Bug Description

When using a Langchain models like OpenAI and Gemini, HTTP calls going to the models online does not use the HTTP proxy defined for N8N. But go straight to the internet.

I found this as my N8N is not allowed to access internet directly. Allowing direct access resolves the issue

To Reproduce

1. Setup n8n to use a HTTP proxy using environmental variables HTTP_PROXY and HTTPS_PROXY.
2. Turn off access to internet except through proxy
3. Set up and run a simple AI flow using chat trigger, conversational agent and OpenAI model.

Expected behavior

Preferably, the AI model nodes should use the proxy defined for n8n. If not there should be a possibility to configure proxy separately for these nodes

Operating System

Docker

n8n Version

1.59.3

Node.js Version

Provided by docker image

Database

SQLite (default)

Execution mode

main (default)