fix(transformation.py): correctly translate 'thinking' param for lite… (#9904)

* fix(transformation.py): correctly translate 'thinking' param for litellm_proxy/ route

Fixes #9892

* test: update test

Author: krrishdholakia

litellm/llms/litellm_proxy/chat/transformation.py

@@ -15,6 +15,21 @@ def get_supported_openai_params(self, model: str) -> List:
         list.append("thinking")
         return list
 
+    def _map_openai_params(
+        self,
+        non_default_params: dict,
+        optional_params: dict,
+        model: str,
+        drop_params: bool,
+    ) -> dict:
+        supported_openai_params = self.get_supported_openai_params(model)
+        for param, value in non_default_params.items():
+            if param == "thinking":
+                optional_params.setdefault("extra_body", {})["thinking"] = value
+            elif param in supported_openai_params:
+                optional_params[param] = value
+        return optional_params
+
     def _get_openai_compatible_provider_info(
         self, api_base: Optional[str], api_key: Optional[str]
     ) -> Tuple[Optional[str], Optional[str]]:

tests/llm_translation/test_litellm_proxy_provider.py

@@ -449,3 +449,19 @@ def test_litellm_gateway_from_sdk_with_response_cost_in_additional_headers():
         )
 
         assert response._hidden_params["response_cost"] == 120
+
+
+def test_litellm_gateway_from_sdk_with_thinking_param():
+    try: 
+        response = litellm.completion(
+            model="litellm_proxy/anthropic.claude-3-7-sonnet-20250219-v1:0",
+            messages=[{"role": "user", "content": "Hello world"}],
+            api_base="http://0.0.0.0:4000",
+            api_key="sk-PIp1h0RekR",
+            # client=openai_client,
+            thinking={"type": "enabled", "max_budget": 100},
+        )
+        pytest.fail("Expected an error to be raised")
+    except Exception as e:
+        assert "Connection error." in str(e)
+    

tests/llm_translation/test_optional_params.py

@@ -1410,7 +1410,7 @@ def test_litellm_proxy_thinking_param():
         custom_llm_provider="litellm_proxy",
         thinking={"type": "enabled", "budget_tokens": 1024},
     )
-    assert optional_params["thinking"] == {"type": "enabled", "budget_tokens": 1024}
+    assert optional_params["extra_body"]["thinking"] == {"type": "enabled", "budget_tokens": 1024}
 
 def test_gemini_modalities_param():
     optional_params = get_optional_params(