Skip to content

The agent engineering platform.

langchain[patch]: add async methods to ConversationSummaryBufferMemory (#20956)

Thomas Meike40c02cedaffe
diff --git a/libs/langchain/langchain/memory/summary.py b/libs/langchain/langchain/memory/summary.py
index b7cb996cb5..1ae2418401 100644
--- a/libs/langchain/langchain/memory/summary.py
+++ b/libs/langchain/langchain/memory/summary.py
@@ -34,6 +34,18 @@ class SummarizerMixin(BaseModel):
chain = LLMChain(llm=self.llm, prompt=self.prompt)
return chain.predict(summary=existing_summary, new_lines=new_lines)
+ async def apredict_new_summary(
+ self, messages: List[BaseMessage], existing_summary: str
+ ) -> str:
+ new_lines = get_buffer_string(
+ messages,
+ human_prefix=self.human_prefix,
+ ai_prefix=self.ai_prefix,
+ )
+
+ chain = LLMChain(llm=self.llm, prompt=self.prompt)
+ return await chain.apredict(summary=existing_summary, new_lines=new_lines)
+
class ConversationSummaryMemory(BaseChatMemory, SummarizerMixin):
"""Conversation summarizer to chat memory."""
diff --git a/libs/langchain/langchain/memory/summary_buffer.py b/libs/langchain/langchain/memory/summary_buffer.py
index 389da7e421..f17f011ab9 100644
--- a/libs/langchain/langchain/memory/summary_buffer.py
+++ b/libs/langchain/langchain/memory/summary_buffer.py
@@ -19,6 +19,11 @@ class ConversationSummaryBufferMemory(BaseChatMemory, SummarizerMixin):
"""String buffer of memory."""
return self.load_memory_variables({})[self.memory_key]
+ async def abuffer(self) -> Union[str, List[BaseMessage]]:
+ """Async memory buffer."""
+ memory_variables = await self.aload_memory_variables({})
+ return memory_variables[self.memory_key]
+
@property
def memory_variables(self) -> List[str]:
"""Will always return list of memory variables.
@@ -43,6 +48,22 @@ class ConversationSummaryBufferMemory(BaseChatMemory, SummarizerMixin):
)
return {self.memory_key: final_buffer}
+ async def aload_memory_variables(self, inputs: Dict[str, Any]) -> Dict[str, Any]:
+ """Asynchronously return key-value pairs given the text input to the chain."""
+ buffer = await self.chat_memory.aget_messages()
+ if self.moving_summary_buffer != "":
+ first_messages: List[BaseMessage] = [
+ self.summary_message_cls(content=self.moving_summary_buffer)
+ ]
+ buffer = first_messages + buffer
+ if self.return_messages:
+ final_buffer: Any = buffer
+ else:
+ final_buffer = get_buffer_string(
+ buffer, human_prefix=self.human_prefix, ai_prefix=self.ai_prefix
+ )
+ return {self.memory_key: final_buffer}
+
@root_validator()
def validate_prompt_input_variables(cls, values: Dict) -> Dict:
"""Validate that prompt input variables are consistent."""
@@ -60,6 +81,13 @@ class ConversationSummaryBufferMemory(BaseChatMemory, SummarizerMixin):
super().save_context(inputs, outputs)
self.prune()
+ async def asave_context(
+ self, inputs: Dict[str, Any], outputs: Dict[str, str]
+ ) -> None:
+ """Asynchronously save context from this conversation to buffer."""
+ await super().asave_context(inputs, outputs)
+ await self.aprune()
+
def prune(self) -> None:
"""Prune buffer if it exceeds max token limit"""
buffer = self.chat_memory.messages
@@ -73,7 +101,25 @@ class ConversationSummaryBufferMemory(BaseChatMemory, SummarizerMixin):
pruned_memory, self.moving_summary_buffer
)
+ async def aprune(self) -> None:
+ """Asynchronously prune buffer if it exceeds max token limit"""
+ buffer = self.chat_memory.messages
+ curr_buffer_length = self.llm.get_num_tokens_from_messages(buffer)
+ if curr_buffer_length > self.max_token_limit:
+ pruned_memory = []
+ while curr_buffer_length > self.max_token_limit:
+ pruned_memory.append(buffer.pop(0))
+ curr_buffer_length = self.llm.get_num_tokens_from_messages(buffer)
+ self.moving_summary_buffer = await self.apredict_new_summary(
+ pruned_memory, self.moving_summary_buffer
+ )
+
def clear(self) -> None:
"""Clear memory contents."""
super().clear()
self.moving_summary_buffer = ""
+
+ async def aclear(self) -> None:
+ """Asynchronously clear memory contents."""
+ await super().aclear()
+ self.moving_summary_buffer = ""
diff --git a/libs/langchain/tests/integration_tests/chains/test_memory.py b/libs/langchain/tests/integration_tests/chains/test_memory.py
deleted file mode 100644
index 826380b856..0000000000
--- a/libs/langchain/tests/integration_tests/chains/test_memory.py
+++ /dev/null
@@ -1,32 +0,0 @@
-"""Test memory functionality."""
-
-from langchain.memory.summary_buffer import ConversationSummaryBufferMemory
-from tests.unit_tests.llms.fake_llm import FakeLLM
-
-
-def test_summary_buffer_memory_no_buffer_yet() -> None:
- """Test ConversationSummaryBufferMemory when no inputs put in buffer yet."""
- memory = ConversationSummaryBufferMemory(llm=FakeLLM(), memory_key="baz")
- output = memory.load_memory_variables({})
- assert output == {"baz": ""}
-
-
-def test_summary_buffer_memory_buffer_only() -> None:
- """Test ConversationSummaryBufferMemory when only buffer."""
- memory = ConversationSummaryBufferMemory(llm=FakeLLM(), memory_key="baz")
- memory.save_context({"input": "bar"}, {"output": "foo"})
- assert memory.buffer == ["Human: bar\nAI: foo"]
- output = memory.load_memory_variables({})
- assert output == {"baz": "Human: bar\nAI: foo"}
-
-
-def test_summary_buffer_memory_summary() -> None:
- """Test ConversationSummaryBufferMemory when only buffer."""
- memory = ConversationSummaryBufferMemory(
- llm=FakeLLM(), memory_key="baz", max_token_limit=13
- )
- memory.save_context({"input": "bar"}, {"output": "foo"})
- memory.save_context({"input": "bar1"}, {"output": "foo1"})
- assert memory.buffer == ["Human: bar1\nAI: foo1"]
- output = memory.load_memory_variables({})
- assert output == {"baz": "foo\nHuman: bar1\nAI: foo1"}
diff --git a/libs/langchain/tests/unit_tests/chains/test_summary_buffer_memory.py b/libs/langchain/tests/unit_tests/chains/test_summary_buffer_memory.py
new file mode 100644
index 0000000000..f6651c50c2
--- /dev/null
+++ b/libs/langchain/tests/unit_tests/chains/test_summary_buffer_memory.py
@@ -0,0 +1,62 @@
+"""Test memory functionality."""
+
+from langchain.memory.summary_buffer import ConversationSummaryBufferMemory
+from tests.unit_tests.llms.fake_llm import FakeLLM
+
+
+def test_summary_buffer_memory_no_buffer_yet() -> None:
+ """Test ConversationSummaryBufferMemory when no inputs put in buffer yet."""
+ memory = ConversationSummaryBufferMemory(llm=FakeLLM(), memory_key="baz")
+ output = memory.load_memory_variables({})
+ assert output == {"baz": ""}
+
+
+async def test_summary_buffer_memory_no_buffer_yet_async() -> None:
+ """Test ConversationSummaryBufferMemory when no inputs put in buffer yet."""
+ memory = ConversationSummaryBufferMemory(llm=FakeLLM(), memory_key="baz")
+ output = await memory.aload_memory_variables({})
+ assert output == {"baz": ""}
+
+
+def test_summary_buffer_memory_buffer_only() -> None:
+ """Test ConversationSummaryBufferMemory when only buffer."""
+ memory = ConversationSummaryBufferMemory(llm=FakeLLM(), memory_key="baz")
+ memory.save_context({"input": "bar"}, {"output": "foo"})
+ assert memory.buffer == "Human: bar\nAI: foo"
+ output = memory.load_memory_variables({})
+ assert output == {"baz": "Human: bar\nAI: foo"}
+
+
+async def test_summary_buffer_memory_buffer_only_async() -> None:
+ """Test ConversationSummaryBufferMemory when only buffer."""
+ memory = ConversationSummaryBufferMemory(llm=FakeLLM(), memory_key="baz")
+ await memory.asave_context({"input": "bar"}, {"output": "foo"})
+ assert memory.buffer == "Human: bar\nAI: foo"
+ output = await memory.aload_memory_variables({})
+ assert output == {"baz": "Human: bar\nAI: foo"}
+
+
+def test_summary_buffer_memory_summary() -> None:
+ """Test ConversationSummaryBufferMemory when only buffer."""
+ llm = FakeLLM(queries={0: "summary"}, sequential_responses=True)
+ memory = ConversationSummaryBufferMemory(
+ llm=llm, memory_key="baz", max_token_limit=5
+ )
+ memory.save_context({"input": "bar"}, {"output": "foo"})
+ memory.save_context({"input": "bar1"}, {"output": "foo1"})
+ assert memory.buffer == "System: summary\nHuman: bar1\nAI: foo1"
+ output = memory.load_memory_variables({})
+ assert output == {"baz": "System: summary\nHuman: bar1\nAI: foo1"}
+
+
+async def test_summary_buffer_memory_summary_async() -> None:
+ """Test ConversationSummaryBufferMemory when only buffer."""
+ llm = FakeLLM(queries={0: "summary"}, sequential_responses=True)
+ memory = ConversationSummaryBufferMemory(
+ llm=llm, memory_key="baz", max_token_limit=5
+ )
+ await memory.asave_context({"input": "bar"}, {"output": "foo"})
+ await memory.asave_context({"input": "bar1"}, {"output": "foo1"})
+ assert memory.buffer == "System: summary\nHuman: bar1\nAI: foo1"
+ output = await memory.aload_memory_variables({})
+ assert output == {"baz": "System: summary\nHuman: bar1\nAI: foo1"}