[Question]: How to set HTTP proxy for calling AI providers?

[nth347]

What features would you like to see added?

Does LibreChat support setting HTTP Proxy for outgoing API calls to AI providers?

More details

I'm attempting to configure LibreChat to route all API calls to AI models through a proxy server running on my host machine (macOS). I've set the PROXY environment variable in my .env file as follows:

PROXY=http://host.docker.internal:8080

However, after restarting the Docker containers using docker compose up -d --force-recreate , traffic does not appear to be routed through the specified proxy.

Please let me know if I am missing something.

Which components are impacted by your request?

General

Pictures

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

Can you share which AI providers? most are configured to use an HTTP proxy if one is set

If possible, please also share your librechat.yaml file

[nth347]

Hello @danny-avila, I’m new to LibreChat. I haven’t made any changes to the .yaml file, so it’s still the default one. I would like to set an HTTP proxy for the following providers: OpenAI, Google, AWS Bedrock, and Anthropic. Could you please let me know if there’s a global configuration option for this?

[danny-avila]

The Google client doesn't support proxying though. Last I checked their SDK didn't support it but this might've changed by now. Every other provider you mentioned should support it

[nth347]

Do you mean that the PROXY option might work for other providers but not for Google by now?

[danny-avila]

yes

[nth347]

Hi @danny-avila,

I tested the <PROVIDER>_REVERSE_PROXY option for Google, OpenAI, and Anthropic, and it worked perfectly. Thank you for your kind support!

I was wondering if it would be possible to:

• Retrieve the current username from the context of LibreChat.
• Set the current username as an HTTP request header when making requests to the AI provider.

My goal is to monitor AI usages of AD users of the enterprise.