feat(config): accept STRIX_REASONING_EFFORT=max for providers that support it (#956)

Co-authored-by: Ahmed Allam <allam@usestrix.com>

strix/config/models.py

@@ -83,10 +83,13 @@ class _CodexResponsesModel(OpenAIResponsesModel):
         effort = self._reasoning_effort
         if effort and effort != "none":
             # Clamp to efforts the backend accepts.
-            if effort == "minimal":
-                effort = "low"
-            elif effort == "xhigh":
-                effort = "high"
+            match effort:
+                case "minimal":
+                    effort = "low"
+                case "xhigh" | "max":
+                    effort = "high"
+                case _:
+                    pass
             overrides = overrides.resolve(ModelSettings(reasoning=Reasoning(effort=effort)))
         return model_settings.resolve(overrides)
 

strix/config/settings.py

@@ -8,7 +8,7 @@ from pydantic import AliasChoices, Field
 from pydantic_settings import BaseSettings, SettingsConfigDict
 
 
-ReasoningEffort = Literal["none", "minimal", "low", "medium", "high", "xhigh"]
+ReasoningEffort = Literal["none", "minimal", "low", "medium", "high", "xhigh", "max"]
 
 _BASE_CONFIG = SettingsConfigDict(
     case_sensitive=False,

strix/core/inputs.py

@@ -147,7 +147,7 @@ def make_model_settings(
         and model_supports_reasoning(model_name)
     ):
         model_settings = model_settings.resolve(
-            ModelSettings(reasoning=Reasoning(effort=reasoning_effort)),
+            _reasoning_settings(reasoning_effort, model_settings.extra_args),
         )
     if force_required_tool_choice and _accepts_required_tool_choice(model_name):
         model_settings = model_settings.resolve(ModelSettings(tool_choice="required"))
@@ -162,6 +162,22 @@ def make_model_settings(
     return model_settings
 
 
+def _reasoning_settings(
+    effort: ReasoningEffort,
+    extra_args: dict[str, Any] | None,
+) -> ModelSettings:
+    """``max`` is not in the OpenAI SDK's ``Reasoning.effort`` enum, so send it as
+    a raw body field instead — also keeping it clear of LiteLLM's DeepSeek mapping,
+    which collapses every ``reasoning_effort`` level to plain thinking-enabled.
+    Providers that don't support ``max`` reject the request.
+    """
+    if effort != "max":
+        return ModelSettings(reasoning=Reasoning(effort=effort))
+    return ModelSettings(
+        extra_args={**(extra_args or {}), "extra_body": {"reasoning_effort": "max"}},
+    )
+
+
 def _prompt_cache_extra_args(model_name: str) -> dict[str, Any] | None:
     """LiteLLM ``cache_control_injection_points`` for Claude prompt caching.
 

strix/interface/main.py

@@ -171,8 +171,8 @@ def validate_environment() -> None:
                     error_text.append("• ", style="white")
                     error_text.append("STRIX_REASONING_EFFORT", style="bold cyan")
                     error_text.append(
-                        " - Reasoning effort level: none, minimal, low, medium, high, xhigh "
-                        "(default: high)\n",
+                        " - Reasoning effort level: none, minimal, low, medium, high, xhigh, "
+                        "max (default: high)\n",
                         style="white",
                     )