.adecisions( + await getattr(litellm, f"a{api}")( model="openai/gpt-6-luna", api_key="caller-key", **request_kwargs, **invalid_safety_identifier ) @@ -1572,9 +1567,9 @@ async def test_non_string_safety_identifier_is_rejected_before_http( @pytest.mark.asyncio -@pytest.mark.parametrize("request_kwargs", _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) +@pytest.mark.parametrize(("api", "request_kwargs"), _PREDICATE_REQUESTS, ids=_PREDICATE_REQUEST_IDS) async def test_non_string_safety_identifier_is_dropped_from_the_wire_under_drop_params( - request_kwargs: Mapping[str, object], respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch + api: str, request_kwargs: Mapping[str, object], respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch ) -> None: monkeypatch.setattr(litellm, "drop_params", False) monkeypatch.setattr(litellm, "api_base", None) @@ -1585,7 +1580,7 @@ async def test_non_string_safety_identifier_is_dropped_from_the_wire_under_drop_ {"safety_identifier": 7, "drop_params": True} ) - await litellm.adecisions( + await getattr(litellm, f"a{api}")( model="openai/gpt-6-luna", api_key="caller-key", **request_kwargs, **dropped_safety_identifier ) @@ -1622,7 +1617,7 @@ async def test_azure_ai_decision_posts_deployment_to_foundry_systemone_route( json={**_RESPONSE, "model": "microsoft-decision-1"} ) - response: Final = await litellm.adecisions( + response: Final = await litellm.asystemone( model="azure_ai/decision-1", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -1654,7 +1649,7 @@ async def test_azure_ai_decision_reads_foundry_env_and_keeps_gateway_path_prefix json=_RESPONSE ) - await litellm.adecisions( + await litellm.asystemone( model="azure_ai/decision-1", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, @@ -1673,7 +1668,7 @@ async def test_azure_ai_decision_requires_api_base_before_http( monkeypatch.setenv("AZURE_AI_API_KEY", "foundry-key") with pytest.raises(litellm.BadRequestError, match="Missing AZURE_AI_API_BASE"): - await litellm.adecisions( + await litellm.asystemone( model="azure_ai/decision-1", state="review", questions={"is_defect": {"type": "noul", "instructions": "Is this a defect?"}}, diff --git a/tests/unit/llms/azure/test_azure_common_utils.py b/tests/unit/llms/azure/test_azure_common_utils.py index 302175c46f8..b277cac1f29 100644 --- a/tests/unit/llms/azure/test_azure_common_utils.py +++ b/tests/unit/llms/azure/test_azure_common_utils.py @@ -471,6 +471,7 @@ def test_default_max_retries_env_var_reaches_azure_sdk_client(): "llm_passthrough_route", "asearch", "adecisions", + "asystemone", "avector_store_create", "avector_store_search", "acreate_skill", diff --git a/tests/unit/llms/vertex_ai/batches/test_transformation.py b/tests/unit/llms/vertex_ai/batches/test_transformation.py index cf5e11b35b1..064845a7266 100644 --- a/tests/unit/llms/vertex_ai/batches/test_transformation.py +++ b/tests/unit/llms/vertex_ai/batches/test_transformation.py @@ -22,10 +22,7 @@ import pytest from litellm.llms.vertex_ai.batches.transformation import ( # noqa: E402 VertexAIBatchTransformation, ) -from litellm.llms.vertex_ai.common_utils import ( # noqa: E402 - VertexAIError, - convert_vertex_datetime_to_openai_datetime, -) +from litellm.llms.vertex_ai.common_utils import VertexAIError # noqa: E402 from litellm.types.utils import LiteLLMBatch # noqa: E402 T = VertexAIBatchTransformation @@ -142,10 +139,7 @@ def test_transform_vertex_response_full_mapping(): assert isinstance(batch, LiteLLMBatch) assert batch.id == "3814889423749775360" assert batch.completion_window == "24h" - # created_at is parsed via the shared helper (uses local tz); assert the - # transform forwards createTime through that helper rather than a hardcoded - # epoch that would be tz-dependent - assert batch.created_at == convert_vertex_datetime_to_openai_datetime("2024-12-04T21:53:12.120184Z") + assert batch.created_at == 1733349192 assert batch.endpoint == "" assert batch.object == "batch" assert batch.input_file_id == "gs://bucket/in.jsonl" diff --git a/tests/unit/llms/vertex_ai/test_vertex_ai_common_utils.py b/tests/unit/llms/vertex_ai/test_vertex_ai_common_utils.py index 8407007e1d4..c0b1173db29 100644 --- a/tests/unit/llms/vertex_ai/test_vertex_ai_common_utils.py +++ b/tests/unit/llms/vertex_ai/test_vertex_ai_common_utils.py @@ -1,3 +1,5 @@ +import time +from collections.abc import Iterator from unittest.mock import patch import pytest @@ -5,6 +7,7 @@ import pytest from litellm.constants import DEFAULT_MAX_RECURSE_DEPTH from litellm.llms.vertex_ai.common_utils import ( convert_anyof_null_to_nullable, + convert_vertex_datetime_to_openai_datetime, get_vertex_location_from_url, get_vertex_project_id_from_url, get_vertex_url, @@ -1882,3 +1885,31 @@ def test_get_vertex_ai_lyria_model_info_is_none_for_non_lyria_speech_models(mode assert get_vertex_ai_lyria_model_info(model=model) is None +@pytest.fixture +def host_zone(request: pytest.FixtureRequest, monkeypatch: pytest.MonkeyPatch) -> Iterator[None]: + monkeypatch.setenv("TZ", request.param) + time.tzset() + yield + monkeypatch.undo() + time.tzset() + + +@pytest.mark.parametrize( + "host_zone", ["UTC", "America/Los_Angeles", "Asia/Kolkata", "Pacific/Kiritimati"], indirect=True +) +@pytest.mark.parametrize( + ("vertex_datetime", "utc_epoch"), + [ + ("2024-12-04T21:53:12.120184Z", 1733349192), + ("2026-10-08T23:41:07.902Z", 1791502867), + ("2026-01-15T08:00:00.000000Z", 1768464000), + ("2026-09-09T21:39:16Z", 1788989956), + ("2026-10-09T08:19:53.576944123Z", 1791533993), + ], +) +@pytest.mark.usefixtures("host_zone") +@pytest.mark.skipif(not hasattr(time, "tzset"), reason="switching the process timezone needs time.tzset()") +def test_convert_vertex_datetime_to_openai_datetime_is_the_utc_epoch_on_any_host_zone( + vertex_datetime: str, utc_epoch: int +) -> None: + assert convert_vertex_datetime_to_openai_datetime(vertex_datetime) == utc_epoch diff --git a/tests/unit/proxy/guardrails/guardrail_hooks/test_decision_model.py b/tests/unit/proxy/guardrails/guardrail_hooks/test_decision_model.py index 1c2b297f238..54a73690840 100644 --- a/tests/unit/proxy/guardrails/guardrail_hooks/test_decision_model.py +++ b/tests/unit/proxy/guardrails/guardrail_hooks/test_decision_model.py @@ -414,7 +414,7 @@ def test_initialize_rejects_invalid_streaming_scan_settings(mock_mgr, overrides, def test_add_guardrail_settings_returns_decisions_providers(): import asyncio - from litellm.decisions.main import supported_decisions_providers + from litellm.decisions.call import supported_decisions_providers from litellm.proxy.guardrails.guardrail_endpoints import get_guardrail_ui_settings settings: Final = asyncio.run(get_guardrail_ui_settings()) diff --git a/tests/unit/systemone/__init__.py b/tests/unit/systemone/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/tests/unit/systemone/test_main.py b/tests/unit/systemone/test_main.py new file mode 100644 index 00000000000..342b7f1ef97 --- /dev/null +++ b/tests/unit/systemone/test_main.py @@ -0,0 +1,138 @@ +from __future__ import annotations + +import asyncio +import json +from collections.abc import Mapping +from types import MappingProxyType +from typing import Final + +import pytest +import respx +from respx.models import Call + +import litellm +from litellm.integrations.custom_logger import CustomLogger +from litellm.litellm_core_utils.logging_worker import GLOBAL_LOGGING_WORKER +from litellm.types.decisions import DecisionsResponse, OpenAIDecisionResponse + +_STATE: Final = "Customer asked for a refund" +_SYSTEMONE_QUESTIONS: Final[Mapping[str, object]] = MappingProxyType( + {"is_refund": {"type": "noul", "instructions": "Is this a refund request?"}} +) +_OPENAI_QUESTIONS: Final[tuple[Mapping[str, object], ...]] = ( + {"type": "predicate", "name": "is_refund", "instructions": "Is this a refund request?"}, +) +_SYSTEMONE_RESPONSE: Final[Mapping[str, object]] = { + "model": "jev-1.13", + "answers": {"is_refund": {"type": "noul", "noul": 0.93}}, + "usage": {"input_tokens": 12, "output_tokens": 1}, +} +_TYPESAFE_URL: Final = "https://api.typesafe.ai/v1/systemone" + + +@pytest.fixture(autouse=True) +def _httpx_transport(monkeypatch: pytest.MonkeyPatch) -> None: + monkeypatch.setattr(litellm, "disable_aiohttp_transport", True) + litellm.in_memory_llm_clients_cache.flush_cache() + + +class _CallTypeRecorder(CustomLogger): + def __init__(self) -> None: + super().__init__() + self.call_types: list[str] = [] # mutable-ok: test recorder collects logger callbacks + + async def async_log_success_event(self, kwargs, response_obj, start_time, end_time) -> None: + self.call_types.append(kwargs["call_type"]) + + +@pytest.mark.asyncio +@pytest.mark.parametrize("use_async", (True, False), ids=("asystemone", "systemone")) +async def test_systemone_sends_the_systemone_wire_format_and_returns_a_systemone_response( + use_async: bool, respx_mock: respx.MockRouter +) -> None: + route: Final = respx_mock.post(_TYPESAFE_URL).respond(json=_SYSTEMONE_RESPONSE) + kwargs: Final[Mapping[str, object]] = MappingProxyType( + {"model": "typesafe/jev-1.13", "state": _STATE, "questions": _SYSTEMONE_QUESTIONS, "api_key": "caller-key"} + ) + + response: Final = await litellm.asystemone(**kwargs) if use_async else litellm.systemone(**kwargs) + + assert route.called + call: Final[Call] = respx_mock.calls.last + assert json.loads(call.request.content) == { + "model": "jev-1.13", + "state": _STATE, + "questions": {"is_refund": {"type": "noul", "instructions": "Is this a refund request?"}}, + } + assert isinstance(response, DecisionsResponse) + assert response.answers["is_refund"].model_dump(mode="json") == {"type": "noul", "noul": 0.93} + + +@pytest.mark.asyncio +async def test_systemone_rejects_openai_input_and_points_at_decisions(respx_mock: respx.MockRouter) -> None: + route: Final = respx_mock.post(_TYPESAFE_URL).respond(json=_SYSTEMONE_RESPONSE) + + with pytest.raises(litellm.BadRequestError, match=r"litellm\.decisions") as caught: + await litellm.asystemone( + model="typesafe/jev-1.13", input=_STATE, questions=_SYSTEMONE_QUESTIONS, api_key="caller-key" + ) + + assert caught.value.status_code == 400 + assert not route.called + + +@pytest.mark.asyncio +@pytest.mark.parametrize("use_async", (True, False), ids=("adecisions", "decisions")) +async def test_decisions_with_state_still_sends_the_systemone_format_and_warns( + use_async: bool, respx_mock: respx.MockRouter +) -> None: + route: Final = respx_mock.post(_TYPESAFE_URL).respond(json=_SYSTEMONE_RESPONSE) + kwargs: Final[Mapping[str, object]] = MappingProxyType( + {"model": "typesafe/jev-1.13", "state": _STATE, "questions": _SYSTEMONE_QUESTIONS, "api_key": "caller-key"} + ) + + with pytest.warns(UserWarning, match=r"litellm\.systemone"): + response: Final = await litellm.adecisions(**kwargs) if use_async else litellm.decisions(**kwargs) + + assert route.called + call: Final[Call] = respx_mock.calls.last + assert json.loads(call.request.content) == { + "model": "jev-1.13", + "state": _STATE, + "questions": {"is_refund": {"type": "noul", "instructions": "Is this a refund request?"}}, + } + assert isinstance(response, DecisionsResponse) + assert response.answers["is_refund"].model_dump(mode="json") == {"type": "noul", "noul": 0.93} + + +@pytest.mark.asyncio +async def test_systemone_rejects_openai_question_lists_before_http(respx_mock: respx.MockRouter) -> None: + route: Final = respx_mock.post(_TYPESAFE_URL).respond(json=_SYSTEMONE_RESPONSE) + + with pytest.raises(litellm.BadRequestError, match="Invalid Decisions request"): + await litellm.asystemone( + model="typesafe/jev-1.13", state=_STATE, questions=_OPENAI_QUESTIONS, api_key="caller-key" + ) + + assert not route.called + + +@pytest.mark.asyncio +async def test_router_asystemone_logs_the_systemone_call_type( + respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch +) -> None: + respx_mock.post(_TYPESAFE_URL).respond(json=_SYSTEMONE_RESPONSE) + recorder: Final = _CallTypeRecorder() + monkeypatch.setattr(litellm, "callbacks", [recorder]) + router: Final = litellm.Router( + model_list=[{"model_name": "jev", "litellm_params": {"model": "typesafe/jev-1.13", "api_key": "caller-key"}}] + ) + + response: Final = await router.asystemone(model="jev", state=_STATE, questions=_SYSTEMONE_QUESTIONS) + await asyncio.sleep(0) + GLOBAL_LOGGING_WORKER.start() + await asyncio.wait_for(GLOBAL_LOGGING_WORKER.flush(), timeout=10.0) + + assert isinstance(response, DecisionsResponse) + assert recorder.call_types == ["asystemone"] + assert not isinstance(response, OpenAIDecisionResponse) diff --git a/ui/litellm-dashboard/src/app/(dashboard)/layout.integration.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/layout.integration.test.tsx new file mode 100644 index 00000000000..537f2115eda --- /dev/null +++ b/ui/litellm-dashboard/src/app/(dashboard)/layout.integration.test.tsx @@ -0,0 +1,189 @@ +import { afterEach, beforeEach, describe, expect, it, vi } from "vitest"; +import userEvent from "@testing-library/user-event"; +import { renderWithProviders, screen, testQueryClient, waitFor, within } from "../../../tests/test-utils"; +import { AuthProvider } from "@/contexts/AuthContext"; +import { Dialog, DialogClose, DialogContent, DialogTitle, DialogTrigger } from "@/components/ui/dialog"; +import { + AlertDialog, + AlertDialogCancel, + AlertDialogContent, + AlertDialogTitle, + AlertDialogTrigger, +} from "@/components/ui/alert-dialog"; +import { Sheet, SheetClose, SheetContent, SheetTitle, SheetTrigger } from "@/components/ui/sheet"; +import Layout from "./layout"; + +vi.unmock("@/app/(dashboard)/hooks/useAuthorized"); +vi.unmock("@/lib/toast"); +vi.mock("next/navigation", () => ({ + useRouter: () => ({ push: vi.fn(), replace: vi.fn() }), + useSearchParams: () => new URLSearchParams(), + usePathname: () => "/ui/guardrails", +})); + +beforeEach(() => { + testQueryClient.clear(); + localStorage.clear(); + const claims = { + key: "sk-layout-test", + user_id: "layout-admin", + user_role: "proxy_admin", + login_method: "sso", + exp: Date.now() / 1000 + 3600, + }; + const token = `${btoa(JSON.stringify({ alg: "none" }))}.${btoa(JSON.stringify(claims))}.test`; + document.cookie = `token=${token}; Path=/`; + vi.stubGlobal( + "fetch", + vi.fn(async (input: RequestInfo | URL) => { + const url = input instanceof Request ? input.url : String(input); + const path = new URL(url, "http://localhost").pathname; + switch (path) { + case "/litellm/.well-known/litellm-ui-config": + return Response.json({ server_root_path: "", proxy_base_url: "", admin_ui_disabled: false }); + case "/get/ui_settings": + case "/get/ui_theme_settings": + return Response.json({ values: {} }); + case "/api/plugins": + case "/team/list": + case "/organization/list": + return Response.json([]); + case "/health/readiness/details": + return Response.json({ status: "healthy" }); + case "/health/license": { + const license = { + has_license: false, + license_type: null, + expiration_date: null, + allowed_features: [], + limits: { max_users: null, max_teams: null }, + }; + return Response.json(license); + } + case "/user/available_users": { + const usage = { + total_users: null, + total_users_used: 1, + total_users_remaining: null, + total_teams: null, + total_teams_used: 0, + total_teams_remaining: null, + }; + return Response.json(usage); + } + case "/sso/get/ui_settings": + return Response.json({}); + case "/get/user_banner": + return Response.json({ enabled: false, message: "", severity: "info" }); + case "/get/latest_release_info": { + const release = { version: "", release_url: "", bug_fixes: 0, new_features: 0, other_updates: 0 }; + return Response.json(release); + } + case "/public/litellm_blog_posts": + return Response.json({ posts: [] }); + default: + throw new Error(`Unexpected request: ${url}`); + } + }), + ); +}); + +afterEach(() => { + testQueryClient.clear(); + vi.unstubAllGlobals(); + document.cookie = "token=; Max-Age=0; Path=/"; + localStorage.clear(); +}); + +describe("dashboard overlay boundaries", () => { + it("should render the page with mobile navigation closed", async () => { + renderWithProviders( + + +

Gateway page

+ + , + ); + + expect(await screen.findByRole("heading", { name: "Gateway page" })).toBeVisible(); + expect(screen.getByRole("button", { name: "Open navigation" })).toBeVisible(); + expect(screen.queryByRole("dialog", { name: "Navigation" })).not.toBeInTheDocument(); + }); + + it.each([ + { + name: "dialog", + role: "dialog", + slot: "dialog-overlay", + page: ( + + Open page overlay + + Page overlay + Close page overlay + + + ), + }, + { + name: "alert dialog", + role: "alertdialog", + slot: "alert-dialog-overlay", + page: ( + + Open page overlay + + Page overlay + Close page overlay + + + ), + }, + { + name: "sheet", + role: "dialog", + slot: "sheet-overlay", + page: ( + + Open page overlay + + Page overlay + Close page overlay + + + ), + }, + ])("should display the page $name backdrop while navigation is closed", async ({ role, slot, page }) => { + const user = userEvent.setup(); + renderWithProviders( + + {page} + , + ); + + await user.click(await screen.findByRole("button", { name: "Open page overlay" })); + await screen.findByRole(role, { name: "Page overlay" }); + expect( + screen.getAllByRole("presentation", { hidden: true }).find((element) => element.dataset.slot === slot), + ).toBeVisible(); + await user.click(screen.getByRole("button", { name: "Close page overlay" })); + }); + + it("should return focus to the navigation trigger after closing the drawer with Escape", async () => { + const user = userEvent.setup(); + renderWithProviders( + + +

Gateway page

+ + , + ); + + const trigger = await screen.findByRole("button", { name: "Open navigation" }); + await user.click(trigger); + const navigation = await screen.findByRole("dialog", { name: "Navigation" }); + expect(within(navigation).getByRole("link", { name: "Virtual Keys" })).toBeVisible(); + await user.keyboard("{Escape}"); + await waitFor(() => expect(trigger).toHaveFocus()); + }); +}); diff --git a/ui/litellm-dashboard/src/app/(dashboard)/layout.test.tsx b/ui/litellm-dashboard/src/app/(dashboard)/layout.test.tsx index 876a4e24d84..552bffb6cd8 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/layout.test.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/layout.test.tsx @@ -1,5 +1,5 @@ import { describe, it, expect, vi, beforeEach, afterEach } from "vitest"; -import { fireEvent, render, screen, waitFor, within } from "@testing-library/react"; +import { fireEvent, renderWithProviders as render, screen, waitFor, within } from "../../../tests/test-utils"; import { usePathname } from "next/navigation"; import { AuthProvider } from "@/contexts/AuthContext"; import Layout from "./layout"; diff --git a/ui/litellm-dashboard/src/app/(dashboard)/layout.tsx b/ui/litellm-dashboard/src/app/(dashboard)/layout.tsx index 81d290cdfe4..9233a56781e 100644 --- a/ui/litellm-dashboard/src/app/(dashboard)/layout.tsx +++ b/ui/litellm-dashboard/src/app/(dashboard)/layout.tsx @@ -1,6 +1,7 @@ "use client"; import React, { Suspense, useState, useRef, useEffect } from "react"; +import { Dialog as DialogPrimitive } from "@base-ui/react/dialog"; import { DashboardHeader } from "@/components/DashboardHeader"; import Navbar from "@/components/navbar"; import LoadingScreen from "@/components/common_components/LoadingScreen"; @@ -115,6 +116,7 @@ function DashboardShell({ children }: { children: React.ReactNode }) { const navigationKey = `${pathname}?${searchParams.toString()}`; const isDesktop = useMediaQuery("(min-width: 768px)", { initializeWithValue: false }); const [mobileNavigationKey, setMobileNavigationKey] = useState(null); + const [mobileNavigationHandle] = useState(() => DialogPrimitive.createHandle()); if (mobileNavigationKey !== null && (isDesktop || mobileNavigationKey !== navigationKey)) { setMobileNavigationKey(null); } @@ -152,11 +154,15 @@ function DashboardShell({ children }: { children: React.ReactNode }) { // sidebar owns its own scroll and the content column scrolls independently, // so the page can't be dragged past the end of the nav. return ( - setMobileNavigationKey(open ? navigationKey : null)}> -
-
- -
+
+
+ +
+ setMobileNavigationKey(open ? navigationKey : null)} + > Navigation setMobileNavigationKey(null)} /> - - -
- - } - > - - - } - /> - - - - - - -
{children}
-
- - -
- + + + +
+ + } + > + + + } + /> + + + + + + +
{children}
+
+ + +
); } diff --git a/ui/litellm-dashboard/src/lib/http/schema.d.ts b/ui/litellm-dashboard/src/lib/http/schema.d.ts index 4ba005584a4..3f786a38765 100644 --- a/ui/litellm-dashboard/src/lib/http/schema.d.ts +++ b/ui/litellm-dashboard/src/lib/http/schema.d.ts @@ -28075,7 +28075,7 @@ export interface components { * CallTypes * @enum {string} */ - CallTypes: "embedding" | "aembedding" | "completion" | "acompletion" | "atext_completion" | "text_completion" | "image_generation" | "aimage_generation" | "image_edit" | "aimage_edit" | "moderation" | "amoderation" | "atranscription" | "transcription" | "aspeech" | "speech" | "rerank" | "arerank" | "search" | "asearch" | "decisions" | "adecisions" | "_arealtime" | "_aresponses_websocket" | "create_batch" | "acreate_batch" | "aretrieve_batch" | "retrieve_batch" | "acancel_batch" | "cancel_batch" | "pass_through_endpoint" | "anthropic_messages" | "aanthropic_messages" | "get_assistants" | "aget_assistants" | "create_assistants" | "acreate_assistants" | "delete_assistant" | "adelete_assistant" | "acreate_thread" | "create_thread" | "aget_thread" | "get_thread" | "a_add_message" | "add_message" | "aget_messages" | "get_messages" | "arun_thread" | "run_thread" | "arun_thread_stream" | "run_thread_stream" | "afile_retrieve" | "file_retrieve" | "afile_delete" | "file_delete" | "afile_list" | "file_list" | "acreate_file" | "create_file" | "afile_content" | "file_content" | "create_fine_tuning_job" | "acreate_fine_tuning_job" | "create_video" | "acreate_video" | "video_generation" | "avideo_generation" | "avideo_retrieve" | "video_retrieve" | "avideo_content" | "video_content" | "video_remix" | "avideo_remix" | "video_list" | "avideo_list" | "video_retrieve_job" | "avideo_retrieve_job" | "video_delete" | "avideo_delete" | "video_create_character" | "avideo_create_character" | "video_get_character" | "avideo_get_character" | "video_edit" | "avideo_edit" | "video_extension" | "avideo_extension" | "vector_store_file_create" | "avector_store_file_create" | "vector_store_file_list" | "avector_store_file_list" | "vector_store_file_retrieve" | "avector_store_file_retrieve" | "vector_store_file_content" | "avector_store_file_content" | "vector_store_file_update" | "avector_store_file_update" | "vector_store_file_delete" | "avector_store_file_delete" | "vector_store_create" | "avector_store_create" | "vector_store_search" | "avector_store_search" | "ingest" | "aingest" | "query" | "aquery" | "create_interaction" | "acreate_interaction" | "create_container" | "acreate_container" | "list_containers" | "alist_containers" | "retrieve_container" | "aretrieve_container" | "delete_container" | "adelete_container" | "list_container_files" | "alist_container_files" | "upload_container_file" | "aupload_container_file" | "create_sandbox" | "acreate_sandbox" | "delete_sandbox" | "adelete_sandbox" | "run_code" | "arun_code" | "code_interpreter_tool" | "acode_interpreter_tool" | "acancel_fine_tuning_job" | "cancel_fine_tuning_job" | "alist_fine_tuning_jobs" | "list_fine_tuning_jobs" | "aretrieve_fine_tuning_job" | "retrieve_fine_tuning_job" | "responses" | "aresponses" | "alist_input_items" | "llm_passthrough_route" | "allm_passthrough_route" | "generate_content" | "agenerate_content" | "generate_content_stream" | "agenerate_content_stream" | "ocr" | "aocr" | "call_mcp_tool" | "list_mcp_tools" | "asend_message" | "send_message" | "acreate_skill"; + CallTypes: "embedding" | "aembedding" | "completion" | "acompletion" | "atext_completion" | "text_completion" | "image_generation" | "aimage_generation" | "image_edit" | "aimage_edit" | "moderation" | "amoderation" | "atranscription" | "transcription" | "aspeech" | "speech" | "rerank" | "arerank" | "search" | "asearch" | "decisions" | "adecisions" | "systemone" | "asystemone" | "_arealtime" | "_aresponses_websocket" | "create_batch" | "acreate_batch" | "aretrieve_batch" | "retrieve_batch" | "acancel_batch" | "cancel_batch" | "pass_through_endpoint" | "anthropic_messages" | "aanthropic_messages" | "get_assistants" | "aget_assistants" | "create_assistants" | "acreate_assistants" | "delete_assistant" | "adelete_assistant" | "acreate_thread" | "create_thread" | "aget_thread" | "get_thread" | "a_add_message" | "add_message" | "aget_messages" | "get_messages" | "arun_thread" | "run_thread" | "arun_thread_stream" | "run_thread_stream" | "afile_retrieve" | "file_retrieve" | "afile_delete" | "file_delete" | "afile_list" | "file_list" | "acreate_file" | "create_file" | "afile_content" | "file_content" | "create_fine_tuning_job" | "acreate_fine_tuning_job" | "create_video" | "acreate_video" | "video_generation" | "avideo_generation" | "avideo_retrieve" | "video_retrieve" | "avideo_content" | "video_content" | "video_remix" | "avideo_remix" | "video_list" | "avideo_list" | "video_retrieve_job" | "avideo_retrieve_job" | "video_delete" | "avideo_delete" | "video_create_character" | "avideo_create_character" | "video_get_character" | "avideo_get_character" | "video_edit" | "avideo_edit" | "video_extension" | "avideo_extension" | "vector_store_file_create" | "avector_store_file_create" | "vector_store_file_list" | "avector_store_file_list" | "vector_store_file_retrieve" | "avector_store_file_retrieve" | "vector_store_file_content" | "avector_store_file_content" | "vector_store_file_update" | "avector_store_file_update" | "vector_store_file_delete" | "avector_store_file_delete" | "vector_store_create" | "avector_store_create" | "vector_store_search" | "avector_store_search" | "ingest" | "aingest" | "query" | "aquery" | "create_interaction" | "acreate_interaction" | "create_container" | "acreate_container" | "list_containers" | "alist_containers" | "retrieve_container" | "aretrieve_container" | "delete_container" | "adelete_container" | "list_container_files" | "alist_container_files" | "upload_container_file" | "aupload_container_file" | "create_sandbox" | "acreate_sandbox" | "delete_sandbox" | "adelete_sandbox" | "run_code" | "arun_code" | "code_interpreter_tool" | "acode_interpreter_tool" | "acancel_fine_tuning_job" | "cancel_fine_tuning_job" | "alist_fine_tuning_jobs" | "list_fine_tuning_jobs" | "aretrieve_fine_tuning_job" | "retrieve_fine_tuning_job" | "responses" | "aresponses" | "alist_input_items" | "llm_passthrough_route" | "allm_passthrough_route" | "generate_content" | "agenerate_content" | "generate_content_stream" | "agenerate_content_stream" | "ocr" | "aocr" | "call_mcp_tool" | "list_mcp_tools" | "asend_message" | "send_message" | "acreate_skill"; /** CallbackDelete */ CallbackDelete: { /** Callback Name */