The Memory Layer for AI Agents - Drop-in memory infrastructure for AI agents and apps. Context that persists. Built for production.
Python agentic-memoryagentic-memory-systemagentsaiai-agentschatgptgenaillmlong-term-memorymemorymemory-managementpythonragstate-management
fix(llms): honor reasoning-model params in AzureOpenAIStructuredLLM (#5548)
ly-wang19a8eeddde6481
diff --git a/mem0/llms/azure_openai_structured.py b/mem0/llms/azure_openai_structured.pyindex 46b88a17c..e9eaf3471 100644--- a/mem0/llms/azure_openai_structured.py+++ b/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_tokenselse: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_effortif response_format:params["response_format"] = response_formatif tools:diff --git a/tests/llms/test_azure_openai_structured.py b/tests/llms/test_azure_openai_structured.pyindex 754a6bd1a..e910ff073 100644--- a/tests/llms/test_azure_openai_structured.py+++ b/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 = modelself.azure_kwargs = azure_kwargs or DummyAzureKwargs()@@ -36,6 +37,7 @@ class DummyConfig:self.max_tokens = max_tokensself.top_p = top_pself.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"] == []++[email protected]("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"++[email protected]("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"