Memory

Memory in an AI application is information the application saves and later supplies to a model or workflow. It is not one automatic capability inside every model call.

Tiny example

Conversation history keeps recent messages for one thread. A long-term store keeps an approved language preference across threads. A checkpoint saves workflow state so an interrupted job can resume.

FDE note

Name the lifecycle precisely: context, history, state, checkpoint, summary, or long-term store. Each needs its own scope, retention, access, correction, and deletion policy. Read message history is not durable memory.