Remote kernel

[insungg]

Hi, and first of all, thank you so much for developing such a wonderful tool.

I wanted to ask whether WLJS supports connecting to a remote kernel. I know Mathematica can do this over SSH, but I’m not sure whether WLJS has similar functionality yet.

My laptop is not very powerful, so I usually run the Wolfram Engine on a remote Linux desktop. I was wondering whether WLJS could be used in a similar way to JupyterLab — for example, by hosting a WLJS server on the remote machine and then accessing it locally through SSH tunneling. Is this currently possible? I went through documentation and it's mentioned very shortly on here(https://github.com/WLJSTeam/wljs-notebook/blob/main/container/README.md) with nginx proxy, but it seems not very clear. Thanks!

[JerryI]

Hi @insungg
Thanks ^^

Remote Kernel is not yet a thing in WLJS app. But we plan to implement this feature if there is enough demand on it (+1 vote from you then 😉).

Usually most problems like this can be solved by moving an entire WLJS system to a remote PC and operate it from a browser (web-interface). WLJS can live perfectly fine in a browser. You will get mostly the same features, expect Rasterize, and exporting graphs to PDF will no longer work there since these features rely on our specific desktop applications renderer.

The easiest way to run WLJS on a remote server is using docker

docker run -it \
  -v ~/wljs:"/home/wljs/WLJS Notebooks" \
  -v ~/wljs/Licensing:/home/wljs/.WolframEngine/Licensing \
  -e PUID=$(id -u) \
  -e PGID=$(id -g) \
  -p 8000:3000 \
  --name wljs \
  ghcr.io/wljsteam/wljs-notebook:main

This will mount you local folder wljs to the container and expose WLJS web-interface at 8000 port in your remote os. Then it is up to you, how you route it.

It will already work, if you just connect to the network of your linux pc (via VPN) and then open a browser at //ipaddress:8000. But then your browser will complain about non-secure connection. For this reason you can spin up a reverse proxy - NGINX server, that will reroute all traffic via HTTPS (see https://github.com/WLJSTeam/wljs-notebook/blob/main/container/README.md#tls-proxy-config) and even add some basic authentication if needed.

Another option: you do all the same with a docker, but forward all HTTP traffic over SSH connection (no NGINX needed). I am not an expert, but VSCode does somewhat similar called SSH port forwarding.

We can try to help you depending on the scenario or your needs.

[insungg]

thanks so much for the detailed walkthrough, I'll give it a try!

[JerryI]

added as a feature request:

#491

[JerryI]

Actually there is a 3rd way of running WLJS if you have Wolfram Engine installed on a remote machine. Just in case if you hate docker containers ;)

Step 1: clone

git clone https://github.com/WLJSTeam/wljs-notebook.git
cd wljs-notebook

Step 2: start it

wolframscript -f Scripts/start.wls

WLJS Notebook uses 1 port for plain HTTP and 2 WebSockets ports for communication: 20560, 20559, and 20563. Make sure to have 1:1 port-forwarding for the last 2 if you plan to connect from a proxy or SSH tunnel.

[insungg]

awesome!! this works for me via ssh tunneling for those three ports. Thank you so much!

[JerryI]

We will make it as an official guide :)ThanksOn 22. Mar 2026, at 06:52, Insung Hwang ***@***.***> wrote: Anyone who wonders remote accessing to wljs on host side Add these command to the host side zshrc (tmux and Wolfran engine should be installed) wljsremote() { local session="${1:-wljs}" local repo="<path to wljs-notebook>" local startwls="$repo/Scripts/start.wls" if ! command -v tmux >/dev/null 2>&1; then echo "Error: tmux not found" >&2 return 1 fi if ! command -v wolframscript >/dev/null 2>&1; then echo "Error: wolframscript not found in PATH" >&2 return 1 fi if [ ! -f "$startwls" ]; then echo "Error: $startwls not found" >&2 return 1 fi if tmux has-session -t "$session" 2>/dev/null; then echo "Error: tmux session '$session' already exists" >&2 return 1 fi if ! tmux new -d -s "$session" \ "bash -lc 'cd \"$repo\" && exec wolframscript -f \"$startwls\"'" then echo "Error: failed to create tmux session '$session'" >&2 return 1 fi echo "Started $session" } killwljsremote() { local session="${1:-wljs}" tmux kill-session -t "$session" } You can lanuch wljs by wljsremote and close it by killwljsremote On client side, add ssh tunnel command to zshrc: listenwljs() { ssh -f -N \ -L 20560:localhost:20560 \ -L 20559:localhost:20559 \ -L 20563:localhost:20563 \ <host name>@<host address> -p <host port, defualt=22> } 152 153 killwljs() { 154 pkill -f "20560:localhost:20560.*20559:localhost:20559.*20563:localhos t:20563" 155 } listenwljs on client side opens ssh tunnel for wljs. Go to localhost:20560 on client side to use wljs!! —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>

[insungg]

Just one comment about this code; it seems that when launched with tmux, the Wolfram kernel sometimes doesn’t respond, so I have to refresh it. This doesn’t happen when it’s launched directly from the shell. Other than that, everything works perfectly.

[JerryI]

It might be related to wolframscript. It does not like, when there are too many sessions. WLJS uses 2 (master kernel and evaluation kernel), which is a legal maximum. May be some processes are not terminated properly and are still running from the previous session. In this case wolframscript will kill them automatically and commit suicide too ;D

[JerryI]

We updated wljs to 3.0.2. Please, do not forget to periodically fetch the changes from the repo ;)