Proxy Configuration Not Applied in N8N AI Agent Module

Questions
1

Hello,
We have deployed an on-premise N8N instance on an OpenShift cluster.
Being part of a large group, our cluster is network-isolated. Therefore, to allow an application to access external URLs, we must use a proxy. However, we have noticed that the IA Agent module does not seem to use the configured proxy, unlike the HTTP module, for example.

  • Is there a way to bypass this behavior?
  • Is this a known bug or a planned change in a future version?

Thank you in advance for your help.

Bonjour,
Nous avons déployé une instance N8N on-premise sur un cluster OpenShift. Étant dans un grand groupe, notre cluster est isolé sur le plan réseau. Ainsi, pour permettre à une application d’accéder à des URL externes, nous devons obligatoirement passer par un proxy.
Nous avons cependant constaté que le module IA Agent ne semble pas utiliser le proxy configuré, contrairement au module HTTP, par exemple.

  • Existe-t-il un moyen de contourner ce comportement ?
  • S’agit-il d’un bug connu ou d’une évolution prévue dans une prochaine version ?

Merci par avance pour votre aide.

2

Hey, @nicolas.matelot hope all is well.

I have my n8n behind personal proxy for inspecting http requests and I do see requests from ai agent hitting the proxy.

The way my n8n is configured for that is:

  • I have mitmproxy running in a separate container
  • mitmproxy generates a certificate, which needs to be added to the trust source of the n8n container
  • this certificate is mounted into n8n:
    • - ./mitmproxy:/opt/custom-certificates
  • when n8n detects there a new certificate, it imports it from docker-entrypoint.sh, which looks like this (it’s already there, no need to change anything):
#!/bin/sh
if [ -d /opt/custom-certificates ]; then
  echo "Trusting custom certificates from /opt/custom-certificates."
  export NODE_OPTIONS="--use-openssl-ca $NODE_OPTIONS"
  export SSL_CERT_DIR=/opt/custom-certificates
  c_rehash /opt/custom-certificates
fi

if [ "$#" -gt 0 ]; then
  # Got started with arguments
  exec n8n "$@"
else
  # Got started without arguments
  exec n8n
fi
  • I have these ENV VARs configured for n8n
    • - HTTP_PROXY=http://mitmproxy:8080
    • - HTTPS_PROXY=http://mitmproxy:8080

At this point when I use AI Agent with OpenAI model, I see the following in mitmproxy:

image
image718×333 15.2 KB

Hope this helps you.