FEEL FREE TO FORK THIS REPO IF YOU WANT

might add api key option to this

ive given up on this so im going to be archiving this, BUT IT STILL WORKS GREAT!! as of now, idk if things will change.

welcome to remote model access

these apps allow you to chat with a llm running on a separate machine which is either on the same internet on your iPhone, Apple watch(coming soon), or Apple TV (coming soon) or on the same tailscale network (Tailnet), from anywhere.

how to set up tailscale to let you use your server endpoint on any machine on your tailnet

Normal method (works only for devices on your Tailnet)

DOESNT WORK FOR THIS APP, USE THE FUNNEL METHOD

📡 Look at your server url that you're hosting from your device and copy the numbers after the ":" at the end, which is your port number. Example: "11435"

📡 Go to the Tailscale admin page and copy your devices' address that you're hosting from. Example "100.20.300.40"

📡 Add the port number to the end of the address, for example, "100.20.300.40:11435"

📡 Enter this url into the corresponding settings area in the app.

funnel method (works with any device, even if its not on your Tailnet) -

it seems like this is the only method that works here

assuming both the llm server is successfully hosting and the app is running on your device:

read this for more information: https://tailscale.com/kb/1223/funnel

📡 Go to the settings tab.

📡 Look at your server url that you're hosting from your device and copy the numbers after the ":" at the end, which is your port number.

📡 Open your terminal, and enter tailscale funnel [Port Number] or for it to always be on in the background, tailscale funnel --bg [Port Number] EXAMPLE: tailscale funnel 1234 or tailscale funnel --bg 1234 respectively, if your port was "1234".

📡 The terminal should display your tailscale funnel url for this now. Add "/v1/chat/completions" to that url and this will act as your tailscale server endpoint for the app.

📡 You can use the same api key in the app even if it's through tailscale, BUT YOU DONT NEED IT as you are self-hosting.

📡 Put the default identifier as the model request name, for example, "gemma-3-4b-it-qat"

📡 It should work now!

how to turn off tailscale funneling (If you want)

📡 Use tailscale funnel -off [Port Number], doesn't matter if it was running in background mode or not, both will turn off with this command.

privacy

No data is collected. Everything is on device except tailscale and your llm. :)

coming soon (eventually) (actually not bc im archiving this)

👀 Sideloading Repo (probably not)

👀 Support for multiple models at a time, but one per chat still (might be complicated as i only have one computer to host llms but we will see)

👀 Apple TV App (maybe this summer)

👀 Central website to also interact with your model. Similar to the apps. (maybe this summer)

👀 Image sending support for compatible models not available through the lmstudio api.