anthropic 0.77.0__py3-none-any.whl → 0.78.0__py3-none-any.whl

anthropic/_version.py

@@ -1,4 +1,4 @@
 # File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
 __title__ = "anthropic"
-__version__ = "0.77.0"  # x-release-please-version
+__version__ = "0.78.0"  # x-release-please-version

anthropic/lib/streaming/_beta_messages.py

@@ -20,6 +20,7 @@ from ._beta_types import (
     BetaThinkingEvent,
     BetaInputJsonEvent,
     BetaSignatureEvent,
+    BetaCompactionEvent,
     ParsedBetaTextEvent,
     ParsedBetaMessageStopEvent,
     ParsedBetaMessageStreamEvent,
@@ -403,6 +404,15 @@ def build_events(
                     )
                 )
             pass
+        elif event.delta.type == "compaction_delta":
+            if content_block.type == "compaction":
+                events_to_fire.append(
+                    build(
+                        BetaCompactionEvent,
+                        type="compaction",
+                        content=content_block.content,
+                    )
+                )
         else:
             # we only want exhaustive checking for linters, not at runtime
             if TYPE_CHECKING:  # type: ignore[unreachable]
@@ -512,6 +522,9 @@ def accumulate_event(
         elif event.delta.type == "signature_delta":
             if content.type == "thinking":
                 content.signature = event.delta.signature
+        elif event.delta.type == "compaction_delta":
+            if content.type == "compaction":
+                content.content = event.delta.content
         else:
             # we only want exhaustive checking for linters, not at runtime
             if TYPE_CHECKING:  # type: ignore[unreachable]
@@ -536,5 +549,7 @@ def accumulate_event(
             current_snapshot.usage.cache_read_input_tokens = event.usage.cache_read_input_tokens
         if event.usage.server_tool_use is not None:
             current_snapshot.usage.server_tool_use = event.usage.server_tool_use
+        if event.usage.iterations is not None:
+            current_snapshot.usage.iterations = event.usage.iterations
 
     return current_snapshot

anthropic/lib/streaming/_beta_types.py

@@ -75,6 +75,13 @@ class BetaInputJsonEvent(BaseModel):
     """
 
 
+class BetaCompactionEvent(BaseModel):
+    type: Literal["compaction"]
+
+    content: Union[str, None]
+    """The compaction content"""
+
+
 class ParsedBetaMessageStopEvent(BetaRawMessageStopEvent, GenericModel, Generic[ResponseFormatT]):
     type: Literal["message_stop"]
 
@@ -97,6 +104,7 @@ ParsedBetaMessageStreamEvent = Annotated[
         BetaThinkingEvent,
         BetaSignatureEvent,
         BetaInputJsonEvent,
+        BetaCompactionEvent,
         BetaRawMessageStartEvent,
         BetaRawMessageDeltaEvent,
         ParsedBetaMessageStopEvent[ResponseFormatT],

anthropic/resources/beta/messages/messages.py

@@ -44,7 +44,7 @@ from ...._exceptions import AnthropicError
 from ...._base_client import make_request_options
 from ...._utils._utils import is_dict
 from ....lib.streaming import BetaMessageStreamManager, BetaAsyncMessageStreamManager
-from ...messages.messages import DEPRECATED_MODELS
+from ...messages.messages import DEPRECATED_MODELS, MODELS_TO_WARN_WITH_THINKING_ENABLED
 from ....types.model_param import ModelParam
 from ....lib._parse._response import ResponseFormatT, parse_beta_response
 from ....lib._parse._transform import transform_schema
@@ -112,6 +112,7 @@ class Messages(SyncAPIResource):
         model: ModelParam,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -230,6 +231,9 @@ class Messages(SyncAPIResource):
               This allows you to control how Claude manages context across multiple requests,
               such as whether to clear function results or not.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           mcp_servers: MCP servers to be utilized in this request
 
           metadata: An object describing metadata about the request.
@@ -406,6 +410,7 @@ class Messages(SyncAPIResource):
         stream: Literal[True],
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -527,6 +532,9 @@ class Messages(SyncAPIResource):
               This allows you to control how Claude manages context across multiple requests,
               such as whether to clear function results or not.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           mcp_servers: MCP servers to be utilized in this request
 
           metadata: An object describing metadata about the request.
@@ -699,6 +707,7 @@ class Messages(SyncAPIResource):
         stream: bool,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -820,6 +829,9 @@ class Messages(SyncAPIResource):
               This allows you to control how Claude manages context across multiple requests,
               such as whether to clear function results or not.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           mcp_servers: MCP servers to be utilized in this request
 
           metadata: An object describing metadata about the request.
@@ -991,6 +1003,7 @@ class Messages(SyncAPIResource):
         model: ModelParam,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -1029,6 +1042,13 @@ class Messages(SyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         merged_output_config = _merge_output_configs(output_config, output_format)
 
         extra_headers = {
@@ -1044,6 +1064,7 @@ class Messages(SyncAPIResource):
                     "model": model,
                     "container": container,
                     "context_management": context_management,
+                    "inference_geo": inference_geo,
                     "mcp_servers": mcp_servers,
                     "metadata": metadata,
                     "output_config": merged_output_config,
@@ -1079,6 +1100,7 @@ class Messages(SyncAPIResource):
         model: ModelParam,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -1116,9 +1138,16 @@ class Messages(SyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         betas = [beta for beta in betas] if is_given(betas) else []
 
-        if "structured-outputs-2025-12-15" not in betas and is_given(output_format):
+        if "structured-outputs-2025-12-15" not in betas:
             # Ensure structured outputs beta is included for parse method
             betas.append("structured-outputs-2025-12-15")
 
@@ -1166,6 +1195,7 @@ class Messages(SyncAPIResource):
                     "model": model,
                     "container": container,
                     "context_management": context_management,
+                    "inference_geo": inference_geo,
                     "mcp_servers": mcp_servers,
                     "metadata": metadata,
                     "output_config": merged_output_config,
@@ -1205,6 +1235,7 @@ class Messages(SyncAPIResource):
         compaction_control: CompactionControl | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         max_iterations: int | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
@@ -1241,6 +1272,7 @@ class Messages(SyncAPIResource):
         max_iterations: int | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -1275,6 +1307,7 @@ class Messages(SyncAPIResource):
         max_iterations: int | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -1307,6 +1340,7 @@ class Messages(SyncAPIResource):
         max_iterations: int | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -1339,6 +1373,20 @@ class Messages(SyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         extra_headers = {
             "X-Stainless-Helper": "BetaToolRunner",
             **strip_not_given({"anthropic-beta": ",".join(str(e) for e in betas) if is_given(betas) else NOT_GIVEN}),
@@ -1362,6 +1410,7 @@ class Messages(SyncAPIResource):
                 "model": model,
                 "container": container,
                 "context_management": context_management,
+                "inference_geo": inference_geo,
                 "mcp_servers": mcp_servers,
                 "metadata": metadata,
                 "output_config": output_config,
@@ -1414,6 +1463,7 @@ class Messages(SyncAPIResource):
         model: ModelParam,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -1445,6 +1495,13 @@ class Messages(SyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         """Create a Message stream"""
         extra_headers = {
             "X-Stainless-Helper-Method": "stream",
@@ -1488,6 +1545,7 @@ class Messages(SyncAPIResource):
                     "output_format": omit,
                     "container": container,
                     "context_management": context_management,
+                    "inference_geo": inference_geo,
                     "mcp_servers": mcp_servers,
                     "service_tier": service_tier,
                     "stop_sequences": stop_sequences,
@@ -1808,6 +1866,7 @@ class AsyncMessages(AsyncAPIResource):
         model: ModelParam,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -1926,6 +1985,9 @@ class AsyncMessages(AsyncAPIResource):
               This allows you to control how Claude manages context across multiple requests,
               such as whether to clear function results or not.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           mcp_servers: MCP servers to be utilized in this request
 
           metadata: An object describing metadata about the request.
@@ -2102,6 +2164,7 @@ class AsyncMessages(AsyncAPIResource):
         stream: Literal[True],
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -2223,6 +2286,9 @@ class AsyncMessages(AsyncAPIResource):
               This allows you to control how Claude manages context across multiple requests,
               such as whether to clear function results or not.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           mcp_servers: MCP servers to be utilized in this request
 
           metadata: An object describing metadata about the request.
@@ -2395,6 +2461,7 @@ class AsyncMessages(AsyncAPIResource):
         stream: bool,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -2516,6 +2583,9 @@ class AsyncMessages(AsyncAPIResource):
               This allows you to control how Claude manages context across multiple requests,
               such as whether to clear function results or not.
 
+          inference_geo: Specifies the geographic region for inference processing. If not specified, the
+              workspace's `default_inference_geo` is used.
+
           mcp_servers: MCP servers to be utilized in this request
 
           metadata: An object describing metadata about the request.
@@ -2687,6 +2757,7 @@ class AsyncMessages(AsyncAPIResource):
         model: ModelParam,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -2725,6 +2796,13 @@ class AsyncMessages(AsyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         merged_output_config = _merge_output_configs(output_config, output_format)
 
         extra_headers = {
@@ -2740,6 +2818,7 @@ class AsyncMessages(AsyncAPIResource):
                     "model": model,
                     "container": container,
                     "context_management": context_management,
+                    "inference_geo": inference_geo,
                     "mcp_servers": mcp_servers,
                     "metadata": metadata,
                     "output_config": merged_output_config,
@@ -2775,6 +2854,7 @@ class AsyncMessages(AsyncAPIResource):
         model: ModelParam,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -2811,9 +2891,16 @@ class AsyncMessages(AsyncAPIResource):
                 DeprecationWarning,
                 stacklevel=3,
             )
+
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
         betas = [beta for beta in betas] if is_given(betas) else []
 
-        if "structured-outputs-2025-12-15" not in betas and is_given(output_format):
+        if "structured-outputs-2025-12-15" not in betas:
             # Ensure structured outputs beta is included for parse method
             betas.append("structured-outputs-2025-12-15")
 
@@ -2861,6 +2948,7 @@ class AsyncMessages(AsyncAPIResource):
                     "model": model,
                     "container": container,
                     "context_management": context_management,
+                    "inference_geo": inference_geo,
                     "mcp_servers": mcp_servers,
                     "output_config": merged_output_config,
                     "metadata": metadata,
@@ -2901,6 +2989,7 @@ class AsyncMessages(AsyncAPIResource):
         max_iterations: int | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -2936,6 +3025,7 @@ class AsyncMessages(AsyncAPIResource):
         max_iterations: int | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -2970,6 +3060,7 @@ class AsyncMessages(AsyncAPIResource):
         max_iterations: int | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -3002,6 +3093,7 @@ class AsyncMessages(AsyncAPIResource):
         max_iterations: int | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         metadata: BetaMetadataParam | Omit = omit,
         output_config: BetaOutputConfigParam | Omit = omit,
@@ -3034,6 +3126,13 @@ class AsyncMessages(AsyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         extra_headers = {
             "X-Stainless-Helper": "BetaToolRunner",
             **strip_not_given({"anthropic-beta": ",".join(str(e) for e in betas) if is_given(betas) else NOT_GIVEN}),
@@ -3057,6 +3156,7 @@ class AsyncMessages(AsyncAPIResource):
                 "model": model,
                 "container": container,
                 "context_management": context_management,
+                "inference_geo": inference_geo,
                 "mcp_servers": mcp_servers,
                 "metadata": metadata,
                 "output_config": output_config,
@@ -3112,6 +3212,7 @@ class AsyncMessages(AsyncAPIResource):
         output_format: None | type[ResponseFormatT] | BetaJSONOutputFormatParam | Omit = omit,
         container: Optional[message_create_params.Container] | Omit = omit,
         context_management: Optional[BetaContextManagementConfigParam] | Omit = omit,
+        inference_geo: Optional[str] | Omit = omit,
         mcp_servers: Iterable[BetaRequestMCPServerURLDefinitionParam] | Omit = omit,
         service_tier: Literal["auto", "standard_only"] | Omit = omit,
         stop_sequences: SequenceNotStr[str] | Omit = omit,
@@ -3140,6 +3241,13 @@ class AsyncMessages(AsyncAPIResource):
                 stacklevel=3,
             )
 
+        if model in MODELS_TO_WARN_WITH_THINKING_ENABLED and thinking and thinking["type"] == "enabled":
+            warnings.warn(
+                f"Using Claude with {model} and 'thinking.type=enabled' is deprecated. Use 'thinking.type=adaptive' instead which results in better model performance in our testing: https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking",
+                UserWarning,
+                stacklevel=3,
+            )
+
         extra_headers = {
             "X-Stainless-Helper-Method": "stream",
             "X-Stainless-Stream-Helper": "beta.messages",
@@ -3181,6 +3289,7 @@ class AsyncMessages(AsyncAPIResource):
                     "output_format": omit,
                     "container": container,
                     "context_management": context_management,
+                    "inference_geo": inference_geo,
                     "mcp_servers": mcp_servers,
                     "service_tier": service_tier,
                     "stop_sequences": stop_sequences,