99% of the time it works, but then sometimes it just does not seem to save previous responses in the memory and therefore the chatbot repeats itself. I’m using a unique memory key e.g. 631899733 and I’m using context window 50, which is more than enough.
Any ideas? I can also confirm it is not due to me saving the workflow which i notice resets the memory or reloading the instance, I’m using the n8n hosted service.
is it possible n8n is updating their instances and it causes loss of memory?
Simple memory is meant for testing.
For real use I’d recommend something like Redis or Supabase (Postgres).
You’ll probably get more consistent results once you connect a service like that.
But to be honest I don’t really know why the simple memory would fail in your example.
Hello, I’m interested to know if the simple memory node is suitable for a production environment.
I saw @solomon’s answer where he doesn’t recommend it, but if it works fine in 99% of cases, I’m confident it’ll be fine for my use case.
You would be mixing your database with n8n’s database.
When you store the chat sessions, you’d probably want to be able to manage that in your database. Maybe even allow the user to see his past chat sessions, or erase them.
If you use the Simple Memory node, you’re using n8n’s database to store the chat sessions.
Most production projects would benefit from having more control over this data and put it in the same database as the rest of the user data.