[inkling] Let Anthropic thinking=disabled map to reasoning effort "none" (#33913)

Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>

python/sglang/srt/entrypoints/openai/serving_chat.py

@@ -2220,6 +2220,12 @@ class OpenAIServingChat(OpenAIServingBase):
             request.reasoning_effort = "medium" if enabled else "no_think"
             return
 
+        if self.reasoning_parser == "inkling":
+            # Effort-conditioned, not toggled: "none" (0.0) is the off switch.
+            if not enabled:
+                request.reasoning_effort = "none"
+                return
+
         config = self.template_manager.reasoning_config
         is_mistral = (config is not None and config.special_case == "mistral") or (
             config is None and self._reasoning_default_mode() == "mistral"

test/registered/unit/entrypoints/openai/test_serving_chat.py

@@ -3054,6 +3054,26 @@ class InklingReasoningEffortTest(unittest.TestCase):
         finally:
             env.clear()
 
+    def test_thinking_disabled_maps_to_no_thinking_effort(self):
+        """Bug regression: Inkling is an always-on parser, so Anthropic
+        thinking={"type": "disabled"} was rejected outright even though effort
+        "none" (0.0) expresses exactly that."""
+        serving = object.__new__(OpenAIServingChat)
+        serving.reasoning_parser = "inkling"
+        serving.template_manager = Mock(reasoning_config=None)
+        serving._reasoning_detector = Mock(reasoning_default="always")
+        request = ChatCompletionRequest(
+            model="test-model", messages=[{"role": "user", "content": "hi"}]
+        )
+
+        serving.apply_reasoning_enabled(request, False)
+        self.assertEqual(request.reasoning_effort, "none")
+
+        # Enabling leaves an effort set via output_config.effort alone.
+        request.reasoning_effort = "low"
+        serving.apply_reasoning_enabled(request, True)
+        self.assertEqual(request.reasoning_effort, "low")
+
     def test_serving_does_not_prefill_model_message(self):
         from sglang.srt.parser.inkling_tokenizer import INKLING_SPECIAL_TOKEN_IDS