Messages API: add reasoning efforts support

docs/control/reference/rest_api/messages.md

@@ -27,6 +27,7 @@ Where `{endpoint_type}` is `chat`, `code`, `agent`, or `lang-graph`. See [URL Pa
 | `tools` | `array[ToolParam]` | No | Tool definitions the model may use. See [Tools](#tools). |
 | `tool_choice` | `ToolChoice` | No | How the model should use tools. See [Tool Choice](#tool-choice). |
 | `thinking` | `ThinkingConfig` | No | Extended thinking configuration. See [Thinking](#thinking). |
+| `reasoning_effort` | `string` | No | Reasoning effort for reasoning models: `"none"`, `"minimal"`, `"low"`, `"medium"`, `"high"`, `"xhigh"`. |
 | `stop_sequences` | `array[string]` | No | Custom stop sequences. |
 | `stream` | `boolean` | No | If `true`, stream the response via server-sent events. |
 | `output_config` | `OutputConfig` | No | Output format configuration (e.g. JSON schema). |
@@ -132,6 +133,7 @@ Extended thinking configuration. Discriminated by `type`:
 |------|--------|-------------|
 | `enabled` | `budget_tokens` (int, min 1024) | Enable thinking with a token budget. |
 | `disabled` | — | Disable thinking. |
+| `adaptive` | — | Claude automatically decides whether to use extended thinking based on request complexity. |
 
 ### Output Config
 

docs/release_notes.md

@@ -5,6 +5,20 @@ hide:
 
 # Release Notes
 
+## v0.4.2
+
+`time: 2026-03-11`
+
+| Product  | Version |
+| ---------|---------|
+| Control API | `60c3b428492056c3a31711d314a6978040f43774` |
+
+??? info "v0.4.2 Release Notes"
+
+    **Updates**
+    - Updated reasoning efforts to match Anthropic's adaptive thinking
+    - Add Dual-LLM prompt flavor 'code'
+
 ## v0.4.1
 
 `time: 2026-02-25`

src/sequrity/control/resources/messages.py

@@ -11,6 +11,7 @@
     MessageParam,
     MetadataParam,
     OutputConfigParam,
+    ReasoningEffort,
     TextBlockParam,
     ThinkingConfigParam,
     ToolChoiceParam,
@@ -43,6 +44,7 @@ def create(
         tools: list[ToolParam | dict] | None = None,
         tool_choice: ToolChoiceParam | dict | None = None,
         thinking: ThinkingConfigParam | dict | None = None,
+        reasoning_effort: ReasoningEffort | None = None,
         stop_sequences: list[str] | None = None,
         stream: Literal[True],
         output_config: OutputConfigParam | dict | None = None,
@@ -76,6 +78,7 @@ def create(
         tools: list[ToolParam | dict] | None = None,
         tool_choice: ToolChoiceParam | dict | None = None,
         thinking: ThinkingConfigParam | dict | None = None,
+        reasoning_effort: ReasoningEffort | None = None,
         stop_sequences: list[str] | None = None,
         stream: Literal[False] | None = None,
         output_config: OutputConfigParam | dict | None = None,
@@ -109,6 +112,7 @@ def create(
         tools: list[ToolParam | dict] | None = None,
         tool_choice: ToolChoiceParam | dict | None = None,
         thinking: ThinkingConfigParam | dict | None = None,
+        reasoning_effort: ReasoningEffort | None = None,
         stop_sequences: list[str] | None = None,
         stream: bool | None = None,
         output_config: OutputConfigParam | dict | None = None,
@@ -141,6 +145,7 @@ def create(
             tools: List of tool definitions available to the model.
             tool_choice: How the model should use the provided tools.
             thinking: Configuration for extended thinking.
+            reasoning_effort: Reasoning effort level for supported models.
             stop_sequences: Custom text sequences that cause the model to stop.
             stream: Whether to stream the response. When ``True``, returns a
                 :class:`SyncStream` of :class:`AnthropicStreamEvent` objects.
@@ -179,6 +184,7 @@ def create(
                 "tools": tools,
                 "tool_choice": tool_choice,
                 "thinking": thinking,
+                "reasoning_effort": reasoning_effort,
                 "top_k": top_k,
                 "top_p": top_p,
                 "output_config": output_config,
@@ -248,6 +254,7 @@ async def create(
         tools: list[ToolParam | dict] | None = None,
         tool_choice: ToolChoiceParam | dict | None = None,
         thinking: ThinkingConfigParam | dict | None = None,
+        reasoning_effort: ReasoningEffort | None = None,
         stop_sequences: list[str] | None = None,
         stream: Literal[True],
         output_config: OutputConfigParam | dict | None = None,
@@ -281,6 +288,7 @@ async def create(
         tools: list[ToolParam | dict] | None = None,
         tool_choice: ToolChoiceParam | dict | None = None,
         thinking: ThinkingConfigParam | dict | None = None,
+        reasoning_effort: ReasoningEffort | None = None,
         stop_sequences: list[str] | None = None,
         stream: Literal[False] | None = None,
         output_config: OutputConfigParam | dict | None = None,
@@ -313,6 +321,7 @@ async def create(
         tools: list[ToolParam | dict] | None = None,
         tool_choice: ToolChoiceParam | dict | None = None,
         thinking: ThinkingConfigParam | dict | None = None,
+        reasoning_effort: ReasoningEffort | None = None,
         stop_sequences: list[str] | None = None,
         stream: bool | None = None,
         output_config: OutputConfigParam | dict | None = None,
@@ -344,6 +353,7 @@ async def create(
                 "tools": tools,
                 "tool_choice": tool_choice,
                 "thinking": thinking,
+                "reasoning_effort": reasoning_effort,
                 "top_k": top_k,
                 "top_p": top_p,
                 "output_config": output_config,

src/sequrity/control/types/headers.py

@@ -478,7 +478,7 @@ class PllmPromptOverrides(BaseModel):
     model_config = ConfigDict(extra="forbid")
 
     flavor: PromptFlavor | str | None = Field(
-        default=None, description="Prompt template variant to use (e.g., 'universal')."
+        default=None, description="Prompt template variant to use (e.g., 'universal', 'code')."
     )
     version: PromptVersion | str | None = Field(
         default=None,

src/sequrity/types/chat_completion/request.py

@@ -325,7 +325,7 @@ class ResponseFormatJsonSchema(BaseModel):
 
 ResponseFormat = ResponseFormatText | ResponseFormatJsonObject | ResponseFormatJsonSchema
 
-ReasoningEffort = Literal["minimal", "low", "medium", "high"]
+ReasoningEffort = Literal["none", "minimal", "low", "medium", "high", "xhigh"]
 
 # =============================================================================
 # Main Request Class
@@ -347,7 +347,7 @@ class ChatCompletionRequest(BaseModel):
     # Optional fields
     reasoning_effort: ReasoningEffort | None = Field(
         default=None,
-        description="Constrains effort on reasoning for reasoning models. Supported values are 'minimal', 'low', 'medium', and 'high'.",
+        description="Constrains effort on reasoning for reasoning models. Supported values are 'none', 'minimal', 'low', 'medium', 'high', and 'xhigh'.",
     )
     response_format: ResponseFormat | None = Field(
         default=None,

src/sequrity/types/messages/request.py

@@ -343,6 +343,8 @@ class ThinkingConfigAdaptiveParam(BaseModel):
     Field(discriminator="type"),
 ]
 
+ReasoningEffort = Literal["none", "minimal", "low", "medium", "high", "xhigh"]
+
 
 # =============================================================================
 # Main Request Class
@@ -372,6 +374,10 @@ class AnthropicMessageRequest(BaseModel):
 
     # Optional fields
     metadata: MetadataParam | None = Field(default=None, description="An object describing metadata about the request.")
+    reasoning_effort: ReasoningEffort | None = Field(
+        default=None,
+        description="Constrains effort on reasoning for reasoning models. Supported values are 'none', 'minimal', 'low', 'medium', 'high', and 'xhigh'.",
+    )
     output_config: OutputConfigParam | None = Field(
         default=None, description="Configuration options for the model's output, such as the output format."
     )