DEV Community

Mike W
Mike W

Posted on

cathedral-memory 0.2.0: LangChain memory adapter

#ai #agents #langchain #python

cathedral-memory 0.2.0 adds native LangChain support.

pip install cathedral-memory[langchain]
Enter fullscreen mode Exit fullscreen mode

CathedralMemory

Drop-in BaseMemory for any LangChain chain or agent.

from cathedral.langchain import CathedralMemory
from langchain.chains import ConversationChain

memory = CathedralMemory(api_key='cathedral_...')
chain = ConversationChain(llm=llm, memory=memory)
Enter fullscreen mode Exit fullscreen mode

On load_memory_variables() it calls wake() for full identity reconstruction. On save_context() it stores the exchange via remember(). That chain now has persistent memory across sessions.

CathedralChatMessageHistory

Drop-in BaseChatMessageHistory for LangChain v0.1+.

from cathedral.langchain import CathedralChatMessageHistory
from langchain.memory import ConversationBufferMemory

history = CathedralChatMessageHistory(api_key='cathedral_...')
memory = ConversationBufferMemory(chat_memory=history, return_messages=True)
Enter fullscreen mode Exit fullscreen mode

What you get for free

  • Drift detection -- check if the agent's self-model has shifted
  • Snapshots -- freeze state at any point
  • Behaviour log -- track what the agent actually does vs what it says
  • Compaction -- trim low-value memories when the store gets bloated

Free tier: 1,000 memories per agent. No credit card. cathedral-ai.com/playground to try the API.

Top comments (0)