Skip to content

The agent engineering platform.

feat(langchain): support async summarization in SummarizationMiddleware (#33918)

ccurme68643153e525
diff --git a/libs/langchain_v1/langchain/agents/middleware/summarization.py b/libs/langchain_v1/langchain/agents/middleware/summarization.py
index 4f1853a0cd..5bc3c08fcc 100644
--- a/libs/langchain_v1/langchain/agents/middleware/summarization.py
+++ b/libs/langchain_v1/langchain/agents/middleware/summarization.py
@@ -183,6 +183,33 @@ class SummarizationMiddleware(AgentMiddleware):
]
}
+ async def abefore_model(self, state: AgentState, runtime: Runtime) -> dict[str, Any] | None: # noqa: ARG002
+ """Process messages before model invocation, potentially triggering summarization."""
+ messages = state["messages"]
+ self._ensure_message_ids(messages)
+
+ total_tokens = self.token_counter(messages)
+ if not self._should_summarize(messages, total_tokens):
+ return None
+
+ cutoff_index = self._determine_cutoff_index(messages)
+
+ if cutoff_index <= 0:
+ return None
+
+ messages_to_summarize, preserved_messages = self._partition_messages(messages, cutoff_index)
+
+ summary = await self._acreate_summary(messages_to_summarize)
+ new_messages = self._build_new_messages(summary)
+
+ return {
+ "messages": [
+ RemoveMessage(id=REMOVE_ALL_MESSAGES),
+ *new_messages,
+ *preserved_messages,
+ ]
+ }
+
def _should_summarize(self, messages: list[AnyMessage], total_tokens: int) -> bool:
"""Determine whether summarization should run for the current token usage."""
if not self._trigger_conditions:
@@ -405,6 +432,23 @@ class SummarizationMiddleware(AgentMiddleware):
except Exception as e: # noqa: BLE001
return f"Error generating summary: {e!s}"
+ async def _acreate_summary(self, messages_to_summarize: list[AnyMessage]) -> str:
+ """Generate summary for the given messages."""
+ if not messages_to_summarize:
+ return "No previous conversation history."
+
+ trimmed_messages = self._trim_messages_for_summary(messages_to_summarize)
+ if not trimmed_messages:
+ return "Previous conversation was too long to summarize."
+
+ try:
+ response = await self.model.ainvoke(
+ self.summary_prompt.format(messages=trimmed_messages)
+ )
+ return response.text.strip()
+ except Exception as e: # noqa: BLE001
+ return f"Error generating summary: {e!s}"
+
def _trim_messages_for_summary(self, messages: list[AnyMessage]) -> list[AnyMessage]:
"""Trim messages to fit within summary generation limits."""
try:
diff --git a/libs/langchain_v1/tests/unit_tests/agents/test_middleware_agent.py b/libs/langchain_v1/tests/unit_tests/agents/test_middleware_agent.py
index 34b95e74ad..2908b865b0 100644
--- a/libs/langchain_v1/tests/unit_tests/agents/test_middleware_agent.py
+++ b/libs/langchain_v1/tests/unit_tests/agents/test_middleware_agent.py
@@ -1468,6 +1468,54 @@ def test_summarization_middleware_full_workflow() -> None:
assert "Generated summary" in summary_message.content
+async def test_summarization_middleware_full_workflow_async() -> None:
+ """Test SummarizationMiddleware complete summarization workflow."""
+
+ class MockModel(BaseChatModel):
+ def _generate(self, messages, **kwargs):
+ return ChatResult(generations=[ChatGeneration(message=AIMessage(content="Blep"))])
+
+ async def _agenerate(self, messages, **kwargs):
+ return ChatResult(generations=[ChatGeneration(message=AIMessage(content="Blip"))])
+
+ @property
+ def _llm_type(self):
+ return "mock"
+
+ middleware = SummarizationMiddleware(
+ model=MockModel(), trigger=("tokens", 1000), keep=("messages", 2)
+ )
+
+ # Mock high token count to trigger summarization
+ def mock_token_counter(messages):
+ return 1500 # Above threshold
+
+ middleware.token_counter = mock_token_counter
+
+ messages = [
+ HumanMessage(content="1"),
+ HumanMessage(content="2"),
+ HumanMessage(content="3"),
+ HumanMessage(content="4"),
+ HumanMessage(content="5"),
+ ]
+
+ state = {"messages": messages}
+ result = await middleware.abefore_model(state, None)
+
+ assert result is not None
+ assert "messages" in result
+ assert len(result["messages"]) > 0
+
+ expected_types = ["remove", "human", "human", "human"]
+ actual_types = [message.type for message in result["messages"]]
+ assert actual_types == expected_types
+ assert [message.content for message in result["messages"][2:]] == ["4", "5"]
+
+ summary_message = result["messages"][1]
+ assert "Blip" in summary_message.text
+
+
def test_summarization_middleware_keep_messages() -> None:
"""Test SummarizationMiddleware with keep parameter specifying messages."""