Yes services on Railway have full access to the internet, we do not block any outgoing traffic nor do we even have the ability to.

I am able to resolve that domain just fine - https://utilities.up.railway.app/dns-lookup?value=api.openai.com&type=ip

This means it's an issue with your application or config, to be able to help here we would need a minimal reproducible example.

What do you mean by minimal reproducible example?

Also, my frontend deployed application does not accurately send a post req to my backend, but when i run it locally with the same urls it sends it. This doesn't make sense to me.

https://stackoverflow.com/help/minimal-reproducible-example

I'll do that. But now that I htink about it, I don't know if my environment variables are just setup properly. I dont know if there could be any other issue. I kept a .env file in my directory with my open ai key and then deployed that to production. Does that work with Railway?

While Railway doesn't do anything to prevent you from using an .env file, please never deploy with an .env file, you need to be using service variables.

https://docs.railway.app/overview/the-basics#service-variables

Okay, gotcha.

As for the example, do you have the ability to ssh into these instances and run a python script? If so, I'll do that. Additionally, to replicate this error it seems like you'd need an OPEN_AI_KEY to even test it. Is there anyway I can send that to you without it being public

A simple open ai chat request works on my own computer but when i try to do it on my deployed app it gives me the initial error.

I do technically have the ability to do SSH into a container, yes, but we do not provide a way for our users do that, nor is that a level of support I can provide.

I can acquire my own API key, no need to share yours.

Sounds good. However, I'm not sure how to test this issue without the ability for you to SSH into the container. Do you have access to my project?

I do, but I don't see why I need that, you need to provide me a minimal reproducible example that I could deploy myself on my own account.

Okay. Here's a file that you can use in deployment. I additionally attached requirements.txt and a dockerfile.

Unless I'm blind, I don't see how you are passing in auth?

By auth do you mean the api key? At least for the new installation of pip openai, it says to set the environment variable "OPENAI_API_KEY". I just set this as a service variable in my deployment.

I don't like the ambiguity of that.

What do you mean?

Yea, I passed it in now and still get the same issue.

I tested if there was a connection to the internet. Doesn't look like it!

So, it wouldn't be able to install the pip packages in my Dockerfile unless there wasn't internet. That must mean there is some sort of blocking I believe

Can I have some help here?

As rc stated in another thread, the issue connecting to google is an application / code level issue, we don't provide coding assistance so we can't help with that.

The OpenAI issue is likely them blocking GCP, this is also out of our hands, you would need to get in contact with OpenAI.

I'm sorry we can't help further.