Well, not 100% useless: I can see its use for applications running inside Docker containers. Yes, there are ways to have GUI applications rendered from the inside of a container, but maybe this is easier than getting the Dockerfile right.
EDIT: nevermind, doing this with Docker seems much easier than I expected [0]. I'll try it tomorrow, I'm curious to see if the proposed solution works on Windows as well.
Yeah, I can’t explain why this project makes me so happy because I struggle to think of any time where I’d need this, but it puts a big, dumb grin on my face.
It reminds me a bit of chindōgu, the Japanese art (?) of useless inventions. There's a particular delight to ingenious, but absurd or useless creations.
Emacs it's full of chindogus. Also, there's geekcode, xroach, megahal/hailo, xneko, aatv and mplayer rendering videos over aalib, aaquake, eforth running in the subleq virtual machine...
With a small framebuffer font aatv was almost watchable over a distance, but OFC fbtv made it obsolete, and ditto with mplayer -vo aalib as movies worked in the framebuffer just as fine as X.
But I remember the BB demo and I still remember these catchy s3m modules...
And, well, not AA, but I still play today tons of text adventures and roguelikes (and BSDgames and such), and my main X environment it's CLI/TUI based except for CWM (Window manager), MPV/MuPDF/NSxiv (images) and djview4 for DJVU files..
Well, you can run apps on any less capable device with ssh and proper terminal display. You can limit data usage by offloading video buffering to the host (however not sure if that's net positive saving). And put the host behind VPN to avoid getting region blocks.
I actually used to tunnel Netscape Navigator via SSH to my Commodore Amiga desktop via an Xorg server way back in the 56K phone modem Internet days from my ISP's SSH user account login, since Amiga didn't have Netscape (and even if it did, the Amiga likely would have choked on it, massive and bloated as Netscape was), and the browser AmigaOS did have just wasn't up to the task of normal day-to-day usage of the Web as it existed back then. Fun times.
Sure am glad of the broadband Internet and modern "powerhouse" PCs we have so readily available today. Hell, even the computer most everyone carries in their pocket these days is infinitely more powerful than the average desktop machines of my childhood. :)
That’s right, it will still be pixelfied text. We could run ocr on the images then convert to text, but that’s an entirely new can of worms. Make a feature request issue on GitHub if you’re interested,
I run a ttyd server to get terminal over https, and I have used carbonyl over that to get work done. That's limited to a web browser (to get access to resources not exposed via the public internet), so having full GUI support is very useful
This is one of those things that pushes the boundaries to nowhere, yet everywhere at the same time whilst being incredibly awesome and something you can show off ad infinitum. Outstanding! Not sure how we’ll implement vdi now! Gives ghost in the shell a whole new meaning.
I had the change a couple of line to make it work because term.everything takes input only from stdin (this way it works of ssh and is pretty broadly compatible across terminals).
1. I had to remap another key to the control key (which is usually used to send signals like sigterm)
2. Then I had to change the timeout in which keys are pressed. When using stdin, you get a keydown event, but you don't get a key up event (ever). So I have to guess when you want to key up. Most of the time, I can send key up right away. But, it looks like doom has some sort of key debounce, so I had to wait 50-100 ms for keyup. Then there is the problem of if you want to walk forward in games you usually hold down up arrow, but now you have to rapidly press it! Not ideal, but it does work, and it it playable.
I used to write telnet games so I know all about keypress up never coming through. Even with immediate mode (so repeated keys will send repeated key codes while held down) it never tells you when it stops. You have to read the buffer ascii byte by byte. Still, awesome to see. Great work!!!
This is interesting, but there was something that was even more impressive many years ago: a GTK theme that rendered all decoration and widgets using text chars and a GDK backend that rendered to text. Combine both and you could run any GTK app on a terminal with legible text and a beautiful TUI.
This is such a cool project. Personally, I think there are so many interesting use cases that can be built on top of Wayland, like https://github.com/udevbe/greenfield
and this
A few years back, I was deploying, operating and debugging a Hadoop cluster with Kerberos enabled behind a firewall with only the SSH port being opened. Without a web browser would have been a much more complicated task. I ended up installing the X11 client on my local macOS and the all Gnome + Firefox on one of the cluster's node. Something that is not doable with Wayland. This project work like a charm, here is a quick example on how to test it inside an Incus container (I had to install 2 additional dependencies).
# Work with Gnome terminal but resolution is much better in something supporting images
apt install -y kitty
kitty
# Create an incus container
incus --project default launch images:ubuntu/24.04 term
incus --project default shell term
# Install dependencies
apt install -y curl firefox libharfbuzz0b libfontconfig1
curl -L -o term https://github.com/mmulet/term.everything/releases/download/0.5.1/term.everything.mmulet.com-dont_forget_to_chmod_+x_this_file
mv term.everything.mmulet.com-dont_forget_to_chmod_+x_this_file term
chmod u+x term
echo '<h1>Hello</h1>' > test.html
# Start firefox, wait for a few seconds
./term firefox test.html
I remember the carbonyl project to run chromium in the terminal that got me really excited (https://github.com/fathyb/carbonyl) but it eventually became unmaintained.
This is pretty much that but supercharged. Definitely really cool to see. Good work!
Looks like you're on Windows? You can run X apps with XMing, I used to do it years ago. You can run the actual X app and use it, not just get a screenshot.
I truly appreciate the relational thinking and pointing out other projects that might interest ppl who are excited about this :) Having said that, term.everything seems to be much larger in scope than a browser, unless I'm mistaken
This could be useful for testing UI elements of apps ...
Modern UI applications are way too tightly coupled for my liking, and difficult to test especially if you don't practice "separation of concerns", e.g. decoupling the app logic from its presentation.
Haven't looked at the full thing but something like this might allow you to write tests for UI apps without actually having the UI backend...
This is pretty cool, I can see this being useful when I need to run a one-off remotely. Not sure about attaching a running program then detaching again, or mirroring... I wouldn't mind being able to SSH to my desktop and manipulate say the running Discord client, or similar.
Another similar thing that I'd been meaning to look into is the RDP remote apps stuff.
It was an example, not the only use case. 99% of what I want to do remote is just fine over SSH (over Wireguard)... Mostly remote VS Code usage since my desktop is much beefier than my laptop when travelling.
I like it. I always want to run things in a terminal. Because
1. I used to think that's more secure than X
2. I always seem to get better audio of the videos that I run in tty and my mouse is much smoother in the tty. Yes I can move mouse in tty.
Also someone mentioned a cool project like carbonyl. They also mentioned brow.sh which I have heard but they described it in detail so that's another plus when term.everything kind of projects come they drag other cool projects to he foreground
Point 1 of mine may be pure superstition.
How term.everything works on tty I don't know maybe it will be horriblebecause of the resolution thing but still it's a nice direction.
I started working on this with the Kitty image protocol, but unfortunately that protocol is really unsuited to this sort of thing. Performance will be awful.
The protocol is sort of:
1. I'd like you to display this PNG. Here's the data: ...
2. Ok I've got the data.
3. Ok now display it at this position.
4. Ok now remove it from the screen.
We're talking motion-PNG here. Just think about how awful that is.
I wish someone would add some kind of AV1-over-terminal protocol. That would be actually useful.
The other thing I was going to try was a custom GUI that used normal terminal text for the text of widgets, but Kitty images for the rest. It's quite a hard problem though.
What you're describing is a graphical shell. If you want it over the network, we have a protocol for that, it's called X. Misusing a terminal for this is fundamentally pointless.
Nope. X is too slow to be useful except on local networks, and alternatives like FreeNX or xpra are difficulty and janky to set up. Also X is dying so you really mean something like waypipe or worse (which also jankily failed to run some external command when I last tried it).
But I don't want that anyway. I want something graphical that's actually integrated into the terminal.
This could be used on build machines I own where I occasionally need to interact with the desktop and/or browser on the machine and vnc or other desktop sharing is impractical or exposes security issues.
Great job! If you tug on this thread long and hard enough, you develop this enough and you get RDP (which you can try via xrdp, GNOME's remoting thing, etc.).
The reason the terminal ecosystem doesn't get much more sophisticated over time isn't just the herd-of-cats fragmentation, but also evaporative cooling: people who do really cool things with terminal come to realize that what they really want is remote desktop (perhaps rootless) and leave terminal stuff as-is while they invest in more sophisticated systems instead.
Wow. This is amazing. I have started running a lot of stuff in containers by default for a whole host of reasons, and this may make my workflow even better on the occasions when I want to run a graphical app.
It is funny but this is what I wished things did when I first started using Linux back in the day. '98-'99 timeframe, then I "learned" better that there was Xorg/X11,etc.
1. Yes, but it depends on your compositor because your compositor needs to be able to run as a nested Wayland client. I think there is support for this in wlroots based Wayland compositors, but I'm not sure if hyperland supports it.
2. I think it will work, but I haven't tried. I redraw the terminal window every time the "termed" window updates. So, if you are playing a video for example and you dynamically resize the window, it should update the size automatically. If you are viewing a static window it might not.
Neat! I did a similar project many years ago just to see if I could with ANSI color stuff to animate video in my terminal. Worked really well, but it looked like absolute butt (unlike this project).
Stupid, love it. Occasionally I'll use shaderglass ascii shader on oled screen to play videos with pixel ratio that makes UI unreadable, but it's charming experience.
I tried recently once again to ditch Windows for Linux. Everything kinda worked, but the MediaTek Wi-Fi drivers were janky and my speed was like 10x slower than it should’ve been. After spending about 10 hours messing around with configs, I realized I was doing literally everything except what I actually wanted to do when I turned on the PC… so I just went back and installed LTSC Windows again.
I don't really understand the relevance of this comment to this thread, but since it's here...I remember running into something somewhat similar when trying to dual-boot Windows for something on a machine I already had Linux installed on, and while I can't remember whether it was actually MediaTek or not, I think it might have been. If my recollection is correct, I ended up figuring out that having the wifi configured to 1 Gbps in Windows somehow reverted the wifi to only 100 Mbps in Linux, and the only way for me to fix it was to boot back into Windows and switch it to whatever it had been by default (I think 100 Mbps?). Not sure if this is something you care enough to actually try out or not, but I figured it couldn't hurt to mention!
Depends on what resolution your terminal is set to. (Not the resolution of the GUI app you use, just the resolution you display it). At low resolution (640x480) it’s pretty performant, but at 4K I can hear my fans going full blast.
I definitely want to make a macOS version, but I haven’t even looked into it yet. So, I don’t know the level of hacking required. It definitely doesn’t sound like anything Apple would have an api for, so it would probably be a vnc or accessibility api trick.
I think there is an API (that was added with sidecar) to create a virtual display. So at best you could retrieve the framebuffer and then display that. I don't think there is an easy exposed way to get per-window information, aside from doing a screencapture (which likely would not work if you also wanted to hide the window).
Interesting… do you have a link to the docs? The easiest thing prob would be to set the window to be fullscreen, but on that virtual display. That would accomplish per window screen capture.
But really, in the last 24 hours term.everything has accumulated 1500 stars, 600 upvotes on HN, 185 upvotes on lobsters (the highest upvoted `show` tag of all time), but despite all of that, you, my friend, are the first to ask about Gwerm.
That means you win the secret prize! A custom Gwerm T-shirt! I’ll send the details to the Gmail you have linked to your account.
saw your email, thanks! As much as I would want a customized Gwerm-Jörmungandr-ouroboros t-shirt (just a random though!), I think you have already more than earned the kudos with your work, so let me just give my thanks to you for something that brightened my day today.
Awww, thanks for your kind words. Tell you what: should the fates align and we ever meet at a tech conference or hackathon, etc, I’ll do what I can so that you leave with some Gwerm-flavored memorabilia.
No you do not. It works on x11 and Wayland host systems. I built the Wayland compositor from scratch and it does not have any dependencies on libwayland. So, you don’t have to install Wayland at all.
This is so cool - thank you! I have a very (ahem) useful purpose for this: I use a command line application that calls back to a browser during authentication and that alone prevented me from doing what I needed/wanted from an ssh terminal... I will now happily laugh my ass off as it launches firefox from inside my terminal every time I use it.
"I feel like every single day I hear about another terminal file viewer. I say, stop making terminal file viewers because you can just use the file viewer you already have! In your terminal!"
LMAO
You could use a terminal graphics protocol to render real graphics. But there is already waypipe¹ to do that kind of remoting. Without using an actual terminal.
> You could use a terminal graphics protocol to render real graphics.
It already does that[1].
> But there is already waypipe¹ to do that kind of remoting.
That requires Wayland on the client side, doesn't it? I don't expect this to be super-practical anyway, but it's fun to see how far you can push a terminal.
[1] "If your terminal supports images (like kitty or iterm2) you can render windows at full resolution (performance may degrade)."
I am sure it was a great and fun learning experience.
Well done !
EDIT: nevermind, doing this with Docker seems much easier than I expected [0]. I'll try it tomorrow, I'm curious to see if the proposed solution works on Windows as well.
[0] https://medium.com/@priyamsanodiya340/running-gui-applicatio...
WSL had non-accelerated wayland support at one point in time.
Was it before "most distros"?
I don't understand what you think that amounts to.
Wayland support in Linux-land typically means that the software supports running in your wayland compositor.
Windows famously is not a wayland compositor, no matter how much you try to bend reality.
But I remember the BB demo and I still remember these catchy s3m modules...
https://aa-project.sourceforge.net/bb/
And, well, not AA, but I still play today tons of text adventures and roguelikes (and BSDgames and such), and my main X environment it's CLI/TUI based except for CWM (Window manager), MPV/MuPDF/NSxiv (images) and djview4 for DJVU files..
your animal brain hardwired to discern the direction of gaze of the eyes facing you (citations appreciated)...
Helps me find the mouse cursor on a big screen if I lose track, even with small parallax angle.
On macOS, I just wiggle the mouse back and forth, and the cursor gets really big, it pops out at you.
I generally don't keep `xeyes` running. But it's a righteous, venerable hack.
Thanks for the reminder
Sure am glad of the broadband Internet and modern "powerhouse" PCs we have so readily available today. Hell, even the computer most everyone carries in their pocket these days is infinitely more powerful than the average desktop machines of my childhood. :)
[1] https://www.brow.sh/
I run a ttyd server to get terminal over https, and I have used carbonyl over that to get work done. That's limited to a web browser (to get access to resources not exposed via the public internet), so having full GUI support is very useful
But can it run doom?
I had the change a couple of line to make it work because term.everything takes input only from stdin (this way it works of ssh and is pretty broadly compatible across terminals).
1. I had to remap another key to the control key (which is usually used to send signals like sigterm)
2. Then I had to change the timeout in which keys are pressed. When using stdin, you get a keydown event, but you don't get a key up event (ever). So I have to guess when you want to key up. Most of the time, I can send key up right away. But, it looks like doom has some sort of key debounce, so I had to wait 50-100 ms for keyup. Then there is the problem of if you want to walk forward in games you usually hold down up arrow, but now you have to rapidly press it! Not ideal, but it does work, and it it playable.
I used to write telnet games so I know all about keypress up never coming through. Even with immediate mode (so repeated keys will send repeated key codes while held down) it never tells you when it stops. You have to read the buffer ascii byte by byte. Still, awesome to see. Great work!!!
http://zemljanka.sourceforge.net/cursed/screenshots/
This is pretty much that but supercharged. Definitely really cool to see. Good work!
For example...
It looks like this: https://imgur.com/a/Eq2ToVOObviously no input though, you would have to use xdotool! The main benefit is that you probably already have all these tools installed :)
https://github.com/fathyb/carbonyl
P.S. This is very cool btw.
https://github.com/chase/awrit
Modern UI applications are way too tightly coupled for my liking, and difficult to test especially if you don't practice "separation of concerns", e.g. decoupling the app logic from its presentation.
Haven't looked at the full thing but something like this might allow you to write tests for UI apps without actually having the UI backend...
Another similar thing that I'd been meaning to look into is the RDP remote apps stuff.
Hopefully supports iPadOS one day.
Absolutely no mouse support though, anyone know of an iPad ssh client with mouse support?
(still working on getting vscode to run smoothly)
Blink terminal for iOS and iPadOS. Ships with vscode support built in too.
Prompt 3 from Panic
Also someone mentioned a cool project like carbonyl. They also mentioned brow.sh which I have heard but they described it in detail so that's another plus when term.everything kind of projects come they drag other cool projects to he foreground
Point 1 of mine may be pure superstition.
How term.everything works on tty I don't know maybe it will be horriblebecause of the resolution thing but still it's a nice direction.
The protocol is sort of:
1. I'd like you to display this PNG. Here's the data: ...
2. Ok I've got the data.
3. Ok now display it at this position.
4. Ok now remove it from the screen.
We're talking motion-PNG here. Just think about how awful that is.
I wish someone would add some kind of AV1-over-terminal protocol. That would be actually useful.
The other thing I was going to try was a custom GUI that used normal terminal text for the text of widgets, but Kitty images for the rest. It's quite a hard problem though.
But I don't want that anyway. I want something graphical that's actually integrated into the terminal.
The reason the terminal ecosystem doesn't get much more sophisticated over time isn't just the herd-of-cats fragmentation, but also evaporative cooling: people who do really cool things with terminal come to realize that what they really want is remote desktop (perhaps rootless) and leave terminal stuff as-is while they invest in more sophisticated systems instead.
I love it.
1: https://github.com/Julien-cpsn/desktop-tui
https://news.ycombinator.com/item?id=45203700
Good luck man, you rock
- doesTerm.everything run inside tmux with automatic window resizing? I guess not, but it would be cool
2. I think it will work, but I haven't tried. I redraw the terminal window every time the "termed" window updates. So, if you are playing a video for example and you dynamically resize the window, it should update the size automatically. If you are viewing a static window it might not.
Nicely done!
I would go for weeks just in a large framebuffer terminal, no GUI running. And I still run some servers that way.
Terminally insanely great!
https://github.com/enfp-dev-studio/node-mac-virtual-display/...
It isn't: and even copy paste is hard. Clever people write apps that are bash_completion friendly.
If first main arg is bash friendly
mycli myfunc ...
Myour whole cliapp becomes "discoverable" with one tab keystroke that you probably already typed hopefully anyway.
Never need to advertise a new feature.
Deprecate by removing from completion without breaking scripts.
Then _everything_ already is in your cli, because someone already did it.
That means you win the secret prize! A custom Gwerm T-shirt! I’ll send the details to the Gmail you have linked to your account.
A note to myself: this won't work in the text mode.
Thanks for sharing!
https://github.com/wayland-transpositor/wprs
Also, I'm lost for words, this is plain awesome.
1. https://gitlab.freedesktop.org/mstoeckl/waypipe
It already does that[1].
> But there is already waypipe¹ to do that kind of remoting.
That requires Wayland on the client side, doesn't it? I don't expect this to be super-practical anyway, but it's fun to see how far you can push a terminal.
[1] "If your terminal supports images (like kitty or iterm2) you can render windows at full resolution (performance may degrade)."