Concepts

Memory

Stores and manages conversation history across multiple turns, so a chain or agent can reference what was said earlier.

Why it exists

LLM calls are stateless by default — Memory keeps track of prior messages so multi-turn conversations stay coherent.

View source on GitHub
educhain.py
from educhain.memory import ChatMessageHistory

history = ChatMessageHistory()
history.add_user_message("What's the capital of France?")
history.add_ai_message("Paris.")

Related