Azure OpenAI Chat Model node times out behind proxy

[TechNovaT]

Description

I’m running n8n in a Docker container within a closed network where internet access is only possible through a proxy (no authentication required). I have set the HTTP_PROXY and HTTPS_PROXY environment variables accordingly.

The HTTP Request node works perfectly fine with these proxy settings; however, when I attempt to use the Azure OpenAI Chat Model node, I encounter a “Request timed out” error.

Information on my n8n setup

• n8n version: 1.83.2
• Database (default: SQLite): SQLite
• n8n EXECUTIONS_PROCESS setting (default: own, main): Main
• Running n8n via (Docker, npm, n8n cloud, desktop app): Docker
• Operating system: Oracle Linux 8

My Testing Workflow with Azure OpenAI

Error Logs

Error in handler N8nLlmTracing, handleLLMStart: TypeError: fetch failed
Request timed out.

Expected Behavior

• n8n connects to the external Azure OpenAI API through the proxy.

Actual Behavior

• n8n cannot connect to the external Azure OpenAI API using Azure OpenAI Chat Model
• n8n can connect to Azure OpenAI using HTTP Request node without issues.

Thank You!

Any assistance would be greatly appreciated!
If there’s any additional information you need, let me know!

[Dandy_Dow]

Hey, what’s up?

If the HTTP Request node works fine with your proxy but the Azure OpenAI Chat Model doesn’t, chances are the model node isn’t respecting the HTTP_PROXY and HTTPS_PROXY environment variables.

Here are a few things you can try:

In your Docker container, try adding this to your environment as well:

NODE_OPTIONS=“–require https-proxy-agent”

Another option is using the global-agent package in your container’s entrypoint to force proxy usage globally for all Node.js requests.

If possible, run a small raw Node.js script inside the container using fetch to test if the proxy works outside of n8n nodes. That’ll help narrow things down.

Also, if you’re running n8n in queue mode with separate workers, make sure the proxy env vars are set in all containers (main and workers). Sometimes only the main container has them and the workers fail.

Dandy