diff --git a/litellm/__init__.py b/litellm/__init__.py index be8f59d210b..d202bd41cfe 100644 --- a/litellm/__init__.py +++ b/litellm/__init__.py @@ -1684,6 +1684,9 @@ if TYPE_CHECKING: from .llms.bedrock.messages.mantle_transformation import ( AmazonMantleMessagesConfig as AmazonMantleMessagesConfig, ) + from .llms.bedrock_mantle.messages.transformation import ( + BedrockMantleAnthropicMessagesConfig as BedrockMantleAnthropicMessagesConfig, + ) from .llms.together_ai.chat import TogetherAIConfig as TogetherAIConfig from .llms.together_ai.chat.transformation import ( TogetherAIChatConfig as TogetherAIChatConfig, diff --git a/litellm/_lazy_imports_registry.py b/litellm/_lazy_imports_registry.py index 9cfcb9e41f7..bca04a17250 100644 --- a/litellm/_lazy_imports_registry.py +++ b/litellm/_lazy_imports_registry.py @@ -176,6 +176,7 @@ LLM_CONFIG_NAMES: Final = ( "BedrockClaudePlatformMessagesConfig", "AmazonAnthropicClaudeMessagesConfig", "AmazonMantleMessagesConfig", + "BedrockMantleAnthropicMessagesConfig", "TogetherAIConfig", "TogetherAIChatConfig", "NLPCloudConfig", @@ -746,6 +747,10 @@ _LLM_CONFIGS_IMPORT_MAP: Final = { ".llms.bedrock.messages.mantle_transformation", "AmazonMantleMessagesConfig", ), + "BedrockMantleAnthropicMessagesConfig": ( + ".llms.bedrock_mantle.messages.transformation", + "BedrockMantleAnthropicMessagesConfig", + ), "TogetherAIConfig": (".llms.together_ai.chat", "TogetherAIConfig"), "TogetherAIChatConfig": ( ".llms.together_ai.chat.transformation", diff --git a/litellm/anthropic_beta_headers_config.json b/litellm/anthropic_beta_headers_config.json index eb31cc17a15..1331de4c266 100644 --- a/litellm/anthropic_beta_headers_config.json +++ b/litellm/anthropic_beta_headers_config.json @@ -131,6 +131,41 @@ "web-fetch-2025-09-10": null, "web-search-2025-03-05": null }, + "bedrock_mantle": { + "advanced-tool-use-2025-11-20": "tool-search-tool-2025-10-19", + "advisor-tool-2026-03-01": null, + "bash_20241022": null, + "bash_20250124": null, + "claude-code-20250219": "claude-code-20250219", + "code-execution-2025-08-25": null, + "compact-2026-01-12": "compact-2026-01-12", + "computer-use-2025-01-24": "computer-use-2025-01-24", + "computer-use-2025-11-24": "computer-use-2025-11-24", + "context-1m-2025-08-07": "context-1m-2025-08-07", + "context-management-2025-06-27": "context-management-2025-06-27", + "effort-2025-11-24": "effort-2025-11-24", + "fast-mode-2026-02-01": null, + "files-api-2025-04-14": null, + "fine-grained-tool-streaming-2025-05-14": "fine-grained-tool-streaming-2025-05-14", + "interleaved-thinking-2025-05-14": "interleaved-thinking-2025-05-14", + "mcp-client-2025-04-04": null, + "mcp-client-2025-11-20": null, + "mcp-servers-2025-12-04": null, + "output-128k-2025-02-19": "output-128k-2025-02-19", + "per-turn-control-2026-07-01": "per-turn-control-2026-07-01", + "prompt-caching-scope-2026-01-05": null, + "skills-2025-10-02": null, + "structured-output-2024-03-01": null, + "structured-outputs-2025-11-13": "structured-outputs-2025-11-13", + "text_editor_20241022": null, + "text_editor_20250124": null, + "thinking-binding-controls-2026-08-01": "thinking-binding-controls-2026-08-01", + "token-efficient-tools-2025-02-19": "token-efficient-tools-2025-02-19", + "tool-examples-2025-10-29": "tool-examples-2025-10-29", + "tool-search-tool-2025-10-19": "tool-search-tool-2025-10-19", + "web-fetch-2025-09-10": null, + "web-search-2025-03-05": "web-search-2025-03-05" + }, "vertex_ai": { "advisor-tool-2026-03-01": null, "advanced-tool-use-2025-11-20": "tool-search-tool-2025-10-19", diff --git a/litellm/anthropic_beta_headers_manager.py b/litellm/anthropic_beta_headers_manager.py index abce47c191e..7e7099a53b0 100644 --- a/litellm/anthropic_beta_headers_manager.py +++ b/litellm/anthropic_beta_headers_manager.py @@ -334,7 +334,7 @@ def update_headers_with_filtered_beta( Updated headers dict """ existing_beta: Final = headers.get("anthropic-beta") - if not existing_beta: + if existing_beta is None: return headers # Parse existing beta headers diff --git a/litellm/llms/anthropic/experimental_pass_through/messages/handler.py b/litellm/llms/anthropic/experimental_pass_through/messages/handler.py index 87a4801f987..d87cb0a64f5 100644 --- a/litellm/llms/anthropic/experimental_pass_through/messages/handler.py +++ b/litellm/llms/anthropic/experimental_pass_through/messages/handler.py @@ -651,6 +651,11 @@ def anthropic_messages_handler( "display": "summarized", } + resolved_api_base: Final = ( + dynamic_api_base + if dynamic_api_base is not None and anthropic_messages_provider_config.uses_get_llm_provider_api_base() + else api_base + ) return base_llm_http_handler.anthropic_messages_handler( model=model, messages=strip_provider_specific_fields_from_anthropic_messages(messages), @@ -662,7 +667,7 @@ def anthropic_messages_handler( litellm_params=litellm_params, logging_obj=litellm_logging_obj, api_key=api_key, - api_base=api_base, + api_base=resolved_api_base, stream=stream, kwargs=kwargs, ) diff --git a/litellm/llms/base_llm/anthropic_messages/transformation.py b/litellm/llms/base_llm/anthropic_messages/transformation.py index 8e7c22930fa..101a5e6c58c 100644 --- a/litellm/llms/base_llm/anthropic_messages/transformation.py +++ b/litellm/llms/base_llm/anthropic_messages/transformation.py @@ -128,6 +128,9 @@ class BaseAnthropicMessagesConfig(ABC): """ return True + def uses_get_llm_provider_api_base(self) -> bool: + return False + def get_async_streaming_response_iterator( self, model: str, diff --git a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py index d2be1ad9156..4b52a3bafe6 100644 --- a/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py +++ b/litellm/llms/bedrock/messages/invoke_transformations/anthropic_claude3_transformation.py @@ -1,4 +1,4 @@ -from collections.abc import AsyncIterator +from collections.abc import AsyncIterator, Mapping from types import MappingProxyType from typing import TYPE_CHECKING, Any, Final, cast @@ -445,13 +445,16 @@ class AmazonAnthropicClaudeMessagesConfig( # Bedrock InvokeModel DOES support ``clear_tool_uses_20250919`` under the # ``context-management-2025-06-27`` beta. AWS docs: # https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-anthropic-claude-messages-tool-use.md - _BEDROCK_INVOKE_SUPPORTED_CONTEXT_MANAGEMENT_EDITS: dict[str, str] = { - "compact_20260112": ANTHROPIC_BETA_HEADER_VALUES.COMPACT_2026_01_12.value, - "clear_tool_uses_20250919": ANTHROPIC_BETA_HEADER_VALUES.CONTEXT_MANAGEMENT_2025_06_27.value, - } + _BEDROCK_INVOKE_SUPPORTED_CONTEXT_MANAGEMENT_EDITS: Mapping[str, str] = MappingProxyType( + { + "compact_20260112": ANTHROPIC_BETA_HEADER_VALUES.COMPACT_2026_01_12.value, + "clear_tool_uses_20250919": ANTHROPIC_BETA_HEADER_VALUES.CONTEXT_MANAGEMENT_2025_06_27.value, + } + ) - @staticmethod + @classmethod def _filter_context_management_for_bedrock_invoke( + cls, anthropic_messages_request: dict, beta_set: set, ) -> None: @@ -481,7 +484,7 @@ class AmazonAnthropicClaudeMessagesConfig( anthropic_messages_request.pop("context_management", None) return - supported: Final = AmazonAnthropicClaudeMessagesConfig._BEDROCK_INVOKE_SUPPORTED_CONTEXT_MANAGEMENT_EDITS + supported: Final = cls._BEDROCK_INVOKE_SUPPORTED_CONTEXT_MANAGEMENT_EDITS retained_edits: Final = [e for e in edits if isinstance(e, dict) and e.get("type") in supported] if not retained_edits: anthropic_messages_request.pop("context_management", None) @@ -546,15 +549,16 @@ class AmazonAnthropicClaudeMessagesConfig( if "tool-search-tool-2025-10-19" in beta_set: beta_set.add("tool-examples-2025-10-29") + beta_provider: Final = self.custom_llm_provider or "bedrock" filtered_betas: Final = sorted( filter_and_transform_beta_headers( beta_headers=list(beta_set), - provider="bedrock", + provider=beta_provider, ) ) dropped_user_betas: Final = sorted( - b for b in user_beta_set if not filter_and_transform_beta_headers([b], provider="bedrock") + b for b in user_beta_set if not filter_and_transform_beta_headers([b], provider=beta_provider) ) if dropped_user_betas: verbose_logger.warning( diff --git a/litellm/llms/bedrock_mantle/messages/__init__.py b/litellm/llms/bedrock_mantle/messages/__init__.py new file mode 100644 index 00000000000..e69de29bb2d diff --git a/litellm/llms/bedrock_mantle/messages/transformation.py b/litellm/llms/bedrock_mantle/messages/transformation.py new file mode 100644 index 00000000000..6e975d072ed --- /dev/null +++ b/litellm/llms/bedrock_mantle/messages/transformation.py @@ -0,0 +1,127 @@ +from collections.abc import Mapping +from types import MappingProxyType +from typing import Final + +from pydantic import TypeAdapter + +from litellm.llms.anthropic.experimental_pass_through.messages.transformation import ( + DEFAULT_ANTHROPIC_API_VERSION, +) +from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM +from litellm.llms.bedrock.common_utils import MANTLE_MESSAGES_PATH +from litellm.llms.bedrock.messages.mantle_transformation import AmazonMantleMessagesConfig +from litellm.llms.bedrock_mantle.common_utils import ( + MANTLE_HOST_RE, + BedrockMantleAuthMixin, + resolve_mantle_region, +) +from litellm.secret_managers.main import get_secret_str +from litellm.types.llms.anthropic import ANTHROPIC_BETA_HEADER_VALUES +from litellm.types.router import GenericLiteLLMParams + +_BASE_SUFFIXES_TO_STRIP: Final = ( + MANTLE_MESSAGES_PATH, + "/v1/messages", + "/messages", + "/anthropic/v1", + "/openai/v1", + "/v1", +) +_BODY_FIELDS_MANTLE_READS_FROM_HEADERS: Final = frozenset({"anthropic_version", "anthropic_beta"}) +_ANTHROPIC_BETAS: Final = TypeAdapter(tuple[str, ...]) +_MANTLE_REQUEST: Final = TypeAdapter(dict[str, object]) + + +def build_mantle_native_messages_url(api_base: str | None, litellm_params: Mapping[str, object]) -> str: + region: Final = resolve_mantle_region(MappingProxyType({**litellm_params, "api_base": api_base})) + configured: Final = ( + api_base or get_secret_str("BEDROCK_MANTLE_API_BASE") or f"https://bedrock-mantle.{region}.api.aws" + ).rstrip("/") + stripped: Final = next( + (configured[: -len(suffix)] for suffix in _BASE_SUFFIXES_TO_STRIP if configured.endswith(suffix)), + configured, + ) + host: Final = f"https://bedrock-mantle.{region}.api.aws" if MANTLE_HOST_RE.match(stripped) else stripped + return f"{host}{MANTLE_MESSAGES_PATH}" + + +class BedrockMantleAnthropicMessagesConfig(BedrockMantleAuthMixin, AmazonMantleMessagesConfig): + _BEDROCK_INVOKE_SUPPORTED_CONTEXT_MANAGEMENT_EDITS: Mapping[str, str] = MappingProxyType( + { + **AmazonMantleMessagesConfig._BEDROCK_INVOKE_SUPPORTED_CONTEXT_MANAGEMENT_EDITS, + "clear_thinking_20251015": ANTHROPIC_BETA_HEADER_VALUES.CONTEXT_MANAGEMENT_2025_06_27.value, + } + ) + + def __init__(self, aws_signer: BaseAWSLLM | None = None) -> None: + AmazonMantleMessagesConfig.__init__(self) + self._aws_signer = aws_signer or self + + @property + def custom_llm_provider(self) -> str | None: + return "bedrock_mantle" + + def uses_get_llm_provider_api_base(self) -> bool: + return True + + def get_complete_url( + self, + api_base: str | None, + api_key: str | None, + model: str, + optional_params: dict, + litellm_params: dict, + stream: bool | None = None, + ) -> str: + return build_mantle_native_messages_url(api_base=api_base, litellm_params=litellm_params) + + def validate_anthropic_messages_environment( + self, + headers: dict, + model: str, + messages: list[dict], + optional_params: dict, + litellm_params: dict, + api_key: str | None = None, + api_base: str | None = None, + ) -> tuple[dict, str | None]: + merged_headers, resolved_api_base = super().validate_anthropic_messages_environment( + headers=headers, + model=model, + messages=messages, + optional_params=optional_params, + litellm_params=litellm_params, + api_key=api_key, + api_base=api_base, + ) + if any(name.lower() == "anthropic-version" for name in merged_headers): + return merged_headers, resolved_api_base + return { # mutable-ok: the base class contract returns a dict the handler signs into in place + **merged_headers, + "anthropic-version": DEFAULT_ANTHROPIC_API_VERSION, + }, resolved_api_base + + def transform_anthropic_messages_request( + self, + model: str, + messages: list[dict], + anthropic_messages_optional_request_params: dict, + litellm_params: GenericLiteLLMParams, + headers: dict, + ) -> dict: + request: Final = _MANTLE_REQUEST.validate_python( + super().transform_anthropic_messages_request( + model=model, + messages=messages, + anthropic_messages_optional_request_params=anthropic_messages_optional_request_params, + litellm_params=litellm_params, + headers=headers, + ), + ) + betas: Final = request.get("anthropic_beta") + if betas is not None: + header_betas: Final = ",".join(_ANTHROPIC_BETAS.validate_python(betas)) + headers["anthropic-beta"] = header_betas # rebind-ok: the handler signs and sends this same dict + return { # mutable-ok: the base class contract returns the dict the handler serializes as the body + key: value for key, value in request.items() if key not in _BODY_FIELDS_MANTLE_READS_FROM_HEADERS + } diff --git a/litellm/model_prices_and_context_window_backup.json b/litellm/model_prices_and_context_window_backup.json index 93a9e5e3c30..97a38ac1657 100644 --- a/litellm/model_prices_and_context_window_backup.json +++ b/litellm/model_prices_and_context_window_backup.json @@ -60564,6 +60564,34 @@ "supports_tool_choice": true, "supports_vision": true }, + "bedrock_mantle/anthropic.claude-haiku-4-5": { + "cache_creation_input_token_cost": 1.25e-06, + "cache_creation_input_token_cost_above_1hr": 2e-06, + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "litellm_provider": "bedrock_mantle", + "supports_tool_search": true, + "max_input_tokens": 200000, + "max_output_tokens": 64000, + "max_tokens": 64000, + "mode": "chat", + "output_cost_per_token": 5e-06, + "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", + "supports_assistant_prefill": true, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 4096, + "input_cost_per_token_batches": 5e-07, + "output_cost_per_token_batches": 2.5e-06 + }, "us.xai.grok-4.6": { "input_cost_per_token": 2.2e-06, "output_cost_per_token": 6.6e-06, diff --git a/litellm/proxy/litellm_pre_call_utils.py b/litellm/proxy/litellm_pre_call_utils.py index 9a973755894..e415a78f412 100644 --- a/litellm/proxy/litellm_pre_call_utils.py +++ b/litellm/proxy/litellm_pre_call_utils.py @@ -3418,7 +3418,12 @@ async def add_guardrails_from_policy_engine( _ANTHROPIC_API_HEADER_PROVIDERS: Final = ",".join( - (LlmProviders.ANTHROPIC.value, LlmProviders.BEDROCK.value, LlmProviders.VERTEX_AI.value) + ( + LlmProviders.ANTHROPIC.value, + LlmProviders.BEDROCK.value, + LlmProviders.BEDROCK_MANTLE.value, + LlmProviders.VERTEX_AI.value, + ) ) _ANTHROPIC_OAUTH_CREDENTIAL_PROVIDERS: Final = LlmProviders.ANTHROPIC.value diff --git a/litellm/utils.py b/litellm/utils.py index 9a80b115d4b..da2b3da6302 100644 --- a/litellm/utils.py +++ b/litellm/utils.py @@ -5624,6 +5624,12 @@ def _get_model_info_from_generalization( return None +def _strip_mantle_region_prefix(model: str) -> str: + from litellm.llms.bedrock_mantle.common_utils import split_mantle_region_prefix + + return split_mantle_region_prefix(model)[1] + + def _get_potential_model_names(model: str, custom_llm_provider: str | None) -> PotentialModelNamesAndCustomLLMProvider: if custom_llm_provider is None: # Get custom_llm_provider @@ -5656,20 +5662,30 @@ def _get_potential_model_names(model: str, custom_llm_provider: str | None) -> P split_model = strip_bedrock_routing_prefix(split_model) + region_free_split_model: Final = ( + _strip_mantle_region_prefix(split_model) if custom_llm_provider == "bedrock_mantle" else split_model + ) + region_free_combined_stripped_model_name: Final = ( + f"bedrock_mantle/{_strip_model_name(model=region_free_split_model, custom_llm_provider=custom_llm_provider)}" + if custom_llm_provider == "bedrock_mantle" + else combined_stripped_model_name + ) provider_model_info: Final = ( - ProviderConfigManager.get_provider_model_info(model=split_model, provider=LlmProviders(custom_llm_provider)) + ProviderConfigManager.get_provider_model_info( + model=region_free_split_model, provider=LlmProviders(custom_llm_provider) + ) if custom_llm_provider in LlmProvidersSet else None ) provider_cost_key: Final = ( - provider_model_info.get_model_cost_key(split_model) if provider_model_info is not None else None + provider_model_info.get_model_cost_key(region_free_split_model) if provider_model_info is not None else None ) return PotentialModelNamesAndCustomLLMProvider( - split_model=split_model, + split_model=region_free_split_model, combined_model_name=combined_model_name, stripped_model_name=stripped_model_name, - combined_stripped_model_name=combined_stripped_model_name, + combined_stripped_model_name=region_free_combined_stripped_model_name, provider_prefixed_model_name=provider_cost_key or provider_prefixed_model_name, custom_llm_provider=cast(str, custom_llm_provider), ) @@ -8681,6 +8697,13 @@ class ProviderConfigManager: from litellm.llms.bedrock.common_utils import BedrockModelInfo return BedrockModelInfo.get_bedrock_provider_config_for_messages_api(model) + elif litellm.LlmProviders.BEDROCK_MANTLE == provider: + if "claude" in model_lower: + from litellm.llms.bedrock_mantle.messages.transformation import ( + BedrockMantleAnthropicMessagesConfig, + ) + + return BedrockMantleAnthropicMessagesConfig() elif litellm.LlmProviders.VERTEX_AI == provider: if "claude" in model_lower: from litellm.llms.vertex_ai.vertex_ai_partner_models.anthropic.experimental_pass_through.transformation import ( diff --git a/model_prices_and_context_window.json b/model_prices_and_context_window.json index 93a9e5e3c30..97a38ac1657 100644 --- a/model_prices_and_context_window.json +++ b/model_prices_and_context_window.json @@ -60564,6 +60564,34 @@ "supports_tool_choice": true, "supports_vision": true }, + "bedrock_mantle/anthropic.claude-haiku-4-5": { + "cache_creation_input_token_cost": 1.25e-06, + "cache_creation_input_token_cost_above_1hr": 2e-06, + "cache_read_input_token_cost": 1e-07, + "input_cost_per_token": 1e-06, + "litellm_provider": "bedrock_mantle", + "supports_tool_search": true, + "max_input_tokens": 200000, + "max_output_tokens": 64000, + "max_tokens": 64000, + "mode": "chat", + "output_cost_per_token": 5e-06, + "source": "https://aws.amazon.com/about-aws/whats-new/2025/10/claude-4-5-haiku-anthropic-amazon-bedrock", + "supports_assistant_prefill": true, + "supports_computer_use": true, + "supports_function_calling": true, + "supports_pdf_input": true, + "supports_prompt_caching": true, + "supports_reasoning": true, + "supports_response_schema": true, + "supports_tool_choice": true, + "supports_vision": true, + "supports_native_structured_output": true, + "supports_parallel_tool_use_config": true, + "prompt_cache_min_tokens": 4096, + "input_cost_per_token_batches": 5e-07, + "output_cost_per_token_batches": 2.5e-06 + }, "us.xai.grok-4.6": { "input_cost_per_token": 2.2e-06, "output_cost_per_token": 6.6e-06, diff --git a/tests/proxy_unit_tests/test_proxy_utils.py b/tests/proxy_unit_tests/test_proxy_utils.py index 7cdd7365209..1134f41a940 100644 --- a/tests/proxy_unit_tests/test_proxy_utils.py +++ b/tests/proxy_unit_tests/test_proxy_utils.py @@ -2003,7 +2003,7 @@ def test_provider_specific_header(): ) # Verify multi-provider support: anthropic headers work across multiple providers assert data["provider_specific_header"] == { - "custom_llm_provider": "anthropic,bedrock,vertex_ai", + "custom_llm_provider": "anthropic,bedrock,bedrock_mantle,vertex_ai", "extra_headers": { "anthropic-beta": "prompt-caching-2024-07-31", }, @@ -2075,7 +2075,7 @@ def test_provider_specific_header_multi_provider(): assert "provider_specific_header" in data assert ( data["provider_specific_header"]["custom_llm_provider"] - == "anthropic,bedrock,vertex_ai" + == "anthropic,bedrock,bedrock_mantle,vertex_ai" ) assert data["provider_specific_header"]["extra_headers"] == { "anthropic-beta": "context-1m-2025-08-07", diff --git a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py index e8bfcb86bf6..cc4eb1d4136 100644 --- a/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py +++ b/tests/test_litellm/llms/anthropic/experimental_pass_through/messages/test_anthropic_experimental_pass_through_messages_handler.py @@ -1440,6 +1440,46 @@ async def test_anthropic_messages_leaves_non_provider_failures_unmapped(): assert "Traceback" not in str(excinfo.value) +def _recording_client(seen_urls: list[str]) -> AsyncHTTPHandler: + def record_and_answer(request: httpx.Request) -> httpx.Response: + seen_urls.append(str(request.url)) + return httpx.Response( + 200, + json={ + "id": "msg_test", + "type": "message", + "role": "assistant", + "model": "deepseek-chat", + "content": [{"type": "text", "text": "pong"}], + "stop_reason": "end_turn", + "stop_sequence": None, + "usage": {"input_tokens": 3, "output_tokens": 1}, + }, + ) + + upstream = AsyncHTTPHandler() + upstream.client = httpx.AsyncClient(transport=httpx.MockTransport(record_and_answer)) + return upstream + + +@pytest.mark.asyncio +async def test_provider_messages_api_base_env_is_not_shadowed_by_the_chat_default(monkeypatch): + from litellm.llms.anthropic.experimental_pass_through.messages import handler + + monkeypatch.delenv("DEEPSEEK_API_BASE", raising=False) + monkeypatch.setenv("DEEPSEEK_ANTHROPIC_API_BASE", "https://deepseek.internal.example/anthropic") + seen_urls: list[str] = [] + + await handler.anthropic_messages( + max_tokens=16, + messages=[{"role": "user", "content": "ping"}], + model="deepseek/deepseek-chat", + api_key="sk-test", + client=_recording_client(seen_urls), + ) + + assert seen_urls == ["https://deepseek.internal.example/anthropic/v1/messages"] + @pytest.mark.asyncio async def test_anthropic_messages_forwards_safeguards_and_unknown_beta_to_anthropic(): """Shapes are what Claude Code 2.1.278 sends and api.anthropic.com returns, captured 2026-09-21.""" diff --git a/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_messages_transformation.py b/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_messages_transformation.py new file mode 100644 index 00000000000..6bacf8f3d94 --- /dev/null +++ b/tests/test_litellm/llms/bedrock_mantle/test_bedrock_mantle_messages_transformation.py @@ -0,0 +1,484 @@ +""" +Unit tests for the bedrock_mantle native Anthropic Messages route. + +Mantle serves its Claude models only on `/anthropic/v1/messages` (the OpenAI +paths reject them), so `bedrock_mantle/anthropic.claude-*` requests on +/v1/messages must hit that endpoint directly instead of the chat-completions +bridge. These tests lock the dispatcher gate, the URL derivation from the +OpenAI-surface base that get_llm_provider pre-fills, the version header, the +Bearer/SigV4 auth chain, and the wire request through the public entrypoint. +""" + +import json +from unittest.mock import MagicMock + +import httpx +import pytest +import respx + +import litellm +from litellm.caching.llm_caching_handler import LLMClientCache +from litellm.llms.bedrock.base_aws_llm import BaseAWSLLM +from litellm.llms.bedrock_mantle.messages.transformation import ( + BedrockMantleAnthropicMessagesConfig, + build_mantle_native_messages_url, +) +from litellm.types.router import GenericLiteLLMParams +from litellm.utils import ProviderConfigManager + +MESSAGES_PATH = "/anthropic/v1/messages" + + +@pytest.fixture(autouse=True) +def _httpx_transport_with_fresh_clients(monkeypatch): + monkeypatch.setattr(litellm, "disable_aiohttp_transport", True) + monkeypatch.setattr(litellm, "in_memory_llm_clients_cache", LLMClientCache()) + + +@pytest.fixture(autouse=True) +def _no_ambient_mantle_env(monkeypatch): + monkeypatch.delenv("AWS_BEARER_TOKEN_BEDROCK", raising=False) + monkeypatch.delenv("BEDROCK_MANTLE_API_KEY", raising=False) + monkeypatch.delenv("BEDROCK_MANTLE_API_BASE", raising=False) + monkeypatch.delenv("BEDROCK_MANTLE_REGION", raising=False) + monkeypatch.delenv("AWS_REGION_NAME", raising=False) + monkeypatch.delenv("AWS_REGION", raising=False) + + +def _anthropic_response() -> httpx.Response: + return httpx.Response( + status_code=200, + json={ + "id": "msg_test", + "type": "message", + "role": "assistant", + "model": "anthropic.claude-sonnet-5", + "content": [{"type": "text", "text": "pong"}], + "stop_reason": "end_turn", + "stop_sequence": None, + "usage": {"input_tokens": 3, "output_tokens": 1}, + }, + ) + + +_SSE_EVENTS = ( + ( + "message_start", + { + "type": "message_start", + "message": { + "id": "msg_stream", + "type": "message", + "role": "assistant", + "model": "anthropic.claude-sonnet-5", + "content": [], + "stop_reason": None, + "stop_sequence": None, + "usage": {"input_tokens": 3, "output_tokens": 1}, + }, + }, + ), + ("content_block_start", {"type": "content_block_start", "index": 0, "content_block": {"type": "text", "text": ""}}), + ( + "content_block_delta", + {"type": "content_block_delta", "index": 0, "delta": {"type": "text_delta", "text": "pong"}}, + ), + ("content_block_stop", {"type": "content_block_stop", "index": 0}), + ("message_delta", {"type": "message_delta", "delta": {"stop_reason": "end_turn"}, "usage": {"output_tokens": 1}}), + ("message_stop", {"type": "message_stop"}), +) + + +def _sse_response() -> httpx.Response: + body = "".join(f"event: {event}\ndata: {json.dumps(payload)}\n\n" for event, payload in _SSE_EVENTS).encode() + return httpx.Response(status_code=200, content=body, headers={"content-type": "text/event-stream"}) + + +def _mantle_messages_route(region: str) -> respx.Route: + return respx.post(f"https://bedrock-mantle.{region}.api.aws{MESSAGES_PATH}") + + +def _sent_body(route: respx.Route) -> dict: + return json.loads(route.calls.last.request.content) + + +class TestDispatch: + def test_claude_models_get_the_native_messages_config(self): + config = ProviderConfigManager.get_provider_anthropic_messages_config( + model="anthropic.claude-sonnet-5", provider=litellm.LlmProviders.BEDROCK_MANTLE + ) + assert isinstance(config, BedrockMantleAnthropicMessagesConfig) + assert config.custom_llm_provider == "bedrock_mantle" + + @pytest.mark.parametrize("model", ["openai.gpt-5.6-sol", "openai.gpt-oss-120b-1:0", "google.gemma-4-31b"]) + def test_non_claude_models_keep_the_bridge(self, model): + assert ( + ProviderConfigManager.get_provider_anthropic_messages_config( + model=model, provider=litellm.LlmProviders.BEDROCK_MANTLE + ) + is None + ) + + +class TestURL: + @pytest.mark.parametrize( + "api_base", + [ + "https://bedrock-mantle.us-east-1.api.aws/v1", + "https://bedrock-mantle.us-east-1.api.aws/openai/v1", + "https://bedrock-mantle.us-east-1.api.aws/openai/v1/", + "https://bedrock-mantle.us-east-1.api.aws", + "https://bedrock-mantle.us-east-1.api.aws/anthropic/v1/messages", + ], + ) + def test_prefilled_openai_base_becomes_the_messages_endpoint(self, api_base): + url = build_mantle_native_messages_url(api_base, {"aws_region_name": "us-east-1"}) + assert url == f"https://bedrock-mantle.us-east-1.api.aws{MESSAGES_PATH}" + + def test_aws_region_name_wins_over_the_prefilled_host_region(self): + url = build_mantle_native_messages_url( + "https://bedrock-mantle.us-east-1.api.aws/v1", {"aws_region_name": "us-east-2"} + ) + assert url == f"https://bedrock-mantle.us-east-2.api.aws{MESSAGES_PATH}" + + def test_host_region_is_used_when_no_region_param(self): + url = build_mantle_native_messages_url("https://bedrock-mantle.eu-west-1.api.aws/v1", {}) + assert url == f"https://bedrock-mantle.eu-west-1.api.aws{MESSAGES_PATH}" + + def test_custom_host_is_preserved(self): + url = build_mantle_native_messages_url("https://vpce-abc.bedrock-mantle.example.com/v1", {}) + assert url == f"https://vpce-abc.bedrock-mantle.example.com{MESSAGES_PATH}" + + def test_env_base_is_used_without_api_base(self, monkeypatch): + monkeypatch.setenv("BEDROCK_MANTLE_API_BASE", "https://mantle-proxy.internal/openai/v1") + assert build_mantle_native_messages_url(None, {}) == f"https://mantle-proxy.internal{MESSAGES_PATH}" + + def test_default_host_comes_from_mantle_region_env(self, monkeypatch): + monkeypatch.setenv("BEDROCK_MANTLE_REGION", "ap-northeast-1") + assert ( + build_mantle_native_messages_url(None, {}) + == f"https://bedrock-mantle.ap-northeast-1.api.aws{MESSAGES_PATH}" + ) + + def test_config_get_complete_url_reads_litellm_params(self): + config = BedrockMantleAnthropicMessagesConfig() + url = config.get_complete_url( + api_base="https://bedrock-mantle.us-east-1.api.aws/v1", + api_key=None, + model="anthropic.claude-sonnet-5", + optional_params={}, + litellm_params={"aws_region_name": "us-west-2"}, + ) + assert url == f"https://bedrock-mantle.us-west-2.api.aws{MESSAGES_PATH}" + + +class TestEnvironment: + def _validate(self, headers: dict, litellm_params: dict) -> dict: + config = BedrockMantleAnthropicMessagesConfig() + merged, _ = config.validate_anthropic_messages_environment( + headers=headers, + model="anthropic.claude-sonnet-5", + messages=[], + optional_params={}, + litellm_params=litellm_params, + ) + return merged + + def test_adds_the_anthropic_version_header(self): + assert self._validate({}, {})["anthropic-version"] == "2023-06-01" + + def test_keeps_a_caller_supplied_version_header(self): + merged = self._validate({"Anthropic-Version": "2024-01-01"}, {}) + assert merged["Anthropic-Version"] == "2024-01-01" + assert "anthropic-version" not in merged + + def test_project_id_becomes_the_workspace_header(self): + assert self._validate({}, {"aws_bedrock_project_id": "proj_123"})["anthropic-workspace"] == "proj_123" + + +class TestRequestBody: + def test_body_carries_model_and_stream_but_not_the_invoke_version(self): + config = BedrockMantleAnthropicMessagesConfig() + body = config.transform_anthropic_messages_request( + model="anthropic.claude-sonnet-5", + messages=[{"role": "user", "content": "ping"}], + anthropic_messages_optional_request_params={"max_tokens": 8, "stream": True}, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + assert body["model"] == "anthropic.claude-sonnet-5" + assert body["stream"] is True + assert body["max_tokens"] == 8 + assert "anthropic_version" not in body + + def test_body_omits_stream_when_not_streaming(self): + config = BedrockMantleAnthropicMessagesConfig() + body = config.transform_anthropic_messages_request( + model="anthropic.claude-sonnet-5", + messages=[{"role": "user", "content": "ping"}], + anthropic_messages_optional_request_params={"max_tokens": 8}, + litellm_params=GenericLiteLLMParams(), + headers={}, + ) + assert "stream" not in body + + +class TestAuth: + def test_bearer_from_api_key_skips_aws_credentials(self): + signer = BaseAWSLLM() + signer.get_credentials = MagicMock(side_effect=AssertionError("must not resolve AWS credentials")) + config = BedrockMantleAnthropicMessagesConfig(aws_signer=signer) + headers, signed = config.sign_request( + headers={"anthropic-version": "2023-06-01"}, + optional_params={}, + request_data={"model": "anthropic.claude-sonnet-5"}, + api_base=f"https://bedrock-mantle.us-east-1.api.aws{MESSAGES_PATH}", + api_key="arg-bearer", + ) + assert headers["Authorization"] == "Bearer arg-bearer" + assert headers["anthropic-version"] == "2023-06-01" + assert signed == b'{"model": "anthropic.claude-sonnet-5"}' + + def test_bearer_from_mantle_env_key(self, monkeypatch): + monkeypatch.setenv("BEDROCK_MANTLE_API_KEY", "env-bearer") + config = BedrockMantleAnthropicMessagesConfig() + headers, _ = config.sign_request( + headers={}, + optional_params={}, + request_data={}, + api_base=f"https://bedrock-mantle.us-east-1.api.aws{MESSAGES_PATH}", + api_key=None, + ) + assert headers["Authorization"] == "Bearer env-bearer" + + def test_sigv4_scope_is_pinned_to_the_url_host_region(self): + config = BedrockMantleAnthropicMessagesConfig() + headers, signed = config.sign_request( + headers={"anthropic-version": "2023-06-01"}, + optional_params={ + "aws_access_key_id": "AKIAEXAMPLE", + "aws_secret_access_key": "c2VjcmV0LXRlc3Qtc2VjcmV0LXRlc3Qtc2VjcmV0", + "aws_region_name": "us-east-1", + }, + request_data={"model": "anthropic.claude-sonnet-5"}, + api_base=f"https://bedrock-mantle.us-west-2.api.aws{MESSAGES_PATH}", + api_key=None, + ) + assert headers["Authorization"].startswith("AWS4-HMAC-SHA256") + assert "/us-west-2/bedrock/aws4_request" in headers["Authorization"] + assert signed == b'{"model": "anthropic.claude-sonnet-5"}' + + +class TestWireRequest: + @pytest.mark.asyncio + @respx.mock + async def test_claude_request_hits_the_native_messages_endpoint(self): + route = _mantle_messages_route("us-east-1").mock(return_value=_anthropic_response()) + + response = await litellm.anthropic_messages( + model="bedrock_mantle/anthropic.claude-sonnet-5", + messages=[{"role": "user", "content": "ping"}], + max_tokens=8, + api_key="test-bearer", + aws_region_name="us-east-1", + ) + + assert response["content"][0]["text"] == "pong" + assert route.call_count == 1 + sent = route.calls.last.request + assert sent.headers["authorization"] == "Bearer test-bearer" + assert sent.headers["anthropic-version"] == "2023-06-01" + assert "x-api-key" not in sent.headers + body = _sent_body(route) + assert body["model"] == "anthropic.claude-sonnet-5" + assert body["messages"] == [{"role": "user", "content": "ping"}] + assert "anthropic_version" not in body + assert "stream" not in body + + @pytest.mark.asyncio + @respx.mock + async def test_region_prefix_selects_the_host_and_is_not_sent_as_model(self): + route = _mantle_messages_route("us-east-2").mock(return_value=_anthropic_response()) + + await litellm.anthropic_messages( + model="bedrock_mantle/us-east-2/anthropic.claude-haiku-4-5", + messages=[{"role": "user", "content": "ping"}], + max_tokens=8, + api_key="test-bearer", + ) + + assert route.call_count == 1 + assert _sent_body(route)["model"] == "anthropic.claude-haiku-4-5" + + @pytest.mark.asyncio + @respx.mock + async def test_streaming_sends_stream_and_passes_the_sse_through(self): + route = _mantle_messages_route("us-east-1").mock(return_value=_sse_response()) + + response = await litellm.anthropic_messages( + model="bedrock_mantle/anthropic.claude-sonnet-5", + messages=[{"role": "user", "content": "ping"}], + max_tokens=8, + stream=True, + api_key="test-bearer", + aws_region_name="us-east-1", + ) + raw = b"".join([chunk async for chunk in response]) + + assert route.call_count == 1 + assert _sent_body(route)["stream"] is True + text = raw.decode() + assert "event: message_start" in text + assert '"text": "pong"' in text + assert "event: message_stop" in text + + @pytest.mark.asyncio + @respx.mock + async def test_sigv4_request_signs_against_the_messages_url(self): + route = _mantle_messages_route("us-east-1").mock(return_value=_anthropic_response()) + + await litellm.anthropic_messages( + model="bedrock_mantle/anthropic.claude-sonnet-5", + messages=[{"role": "user", "content": "ping"}], + max_tokens=8, + aws_access_key_id="AKIAEXAMPLE", + aws_secret_access_key="c2VjcmV0LXRlc3Qtc2VjcmV0LXRlc3Qtc2VjcmV0", + aws_region_name="us-east-1", + ) + + assert route.call_count == 1 + authorization = route.calls.last.request.headers["authorization"] + assert authorization.startswith("AWS4-HMAC-SHA256") + assert "/us-east-1/bedrock/aws4_request" in authorization + + +def _sent_betas(route: respx.Route) -> list[str]: + return route.calls.last.request.headers["anthropic-beta"].split(",") + + +@pytest.mark.usefixtures("local_beta_headers_config") +class TestBetaHeadersOnTheWire: + async def _send(self, **request_params) -> respx.Route: + route = _mantle_messages_route("us-east-1").mock(return_value=_anthropic_response()) + await litellm.anthropic_messages( + model="bedrock_mantle/anthropic.claude-sonnet-5", + messages=[{"role": "user", "content": "ping"}], + max_tokens=8, + api_key="test-bearer", + aws_region_name="us-east-1", + **request_params, + ) + return route + + @pytest.mark.asyncio + @respx.mock + async def test_betas_mantle_accepts_reach_it_in_the_header(self): + route = await self._send( + extra_headers={ + "anthropic-beta": "claude-code-20250219,interleaved-thinking-2025-05-14,context-management-2025-06-27" + } + ) + + assert _sent_betas(route) == [ + "claude-code-20250219", + "context-management-2025-06-27", + "interleaved-thinking-2025-05-14", + ] + + @pytest.mark.asyncio + @respx.mock + async def test_betas_a_proxy_client_sends_reach_mantle_filtered(self): + from litellm.proxy.litellm_pre_call_utils import add_provider_specific_headers_to_request + + proxy_request_data: dict = {} + add_provider_specific_headers_to_request( + data=proxy_request_data, + headers={ + "anthropic-beta": "claude-code-20250219,fast-mode-2026-02-01,interleaved-thinking-2025-05-14", + "anthropic-version": "2023-06-01", + "user-agent": "claude-cli/2.1.239", + }, + ) + + route = await self._send(**proxy_request_data) + + assert _sent_betas(route) == ["claude-code-20250219", "interleaved-thinking-2025-05-14"] + + @pytest.mark.asyncio + @respx.mock + async def test_betas_mantle_rejects_are_dropped_before_the_request(self): + route = await self._send( + extra_headers={"anthropic-beta": "code-execution-2025-08-25,context-1m-2025-08-07,files-api-2025-04-14"} + ) + + assert _sent_betas(route) == ["context-1m-2025-08-07"] + + @pytest.mark.asyncio + @respx.mock + async def test_no_beta_header_is_sent_when_every_value_is_rejected(self): + route = await self._send(extra_headers={"anthropic-beta": "code-execution-2025-08-25"}) + + assert "anthropic-beta" not in route.calls.last.request.headers + + @pytest.mark.asyncio + @respx.mock + async def test_advanced_tool_use_is_renamed_to_the_beta_mantle_knows(self): + route = await self._send(extra_headers={"anthropic-beta": "advanced-tool-use-2025-11-20"}) + + assert "tool-search-tool-2025-10-19" in _sent_betas(route) + assert "advanced-tool-use-2025-11-20" not in _sent_betas(route) + + @pytest.mark.asyncio + @respx.mock + async def test_a_feature_beta_joins_the_callers_betas_in_the_header(self): + route = await self._send( + extra_headers={"anthropic-beta": "context-1m-2025-08-07"}, + context_management={"edits": [{"type": "clear_tool_uses_20250919"}]}, + ) + + assert _sent_betas(route) == ["context-1m-2025-08-07", "context-management-2025-06-27"] + assert _sent_body(route)["context_management"] == {"edits": [{"type": "clear_tool_uses_20250919"}]} + + @pytest.mark.asyncio + @respx.mock + async def test_betas_and_version_never_travel_in_the_body(self): + route = await self._send( + extra_headers={"anthropic-beta": "context-1m-2025-08-07"}, + context_management={"edits": [{"type": "clear_tool_uses_20250919"}]}, + anthropic_version="bedrock-2023-05-31", + ) + + body = _sent_body(route) + assert "anthropic_beta" not in body + assert "anthropic_version" not in body + assert route.calls.last.request.headers["anthropic-version"] == "2023-06-01" + + @pytest.mark.asyncio + @respx.mock + async def test_clear_thinking_edit_is_forwarded_with_thinking_on(self): + edits = [{"type": "clear_thinking_20251015", "keep": "all"}, {"type": "clear_tool_uses_20250919"}] + route = await self._send( + context_management={"edits": edits}, + thinking={"type": "adaptive"}, + ) + + body = _sent_body(route) + assert body["context_management"] == {"edits": edits} + assert body["thinking"] == {"type": "adaptive"} + assert "context-management-2025-06-27" in _sent_betas(route) + + @pytest.mark.asyncio + @respx.mock + async def test_tools_reach_mantle_unchanged(self): + tools = [ + { + "name": "get_weather", + "description": "Look up the weather", + "input_schema": {"type": "object", "properties": {"city": {"type": "string"}}, "required": ["city"]}, + } + ] + route = await self._send(tools=tools, tool_choice={"type": "auto"}) + + body = _sent_body(route) + assert body["tools"] == tools + assert body["tool_choice"] == {"type": "auto"} diff --git a/tests/test_litellm/proxy/test_litellm_pre_call_utils.py b/tests/test_litellm/proxy/test_litellm_pre_call_utils.py index 88d38d74f49..9257a2dd23d 100644 --- a/tests/test_litellm/proxy/test_litellm_pre_call_utils.py +++ b/tests/test_litellm/proxy/test_litellm_pre_call_utils.py @@ -7249,7 +7249,7 @@ CROSS_ACCOUNT_AUTHORIZATION = "Bearer deliberately-configured-pass-through-token SIGV4_PREFIX = "AWS4-HMAC-SHA256" AUTHORIZATION_HEADER_CASINGS = ["authorization", "Authorization", "AUTHORIZATION"] -LEAK_TARGET_PROVIDERS = ["bedrock", "bedrock_converse", "vertex_ai"] +LEAK_TARGET_PROVIDERS = ["bedrock", "bedrock_converse", "bedrock_mantle", "vertex_ai"] BEDROCK_ENDPOINT = ( "https://bedrock-runtime.us-west-2.amazonaws.com/model/us.anthropic.claude-sonnet-4-5-20250929-v1:0/invoke" @@ -7342,6 +7342,28 @@ def test_oauth_credential_entry_is_scoped_to_anthropic_alone(): assert [entry["custom_llm_provider"] for entry in credential_entries] == ["anthropic"] +@pytest.mark.parametrize("custom_llm_provider", ["anthropic", "bedrock", "bedrock_mantle", "vertex_ai"]) +def test_client_anthropic_api_headers_reach_every_anthropic_messages_provider(custom_llm_provider): + client_headers = { + "anthropic-beta": "claude-code-20250219,interleaved-thinking-2025-05-14", + "anthropic-version": "2023-06-01", + "user-agent": "claude-cli/2.1.239", + } + + forwarded = _headers_forwarded_to(client_headers, custom_llm_provider) + + assert forwarded == { + "anthropic-beta": "claude-code-20250219,interleaved-thinking-2025-05-14", + "anthropic-version": "2023-06-01", + } + + +def test_client_anthropic_api_headers_stay_off_openai_compatible_providers(): + forwarded = _headers_forwarded_to({"anthropic-beta": "claude-code-20250219"}, "openai") + + assert forwarded == {} + + def test_no_provider_specific_header_when_client_sends_nothing_anthropic(): data: dict = {} add_provider_specific_headers_to_request( diff --git a/tests/test_litellm/test_anthropic_beta_headers_filtering.py b/tests/test_litellm/test_anthropic_beta_headers_filtering.py index 3c967283abf..d404edb1281 100644 --- a/tests/test_litellm/test_anthropic_beta_headers_filtering.py +++ b/tests/test_litellm/test_anthropic_beta_headers_filtering.py @@ -18,6 +18,7 @@ import pytest import litellm from litellm.anthropic_beta_headers_manager import ( filter_and_transform_beta_headers, + update_headers_with_filtered_beta, update_request_with_filtered_beta, ) @@ -511,3 +512,20 @@ class TestAnthropicBetaHeadersFiltering: assert ( "unknown-header-123" not in filtered ), f"Unknown header should not be in result for {provider}" + + @pytest.mark.parametrize("provider", ["anthropic", "bedrock", "bedrock_mantle", "vertex_ai"]) + def test_blank_anthropic_beta_header_is_removed(self, provider): + headers = {"anthropic-beta": "", "anthropic-version": "2023-06-01"} + + assert update_headers_with_filtered_beta(headers, provider) == {"anthropic-version": "2023-06-01"} + + @pytest.mark.parametrize("provider", ["anthropic", "bedrock", "bedrock_mantle", "vertex_ai"]) + def test_whitespace_only_anthropic_beta_header_is_removed(self, provider): + headers = {"anthropic-beta": " , ", "anthropic-version": "2023-06-01"} + + assert update_headers_with_filtered_beta(headers, provider) == {"anthropic-version": "2023-06-01"} + + def test_absent_anthropic_beta_header_is_left_alone(self): + headers = {"anthropic-version": "2023-06-01"} + + assert update_headers_with_filtered_beta(headers, "bedrock_mantle") == {"anthropic-version": "2023-06-01"} diff --git a/tests/test_litellm/test_cost_calculator.py b/tests/test_litellm/test_cost_calculator.py index da3d022d669..1d6c229f9ce 100644 --- a/tests/test_litellm/test_cost_calculator.py +++ b/tests/test_litellm/test_cost_calculator.py @@ -3522,6 +3522,58 @@ def test_cost_per_token_region_name_applies_to_provider_prefixed_model(_local_mo ) +def test_completion_cost_mantle_native_messages_prices_claude_from_the_bedrock_row(_local_model_cost_map): + """Mantle's native Messages API answers with Anthropic's canonical model name and the proxy + resolves a Mantle region for every call, so the first cost candidate is + bedrock_mantle//claude-sonnet-5. That name has no row of its own and must fall through to + the deployment's bare Bedrock row instead of stopping on an unpriced capability rule at $0.""" + + response = litellm.ModelResponse( + id="msg_x", + choices=[{"index": 0, "message": {"role": "assistant", "content": "hi"}, "finish_reason": "stop"}], + model="claude-sonnet-5", + usage={"prompt_tokens": 100, "completion_tokens": 10, "total_tokens": 110}, + ) + row = litellm.model_cost["anthropic.claude-sonnet-5"] + expected = 100 * row["input_cost_per_token"] + 10 * row["output_cost_per_token"] + assert expected > 0 + + for region_name in ("us-east-1", None): + assert litellm.completion_cost( + completion_response=response, + model="bedrock_mantle/anthropic.claude-sonnet-5", + custom_llm_provider="bedrock_mantle", + region_name=region_name, + ) == pytest.approx(expected) + + +def test_completion_cost_mantle_native_messages_prices_haiku_from_the_mantle_row(_local_model_cost_map): + """Mantle serves Anthropic's un-versioned haiku id, which has no bare Bedrock row (Bedrock's carries + the -20251001-v1:0 suffix), and Claude Code sends every small-fast-model call to it. Both the plain + and the region-prefixed deployment names must price from bedrock_mantle/anthropic.claude-haiku-4-5 + instead of billing $0.""" + + response = litellm.ModelResponse( + id="msg_x", + choices=[{"index": 0, "message": {"role": "assistant", "content": "hi"}, "finish_reason": "stop"}], + model="claude-haiku-4-5", + usage={"prompt_tokens": 100, "completion_tokens": 10, "total_tokens": 110}, + ) + row = litellm.model_cost["bedrock_mantle/anthropic.claude-haiku-4-5"] + expected = 100 * row["input_cost_per_token"] + 10 * row["output_cost_per_token"] + assert expected > 0 + + for model in ( + "bedrock_mantle/anthropic.claude-haiku-4-5", + "bedrock_mantle/us-east-2/anthropic.claude-haiku-4-5", + ): + assert litellm.completion_cost( + completion_response=response, + model=model, + custom_llm_provider="bedrock_mantle", + ) == pytest.approx(expected), model + + def test_select_model_name_keeps_base_model_free_of_region(_local_model_cost_map): """An explicit base_model keeps pricing on that model's own key even when the request carries a region with different regional rates, so the private provider model never widens region pricing.""" diff --git a/tests/test_litellm/test_utils.py b/tests/test_litellm/test_utils.py index 8bdda0490c0..2ccb88b29db 100644 --- a/tests/test_litellm/test_utils.py +++ b/tests/test_litellm/test_utils.py @@ -1163,6 +1163,21 @@ def test_get_model_info_bedrock_regional_inference_profile_pricing(local_model_c assert control["key"] == "au.anthropic.claude-opus-4-8" +def test_get_model_info_bedrock_mantle_region_prefix_falls_back_to_the_mantle_row(local_model_cost_map): + """A Mantle deployment name may carry the region as a prefix (bedrock_mantle/us-east-2/). + That name has no cost row of its own, so pricing must fall through to the region-free + bedrock_mantle/ row instead of raising, while a region that has its own row keeps it.""" + for model, expected_key in ( + ("bedrock_mantle/us-east-2/anthropic.claude-haiku-4-5", "bedrock_mantle/anthropic.claude-haiku-4-5"), + ("bedrock_mantle/us-east-2/openai.gpt-5.6-sol", "bedrock_mantle/openai.gpt-5.6-sol"), + ("bedrock_mantle/us-gov-west-1/openai.gpt-5.4", "bedrock_mantle/us-gov-west-1/openai.gpt-5.4"), + ): + info = litellm.get_model_info(model=model, custom_llm_provider="bedrock_mantle") + assert info["key"] == expected_key, model + assert info["input_cost_per_token"] == litellm.model_cost[expected_key]["input_cost_per_token"], model + assert info["input_cost_per_token"] > 0, model + + def test_openai_models_in_model_info(monkeypatch): monkeypatch.setenv("LITELLM_LOCAL_MODEL_COST_MAP", "True") litellm.model_cost = litellm.get_model_cost_map(url="") @@ -3646,6 +3661,28 @@ class TestGetOptionalParamsTencent: assert isinstance(config, TencentAnthropicMessagesConfig) assert config.custom_llm_provider == "tencent" + def test_bedrock_mantle_claude_messages_config_routing(self): + import litellm + from litellm.llms.bedrock_mantle.messages.transformation import ( + BedrockMantleAnthropicMessagesConfig, + ) + + config = ProviderConfigManager.get_provider_anthropic_messages_config( + model="anthropic.claude-sonnet-5", + provider=litellm.LlmProviders.BEDROCK_MANTLE, + ) + assert isinstance(config, BedrockMantleAnthropicMessagesConfig) + assert config.custom_llm_provider == "bedrock_mantle" + + def test_bedrock_mantle_openai_models_keep_the_messages_bridge(self): + import litellm + + config = ProviderConfigManager.get_provider_anthropic_messages_config( + model="openai.gpt-5.6-sol", + provider=litellm.LlmProviders.BEDROCK_MANTLE, + ) + assert config is None + class TestValidateEnvironmentTencent: """Tests that validate_environment resolves TENCENT_API_KEY for the tencent provider."""