Create example disambiguating langgraph checkpointer and store from short- and long- term memory

XMLWordPrintableJSON

    • Type: Task
    • Resolution: Unresolved
    • Priority: Minor - P4
    • None
    • Affects Version/s: None
    • Component/s: AI/ML, LangGraph
    • None
    • Python Drivers
    • None
    • None
    • None
    • None
    • None
    • None

      Context

      LangGraph – Is a graph abstraction with nodes and agents that speak to each other in a system. This can be agent -> agent or agent -> human -> agent -> agent -> ... 

      LangGraph Checkpointer – Makes a copy of the state of the "graph" (see above) and reserves this state information. This is what is often confused for short-term memory.

      LangGraph Store – This abstraction that's available to keep data. Database storage as a key:value store. We have code for this defined as MongoDBStore.

      Short Term Memory – This is just a generalized term. There is no short term memory explicitly. They're referring to the checkpointer.

      Long Term Memory – This is at length chat history. 

      Since we've been talking about the LangGraph Store (MongoDBStore), everyone has been calling it long-term memory. I was guilty of this too. As that started, we began to think of the checkpointer as short-term memory. The truth is that this is an oversimplification, and I feel like something needs to be created that show off what more these core langgraph features are.

      Definition of done

      • Generate a blog post or article disambiguating this content.
      • Loop in peter.richards@mongodb.com to see where the blog post can be made. 

      Pitfalls

      • There may be existing disambiguation here on what this work is.
        • I don't want to undermine anyone's great work, nor create confusion in attempting to show what the checkpointer and store beyond memory.

            Assignee:
            Unassigned
            Reporter:
            Casey Clements
            Votes:
            0 Vote for this issue
            Watchers:
            2 Start watching this issue

              Created:
              Updated: