Correction in API Base URL for chat/completion requests

[sekulicd]

The current base URL used for chat/completion requests is pointing to an incorrect host(62.4.58.131), causing a failure in request processing.

http://62.4.58.131/gpt4all-lora-q4/v1/chat/completions

@tiero @Janaka-Steph

[tiero]

This seems to be drives from your local IP of your home network?

Yes that should have been localhost and not using ifconfig

[tiero]

this seems connected to https://github.com/premAI-io/prem-app/issues/474

[Janaka-Steph]

Testing docker locally, you can see that baseUrl received from the daemon is wrong. This is my public IP.

[tiero]

I think that is kinda expected, so we need to think if we want to have a special rule for localhost eventually in the frontend or daemon. This is only for development after all

[tiero]

I have a different result instead @Janaka-Steph using https://demo.prem.ninja

The Service Details page shows correctly the baseUrl for LLama service

But then the playground chat calls the wrong baseUrl ie.

[Janaka-Steph]

Can you check TanStack devtool ?

[tiero]

I am on released on demo.prem.ninja, is there a way to do from there? or build the app with Node.js directly on the server?

[tiero]

I am making a new Docker release using latest main, let's see if it fixes

[Janaka-Steph]

We can add devtool on production, hidden and lazy loaded, triggered by a command in browser console

[Janaka-Steph]

The front just use service.baseUrl here

[tiero]

But baseUrl in the service detail page seems correct

[tiero]

Fixed using the latest main!

[Janaka-Steph]

Surprise!