Make Full Use of Postgres Chat Memory

Feature Requests
,
1

Describe the problem

When using the Postgres Chat Memory with AI Agent Nodes I’m not able to add any additional_kwargs, tool_calls, or response_metadata although these fields are added with every record to the database (empty though).
It would really help to have these information (especially timestamps) for further evaluation of chats. My current workaround leaves the Chat Memory out of the AI Agent Node and reads/writes from/to Chat Memory manually before and after the call, but that’s a lot of overhead to maintain if I want to do that with all sub-workflows using other agents. Just adding Postgres Chat Memory to the Agent Node would facilitate this process a lot.

Please share your workflow

Information on your n8n setup

  • n8n version: 1.62.5
  • Database (default: SQLite): PostgreSQL
  • n8n EXECUTIONS_PROCESS setting (default: own, main): regular
  • Running n8n via (Docker, npm, n8n cloud, desktop app): Docker (GCP)
  • Operating system:
2

Excellent suggestion!

That is also my need. I want to add custom data from a tool to a customer and additional_kwargs is awesome for this.

1 Like
3

About timestamps, a good workaround is to create a chat_history table manually (created_at column is automatically filled):

image

1 Like
4

Is there a way to add a timestamps in a way that the AI agent can read it? Because I believe that the AI Agent can only read the json right now?

1 Like
5

Do we have any updates on this feature?

6

Also interested on this, any update so far???

7

Ur update the last row in the table where type is ai(not human) , with the IntermediateSteps added in the content message with extra details like timestampor whatever you want.

I know that is not perfect, but is a quite workaround.