Chat Memory Manager Node Returns Full History Despite Context Window Length Setting

Based on the documentation, the Context Window Length parameter in the Postgres Chat Memory node is used to define how many previous interactions are considered when building the conversation context. However, when the Chat Memory Manager node retrieves messages using that memory, it currently returns all messages stored in the database for that session ID. In other words, the Chat Memory Manager node doesn’t automatically apply the Context Window Length limit—you’re expected to filter or process the messages afterward if you need a limited set for downstream operations.

To help troubleshoot further, please provide the following details:

What is the error message (if any)?

Let us know if any error messages are shown in your workflow or in the logs.

Please share your workflow

Including a summary or diagram of your workflow setup would be useful. In particular, how you’ve connected the Postgres Chat Memory node to the Chat Memory Manager node.

Share the output returned by the last node

Knowing what data is being returned can help clarify whether the issue is with the memory node’s configuration or the way the Chat Memory Manager node processes that data.

Information on your n8n setup

• n8n version: 1.75.2
• Database (default: SQLite):
• n8n EXECUTIONS_PROCESS setting (default: own, main):
• Running n8n via (Docker, npm, n8n cloud, desktop app):
• Operating system:

If all these details are as expected and there’s no error message, then the behavior you described is consistent with the current implementation. The Context Window Length only configures the memory node’s internal behavior for subsequent context-based operations and doesn’t limit the full history retrieval by the Chat Memory Manager node.

Let me know if you can share the additional details or if you need further clarification!