I am using postgresql memory for my agent and after I update to the latest selfhosted version 1.100.1
i am experiencing a lack of speed, for instance just a simple hi takes 8s to process.
I need help.
thanks
I am using postgresql memory for my agent and after I update to the latest selfhosted version 1.100.1
i am experiencing a lack of speed, for instance just a simple hi takes 8s to process.
I need help.
thanks
I’ve been experiencing the same thing. I’m pretty sure it’s a bug. The latest version without this issue is 1.91.3, so it’s been around for quite a while. For now, the only workaround is to downgrade. From time to time, I upgrade my instance to check if they’ve fixed it, but I just tested version 1.103.0 and the bug is still there. It takes around 4 to 5 seconds for an empty flow with just an agent (no system prompt) replying to a simple “Hi” using Gemini Flash 2.0. In version 1.91.3, it takes less than a second.
The same issue happens with the text classifier, so it might be related to all models or possibly all AI-related nodes. I really hope they fix this soon, because it’s impossible to work with complex AI agents if each response takes more than 15 seconds.
Hey Sabatage3
I tried using just for testing the simple memory tool instead of Postgresql and I could down the time to 3-4 seconds.
As I said before, I am just testing for now. Also I am on 1.101.2 version.
I just upgraded my instance to version 1.104.0, and the bug is still present; it’s still taking more than 4 seconds just for a simple “hello” response.
Hi @Gonzalo_Alegre I just upgraded to version 1.108.0 and the issue persists. However, I tried using this env variable N8N_RUNNERS_ENABLED=false, and it seems to perform faster. It looks like the runners feature might be a bottleneck, or maybe it requires additional settings. I think it’s enabled by default. Could you try that env variable and see if it improves performance?
Hey @Sabatage3, how are you doing? Okey I will try it and I’ll let you know if it works.
I had better performance times just by improving my cloud cpu, memory, etc.
I know it is not really a solution but I handle in that way. I will try what you say
After several days of testing, I think N8N_RUNNERS_ENABLED=false did the trick. I got a warning that says it will be true by default in the future, and you won’t be able to turn it off. I hope they don’t do that until they fix the slow responses in agents when runners are enabled.
This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.