Memory — short-term and long-term

In one line

Checkpointer = memory for this chat. Store = a notebook kept across chats.

Key idea

• Short-term (checkpointer): saves message history per thread_id. Lets a job pause and resume.
• Long-term (store): saves facts under a namespace like ("memories", user_id). Add an embeddings index to search by meaning.

from langgraph.checkpoint.memory import InMemorySaver
from langgraph.store.memory import InMemoryStore
graph = builder.compile(checkpointer=InMemorySaver(), store=InMemoryStore())

Production options: PostgresSaver, RedisSaver, MongoDBSaver.

Links

• Persistence
• Add memory
• Full guide — appendix C (memory)

My notes