fix(llms): honor reasoning-model params in AzureOpenAIStructuredLLM (#5548)

Co-authored-by: ly-wang19 <ly-wang19@users.noreply.github.com>

mem0/llms/azure_openai_structured.py

@@ -72,16 +72,26 @@ class AzureOpenAIStructuredLLM(LLMBase):
 
         messages[-1]["content"] = user_prompt
 
+        is_reasoning = self._is_reasoning_model(self.config.model)
         params = {
             "model": self.config.model,
             "messages": messages,
-            "temperature": self.config.temperature,
-            "top_p": self.config.top_p,
         }
-        if self._uses_max_completion_tokens(self.config.model):
+        # Reasoning models (o1/o3/GPT-5 series) reject temperature/top_p; only
+        # forward the sampling params for non-reasoning models. Mirrors the
+        # reasoning-aware handling of OpenAIStructuredLLM (#5458).
+        if not is_reasoning:
+            params["temperature"] = self.config.temperature
+            params["top_p"] = self.config.top_p
+        # Reasoning models require max_completion_tokens rather than max_tokens.
+        if is_reasoning or self._uses_max_completion_tokens(self.config.model):
             params["max_completion_tokens"] = self.config.max_tokens
         else:
             params["max_tokens"] = self.config.max_tokens
+        if is_reasoning:
+            reasoning_effort = getattr(self.config, "reasoning_effort", None)
+            if reasoning_effort:
+                params["reasoning_effort"] = reasoning_effort
         if response_format:
             params["response_format"] = response_format
         if tools:

tests/llms/test_azure_openai_structured.py

@@ -29,6 +29,7 @@ class DummyConfig:
         max_tokens=256,
         top_p=1.0,
         http_client=None,
+        reasoning_effort=None,
     ):
         self.model = model
         self.azure_kwargs = azure_kwargs or DummyAzureKwargs()
@@ -36,6 +37,7 @@ class DummyConfig:
         self.max_tokens = max_tokens
         self.top_p = top_p
         self.http_client = http_client
+        self.reasoning_effort = reasoning_effort
 
 
 @mock.patch("mem0.llms.azure_openai_structured.AzureOpenAI")
@@ -209,3 +211,56 @@ def test_generate_response_with_tools_no_tool_calls(mock_azure_openai):
 
     assert response["content"] == "No tools needed."
     assert response["tool_calls"] == []
+
+
+@mock.patch("mem0.llms.azure_openai_structured.AzureOpenAI")
+def test_reasoning_model_drops_sampling_params(mock_azure_openai):
+    """Reasoning models (o1/o3/GPT-5) reject temperature/max_tokens/top_p."""
+    mock_client = Mock()
+    mock_azure_openai.return_value = mock_client
+
+    config = DummyConfig(
+        model="o3-mini",
+        azure_kwargs=DummyAzureKwargs(api_key="real-key"),
+        reasoning_effort="low",
+    )
+    llm = AzureOpenAIStructuredLLM(config)
+
+    mock_response = Mock()
+    mock_response.choices = [Mock(message=Mock(content="ok"))]
+    mock_client.chat.completions.create.return_value = mock_response
+
+    llm.generate_response([{"role": "user", "content": "Hi"}])
+
+    call_kwargs = mock_client.chat.completions.create.call_args[1]
+    assert "temperature" not in call_kwargs  # reasoning models reject these
+    assert "max_tokens" not in call_kwargs
+    assert "top_p" not in call_kwargs
+    assert call_kwargs["reasoning_effort"] == "low"
+    assert call_kwargs["model"] == "o3-mini"
+
+
+@mock.patch("mem0.llms.azure_openai_structured.AzureOpenAI")
+def test_regular_model_sends_sampling_params(mock_azure_openai):
+    """Regular models still receive the standard sampling params."""
+    mock_client = Mock()
+    mock_azure_openai.return_value = mock_client
+
+    config = DummyConfig(
+        model="gpt-4o",
+        azure_kwargs=DummyAzureKwargs(api_key="real-key"),
+        temperature=0.3,
+    )
+    llm = AzureOpenAIStructuredLLM(config)
+
+    mock_response = Mock()
+    mock_response.choices = [Mock(message=Mock(content="ok"))]
+    mock_client.chat.completions.create.return_value = mock_response
+
+    llm.generate_response([{"role": "user", "content": "Hi"}])
+
+    call_kwargs = mock_client.chat.completions.create.call_args[1]
+    assert call_kwargs["temperature"] == 0.3
+    assert "max_tokens" in call_kwargs  # standard sampling params still forwarded
+    assert "top_p" in call_kwargs
+    assert call_kwargs["model"] == "gpt-4o"