Feedback for “Local API Server”

[therealmonadnomad]

What i am trying to do:
use LibreChat in a container on my laptop to access Jan.AI on my GPU-enabled server.

What error i get when attempting to use LibreChat:
Something went wrong. Here's the specific error message we encountered: An error occurred while processing the request: 403 Invalid host header

How I can duplicate the error with curl:
curl -X 'GET' 'http://aibox:1337/v1/models' -H 'accept: application/json' -H 'Authorization: Bearer [secret]'
returns: "Invalid host header"
(I checked name resolution & that the port is open)

How I can work-around with curl:
curl -X 'GET' 'http://aibox:1337/v1/models' -H 'accept: application/json' -H 'Authorization: Bearer [secret]' --header 'Host: localhost'
returns: the name of the currently loaded model as JSON

Further comment on the curl testing:
I get the same results using curl on the aibox. If i use hostname, or IP address, i need to add the host header localhost to get results. This further cements my thoughts on where it issue lies...

My theory:
Jan.AI API server seems to want the incoming http request to have a host header of localhost. Does that even make sense? If the answer is, "Yes it does require that, there is a good reason, " can someone please explain that good reason?

Here's my LibreChat config to point at my Jan.AI server:

custom:
    - name: "SilverPCJanAI"
      apiKey: "[secret]"
      baseURL: "http://aibox:1337/v1/chat/completions"
      models:
        default: ["Gemma-The-Writer-N-Restless-Quill-10B-D_AU-IQ4_XS"]
      titleConvo: true
      titleModel: "Gemma-The-Writer-N-Restless-Quill-10B-D_AU-IQ4_XS"
      modelDisplayLabel: "Custom AI"

Again, I've verified name resolution and connectivity between the two computers. I get the same results if i test using IP address instead of a hostname.

A few relevant Jan.AI Local API Server settings:
Server host: 0.0.0.0
Server Port: 1337
Trusted Hosts: 10...*

[therealmonadnomad]

Update:
I found that i can work around by using nginx as a proxy to set the host header and then i can access Jan.AI from the other computer. Is that the "correct" solution? Or "should" the API connection not require that header?

[azakharchenko-msol]

same problem