fix: route minimax m3 reasoning controls through profile

Follow up PR #46609's api.minimax.io reasoning report by moving the behavior out of the broad run_agent host gate and into the MiniMax provider profile. Only MiniMax-M3 on the documented OpenAI-compatible /v1 route gets reasoning_split/thinking/reasoning_effort; Anthropic-format MiniMax and non-M3 models keep their existing wire shapes.

Co-authored-by: goku94123 <gooku94123@gmail.com>

tests/plugins/model_providers/test_minimax_profile.py

@@ -118,3 +118,115 @@ class TestMinimaxAuxModelNotHighspeed:
             "is a -highspeed variant — that costs 2x for the same model and "
             "broke #4082 the first time. Revert to plain M2.7 or M3."
         )
+
+
+class TestMinimaxM3OpenAIReasoningWireShape:
+    """MiniMax-M3 on api.minimax.io/v1 gets MiniMax's OpenAI-compatible knobs."""
+
+    def test_m3_openai_route_requests_reasoning_split_by_default(self):
+        import model_tools  # noqa: F401
+        import providers
+
+        profile = providers.get_provider_profile("minimax")
+        assert profile is not None
+        extra_body, top_level = profile.build_api_kwargs_extras(
+            reasoning_config=None,
+            model="MiniMax-M3",
+            base_url="https://api.minimax.io/v1",
+        )
+        assert extra_body == {"reasoning_split": True}
+        assert top_level == {}
+
+    def test_m3_openai_route_maps_explicit_effort_to_adaptive_only(self):
+        import model_tools  # noqa: F401
+        import providers
+
+        profile = providers.get_provider_profile("minimax")
+        assert profile is not None
+        extra_body, top_level = profile.build_api_kwargs_extras(
+            reasoning_config={"enabled": True, "effort": "high"},
+            model="MiniMax-M3",
+            base_url="https://api.minimax.io/v1",
+        )
+        assert extra_body == {
+            "reasoning_split": True,
+            "thinking": {"type": "adaptive"},
+        }
+        assert top_level == {}
+
+    def test_m3_openai_route_does_not_send_reasoning_effort(self):
+        import model_tools  # noqa: F401
+        import providers
+
+        profile = providers.get_provider_profile("minimax")
+        assert profile is not None
+        extra_body, _top_level = profile.build_api_kwargs_extras(
+            reasoning_config={"enabled": True, "effort": "xhigh"},
+            model="MiniMax-M3",
+            base_url="https://api.minimax.io/v1/",
+        )
+        assert extra_body == {
+            "reasoning_split": True,
+            "thinking": {"type": "adaptive"},
+        }
+
+    def test_m3_openai_route_can_disable_thinking(self):
+        import model_tools  # noqa: F401
+        import providers
+
+        profile = providers.get_provider_profile("minimax")
+        assert profile is not None
+        extra_body, top_level = profile.build_api_kwargs_extras(
+            reasoning_config={"enabled": False, "effort": "high"},
+            model="MiniMax-M3",
+            base_url="https://api.minimax.io/v1",
+        )
+        assert extra_body == {
+            "reasoning_split": True,
+            "thinking": {"type": "disabled"},
+        }
+        assert top_level == {}
+
+    @pytest.mark.parametrize(
+        "model,base_url",
+        [
+            ("MiniMax-M2.7", "https://api.minimax.io/v1"),
+            ("MiniMax-M3", "https://api.minimax.io/anthropic"),
+            ("MiniMax-M3", "https://api.minimaxi.com/v1"),
+        ],
+    )
+    def test_non_m3_or_non_global_openai_routes_emit_no_openai_reasoning_knobs(
+        self, model, base_url
+    ):
+        import model_tools  # noqa: F401
+        import providers
+
+        profile = providers.get_provider_profile("minimax")
+        assert profile is not None
+        extra_body, top_level = profile.build_api_kwargs_extras(
+            reasoning_config={"enabled": True, "effort": "high"},
+            model=model,
+            base_url=base_url,
+        )
+        assert extra_body == {}
+        assert top_level == {}
+
+    def test_transport_threads_base_url_to_profile(self):
+        import model_tools  # noqa: F401
+        import providers
+        from agent.transports.chat_completions import ChatCompletionsTransport
+
+        profile = providers.get_provider_profile("minimax")
+        assert profile is not None
+        kwargs = ChatCompletionsTransport().build_kwargs(
+            model="MiniMax-M3",
+            messages=[{"role": "user", "content": "ping"}],
+            tools=None,
+            provider_profile=profile,
+            reasoning_config={"enabled": True, "effort": "medium"},
+            base_url="https://api.minimax.io/v1",
+        )
+        assert kwargs["extra_body"] == {
+            "reasoning_split": True,
+            "thinking": {"type": "adaptive"},
+        }