fix(agents): disable summarization when max_tokens_before_summary is None

Fix SummarizationMiddleware to properly disable all summarization when
max_tokens_before_summary=None, as documented. Previously, summarization
would still occur based on message count when this parameter was None.

The fix adds an early return when max_tokens_before_summary is None,
ensuring summarization is completely disabled regardless of message count
or token count.

Adds test case with messages exceeding messages_to_keep to verify the fix.

Author: karamvirsingh1998

libs/langchain_v1/langchain/agents/middleware/summarization.py

@@ -104,11 +104,12 @@ def before_model(self, state: AgentState, runtime: Runtime) -> dict[str, Any] |
         messages = state["messages"]
         self._ensure_message_ids(messages)
 
+        # If max_tokens_before_summary is None, summarization is disabled
+        if self.max_tokens_before_summary is None:
+            return None
+
         total_tokens = self.token_counter(messages)
-        if (
-            self.max_tokens_before_summary is not None
-            and total_tokens < self.max_tokens_before_summary
-        ):
+        if total_tokens < self.max_tokens_before_summary:
             return None
 
         cutoff_index = self._find_safe_cutoff(messages)

libs/langchain_v1/tests/unit_tests/agents/test_middleware_agent.py

@@ -1056,12 +1056,18 @@ def test_summarization_middleware_no_summarization_cases() -> None:
     model = FakeToolCallingModel()
     middleware = SummarizationMiddleware(model=model, max_tokens_before_summary=1000)
 
-    # Test when summarization is disabled
+    # Test when summarization is disabled with few messages
     middleware_disabled = SummarizationMiddleware(model=model, max_tokens_before_summary=None)
     state = {"messages": [HumanMessage(content="Hello"), AIMessage(content="Hi")]}
     result = middleware_disabled.before_model(state, None)
     assert result is None
 
+    # Test when summarization is disabled with many messages (exceeding messages_to_keep)
+    many_messages = [HumanMessage(content=f"Message {i}") for i in range(25)]
+    state_many = {"messages": many_messages}
+    result = middleware_disabled.before_model(state_many, None)
+    assert result is None
+
     # Test when token count is below threshold
     def mock_token_counter(messages):
         return 500  # Below threshold