How do I send full chat history to AI Agent's memory?

Questions
1

Describe the problem/error/question

I have full chat history in bubble.io (between user and LLM). I want to send the entire chat history to the Agent’s Memory. I cant figure out how to do this.

What is the error message (if any)?

no error message

Please share your workflow

setup
setup1378×682 68.8 KB

(Select the nodes on your canvas and use the keyboard shortcuts CMD+C/CTRL+C and CMD+V/CTRL+V to copy and paste the workflow.)

Share the output returned by the last node


},
"params": 
{
},
"query": 
{
},
"body": 
{
"current_query": 
"User's latest message",
"chat_history": 
[
{
"role": 
"user",
"content": 
"Previous user message 1"
},
{
"role": 
"assistant",
"content": 
"Previous AI response 1"
},
{
"role": 
"user",
"content": 
"Previous user message 2"
}
],
"system_prompt": 
"Your custom system instructions here. Be concise and always end responses with a smiley face."
},

Information on your n8n setup

  • n8n version: Version 1.106.3
  • Database (default: SQLite): bubble.io
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app): cloud
  • Operating system: macOS
2

When you attach a Memory node to your agent, it automatically stores all chat messages.

Chat sessions are identified by a session_id (automatically generated by the Chat Trigger).

If you want persistent history for a specific user, for example, you can use the user_id from your platform instead. This way, every time that user interacts with the agent, it will recall their past conversations.

Considering you’re receiving the request in your webhook, you can send the user_id to n8n too and use that as the session id.

Does this clarifies your question?

3

Thanks @solomon. I understand how to start new chats and continue them later. However, I have old chat conversations before starting with n8n - I previously went directly to OpenRouter and saved all chat history in my bubble database.

I want n8n to get up to speed on a previous conversation, in which n8n never had access to.

Here’s an example conversation history I have with an LLM, which is stored in my bubble databe. It’s the first time n8n is seeing this conversation.

{
},
"query": 
{
},
"body": 
{
"current_query": 
"User's latest message",
"chat_history": 
[
{
"role": 
"user",
"content": 
"Previous user message 1"
},
{
"role": 
"assistant",
"content": 
"Previous AI response 1"
},
{
"role": 
"user",
"content": 
"Previous user message 2"
}
]

I’m able to successfully pass this info to n8n via webhook and use the chat history dynamically in later steps. I just can’t insert to memory. In short - I cant get my n8n agent up to speed with conversations that happened without n8n in the loop.

I thought the chatmemory manager node could help. And it feels like it will, but when I run it never takes my old messages - it never gets up to speed.

memory manager
memory manager1950×774 79.8 KB

settings for memory manager
settings for memory manager3706×1996 428 KB

simple memory
simple memory3736×1980 286 KB

In summary: My problem is that I have old chat history stored in his Bubble database from before I started using n8n, and I want n8n’s memory to recognize and continue those old conversations. While I can pass the old messages into n8n via webhook, the Memory Manager node won’t load them (neither will simple memory off Agent), so n8n never “gets up to speed” on past chats. This makes it impossible to migrate my API with old conversations to n8n

4

I see. n8n doesn’t offer a database for you to edit like Bubble does. You need an external database.

I usually go with Supabase because it’s very easy to use. You can replace your Simple Memory node for any database.

The Simple Memory node uses n8n’s internal database. But there’s no clear way of editing that. And if you’re deploying your solution to the public, you’ll need an external database anyways.

5

bubble is my external database, and I’m sending the chat history from the bubble databse. When I tap the memory [+] it does not give me an option for supabase or bubble.

you can see the only few options for memeory in this image.

databse
databse1256×797 114 KB

6

Supabase uses Postgres under the hood. So you can connect to Supabase using the Postgres option.

If bubble allows external connections to the database, it’ll probably be one of those options too.

7

ooooh!!! this is sad. No bubble absolutely does not allow this. Quick quesitons:

is the simple memory context unlimited? Both in terms of number of chat messages, and time.

So I can return to a previous conversation with 1,000 chats (based on unique sessionID) and in 5 years? and it will pick up right where i left off?

8

simple memory does not persist it will be cleared when n8n instance restarts.

for longtime memory storage you can use persistent storage solution like PostgreSQL it will store as many number of chats as you want and it will be there if you have access to database.

1 Like