Window Buffer Memory (replaced)

Deprecated: replaced by Chat Memory

Window Buffer Memory is no longer in the node palette. Use Chat Memory with Storage: This run only instead. Existing workflows keep running.

Gave an agent a short memory of the last few messages.

What it does

Workflows that already use Window Buffer Memory keep working. The node now runs as a Chat Memory with Storage: This run only and History sent to the model: Last N messages, where N is twice the old window size (a question and its answer are two messages). With the default size of 5, that is the last 10 messages.

The conversation is now really kept for the length of the run (the old node didn’t remember anything). Nothing is saved after the run. When you open the workflow, the node shows that it was upgraded; saving the workflow turns it into a Chat Memory node.

When to use it

Don’t add it to new workflows. Use Chat Memory.

Inputs and settings

设置项	类型	作用
Window Size (k)	Number	Only the last k message exchanges are kept in memory. 默认值：5

Window Size (k) becomes Chat Memory’s Messages, doubled.

Outputs

This node has no output port. It plugs into a slot on another node instead of passing items along.

• Memory slot of Chat Trigger, Tools Agent, RAG Agent

Dependencies and credentials

None. This node needs no connection and has no slots for other nodes.

Example workflow

1. Open a workflow that uses Window Buffer Memory with Window Size (k) 5.
2. Save it. The node becomes a Chat Memory with Storage: This run only, Last N messages and Messages 10.

Common issues

• The agent forgets between runs. That is expected: nothing is saved after the run. Switch the Chat Memory to Conversation or Key storage to remember across runs.
• The Chat Trigger shows no history. A run-only memory can’t give a Chat Trigger a chat history. Use Conversation or Key storage.

Related nodes

• Chat Memory: the node that replaces it.
• Tools Agent: an agent that uses a memory.