← Workflow examples

Chat with conversation memory

Remembers what was said earlier, so the customer does not have to repeat themselves every turn.

Graph preview

How it works

A window memory node (window_memory) is attached to the agent over a link_memory wire. It keeps the most recent messages of the session and injects them into every request, so the user can say “tell me more” without repeating the context.

  1. Entry receives the message.
  2. Memory hands the agent the session history.
  3. Agent answers with that history in context.
  4. Answer returns the result.

What to configure after import

  • The window size: too large costs tokens on every turn, too small cuts the context off.
  • For long conversations look at summary_memory, which compresses the older part of the dialogue instead of dropping it.

Memory is per session: users of a public chat never see each other’s history.