fix(memory): accept llm kwarg in sync Memory.add()/_create_procedural_memory (#5911) (#5953)

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

mem0/memory/main.py

@@ -726,6 +726,7 @@ class Memory(MemoryBase):
         infer: bool = True,
         memory_type: Optional[str] = None,
         prompt: Optional[str] = None,
+        llm=None,
     ):
         """
         Create a new memory.
@@ -802,7 +803,7 @@ class Memory(MemoryBase):
             )
 
         if agent_id is not None and memory_type == MemoryType.PROCEDURAL.value:
-            results = self._create_procedural_memory(messages, metadata=processed_metadata, prompt=prompt)
+            results = self._create_procedural_memory(messages, metadata=processed_metadata, prompt=prompt, llm=llm)
             scale_threshold_notice = detect_scale_threshold_from_add_result(self, results)
             if temporal_usage_notice:
                 display_temporal_usage_notice(self, "sync", "add", *temporal_usage_notice)
@@ -1910,15 +1911,27 @@ class Memory(MemoryBase):
         )
         return memory_id
 
-    def _create_procedural_memory(self, messages, metadata=None, prompt=None):
+    def _create_procedural_memory(self, messages, metadata=None, llm=None, prompt=None):
         """
         Create a procedural memory
 
         Args:
             messages (list): List of messages to create a procedural memory from.
             metadata (dict): Metadata to create a procedural memory from.
+            llm (llm, optional): LLM to use for the procedural memory creation. Defaults to None.
             prompt (str, optional): Prompt to use for the procedural memory creation. Defaults to None.
         """
+        if llm is not None:
+            try:
+                from langchain_core.messages.utils import (
+                    convert_to_messages,  # type: ignore
+                )
+            except Exception:
+                logger.error(
+                    "Import error while loading langchain-core. Please install 'langchain-core' to use procedural memory."
+                )
+                raise
+
         logger.info("Creating procedural memory")
 
         parsed_messages = [
@@ -1931,8 +1944,12 @@ class Memory(MemoryBase):
         ]
 
         try:
-            procedural_memory = self.llm.generate_response(messages=parsed_messages)
-            procedural_memory = remove_code_blocks(procedural_memory)
+            if llm is not None:
+                response = llm.invoke(input=convert_to_messages(parsed_messages))
+                procedural_memory = remove_code_blocks(response.content)
+            else:
+                procedural_memory = self.llm.generate_response(messages=parsed_messages)
+                procedural_memory = remove_code_blocks(procedural_memory)
         except Exception as e:
             logger.error(f"Error generating procedural memory summary: {e}")
             raise

tests/test_memory.py

@@ -1529,3 +1529,85 @@ async def test_async_procedural_memory_langchain_strips_code_blocks(mock_llm_fac
     insert_call = memory.vector_store.insert.call_args
     stored_data = insert_call[1]["payloads"][0]["data"]
     assert "```" not in stored_data
+
+
+@patch("mem0.memory.main.VectorStoreFactory")
+@patch("mem0.memory.main.EmbedderFactory")
+@patch("mem0.memory.main.LlmFactory")
+def test_sync_procedural_memory_accepts_llm_kwarg_and_strips_code_blocks(
+    mock_llm_factory, mock_emb, mock_vs
+):
+    """Regression #5911: sync Memory.add()/Memory._create_procedural_memory must
+    accept the same `llm` kwarg as the async path and route through llm.invoke()."""
+    mock_vs.return_value = MagicMock()
+    mock_emb.return_value = MagicMock()
+    mock_emb.return_value.embed.return_value = [0.1] * 1536
+    mock_llm_factory.return_value = MagicMock()
+
+    from mem0.memory.main import Memory
+
+    config = MemoryConfig()
+    memory = Memory(config)
+    memory.vector_store = MagicMock()
+    memory.vector_store.insert = MagicMock()
+
+    mock_langchain_llm = MagicMock()
+    mock_response = MagicMock()
+    mock_response.content = '```json\n{"key": "value"}\n```'
+    mock_langchain_llm.invoke.return_value = mock_response
+
+    messages = [{"role": "user", "content": "test"}]
+    metadata = {"user_id": "test_user"}
+
+    memory._create_procedural_memory(messages, metadata=metadata, llm=mock_langchain_llm)
+
+    mock_langchain_llm.invoke.assert_called_once()
+    insert_call = memory.vector_store.insert.call_args
+    stored_data = insert_call[1]["payloads"][0]["data"]
+    assert "```" not in stored_data
+
+
+@patch("mem0.memory.main.VectorStoreFactory")
+@patch("mem0.memory.main.EmbedderFactory")
+@patch("mem0.memory.main.LlmFactory")
+def test_sync_memory_add_forwards_llm_to_procedural_memory(
+    mock_llm_factory, mock_emb, mock_vs
+):
+    """Regression #5911: Memory.add() must forward `llm` to _create_procedural_memory."""
+    mock_vs.return_value = MagicMock()
+    mock_emb.return_value = MagicMock()
+    mock_emb.return_value.embed.return_value = [0.1] * 1536
+    mock_llm_factory.return_value = MagicMock()
+
+    from mem0.memory.main import Memory
+
+    config = MemoryConfig()
+    memory = Memory(config)
+    memory.vector_store = MagicMock()
+    memory.vector_store.insert = MagicMock()
+
+    mock_langchain_llm = MagicMock()
+    mock_response = MagicMock()
+    mock_response.content = "procedural summary"
+    mock_langchain_llm.invoke.return_value = mock_response
+
+    forwarded = {"called": False}
+    original_create = memory._create_procedural_memory
+
+    def spy(*args, **kwargs):
+        forwarded["called"] = True
+        forwarded["llm"] = kwargs.get("llm")
+        return original_create(*args, **kwargs)
+
+    memory._create_procedural_memory = spy
+
+    memory.add(
+        [{"role": "user", "content": "Always verify inputs before processing."}],
+        agent_id="agent-1",
+        memory_type="procedural_memory",
+        llm=mock_langchain_llm,
+    )
+
+    assert forwarded["called"] is True
+    assert forwarded["llm"] is mock_langchain_llm
+    mock_langchain_llm.invoke.assert_called_once()