Welcome to KMonad! If you are currently running Linux and want to experiment with this new way of defining keyboard layouts, either:
- Head over to releases for the latest static binary.
- Consult the installation section
We now have (limited) Windows support! Look down below for instructions on how to build your Windows-version of KMonad, or check out our Windows binary. Check out the limitations on windows support below.
KMonad is a keyboard remapping utility written to provide functionality that aligns with that provided by the amazing QMK firmware. The QMK firmware is compiled and installed on programmable keyboards and allows your keyboards to transform from comfortable clacky ("Wahaay!" for mechanical switches) to amazingly useful, by introducing the ability to overlap various maps of keys, have different functionality for the same key when held or tapped, or create buttons that can be tapped multiple times to have different effects.
However, we can't always have our programmable keyboard with us, and I
personally don't like using it with my laptop, because it either becomes a
balancing act, or I have to sit at a table with my laptop uncomfortably far away
from me. Additionally, there are loads of people who do not have programmable
keyboards who might enjoy all the bells and whistles that QMK has to offer, and
so KMonad
was born.
The name KMonad is a wink at my favorite window manager: XMonad, but where XMonad manages your X Window in its X-monad, we manage your Keys in our K-monad (actually, we call it App, but who's checking).
Well, there are a variety of customizations you can make. For example, this duplicates the functionality of another popular Linux utility: XCAPE, which allows you to specify mod-buttons that also emit events when you simply tap them.
For example, look at that useless CapsLock key on your keyboard. It is in a great position, but do we ever use it? Well, what if you could remap CapsLock to a button that behaves like a Ctrl key? That's already quite nice. What if it could also emit 'Esc' when you tap it? Yep: all of that is possible (Note: some of this is already possible through existing tools or udev rules).
Do you write a lot of Haskell code and define your own operators? Do you want your operators to look like this: =>=->->>? That's a lot to type. Why not just define a macro that does it for you?
The same technique can be used to create macros that insert a variety of
accented or umlauted characters, depending on your OS. Why type Right-Alt " e
when you can put this macro in a layer on a button close to home row?
Or more practically: do you really like the Dvorak layout, but do you also wish you had very simple, left-handed access to the common Control z x c v shortcuts? Have a look at this dvorak example to see an easy solution.
For more ideas, check the example
subdirectory, or the syntax
guide
for an overview of different button-types.
KMonad is written in Haskell (with a tiny bit of C). You can either compile it yourself using the instructions mentioned below. Alternatively, the lovely people over at https://github.com/nh2/static-haskell-nix have helped me figure how to compile a static binary that should work basically on any standard 64-bit Linux system. You can find the most recent release on the releases page.
Probably the easiest way to compile KMonad is using stack
. If you do not have stack
installed, check https://docs.haskellstack.org/en/stable/README/ for
instructions on installing it. After compilation, it can be removed again, since
kmomad
does not need to be recompiled upon configuration.
After potentially installing stack
and cloning this repo, you can build
kmonad
by calling:
stack build
Or call the following (CURRENTLY BROKEN, documentation will be fixed up in the future):
stack haddock --no-haddock-deps
to build a KMonad binary and the Haddock documentation. I have put some effort into documenting the code if you want to have a look around. It is nowhere near perfect, and I hope to do more in the future.
stack
will tell you where it saved the compiled binary after which you can
copy it to somewhere on your path.
Windows support was added under Windows10 using a Haskell Platform
installation. Additionally, you might need
to install MinGW to provide gcc
for windows to compile
the C-interface to Windows. Once both the Haskell Platform and MinGW are
installed and available on the path, compiling KMonad should be identical to
Linux, i.e.:
stack build
KMonad currently requires 1, and exactly 1 input argument: a path to a configuration file that describes the keyboard layout to run. For a guide to writing valid configuration files, see the syntax guide or some of the examples.
Once the compiled binary is on the PATH, running KMonad is as simple as:
kmonad /path/to/config/file.kbd
The method of running KMonad under Windows is exactly the same: you use the shell (for example: Powershell) to start KMonad. For example, put the compiled KMonad executable and config file in the same directory, start Powershell, cd to the directory, and run:
./kmonad config_file.kbd
This has the added benefit that, if KMonad experiences issues, you can use your mouse to close the powershell and hopefully release the keyboard-hook.
Note that this interface is extremely provisional and subject to change.
Any kind of internal KMonad error that indicates that something has gone seriously wrong with our representation of the computation will terminate KMonad and display the error to stdout. It is however not uncommon for KMonad to have to reacquire a uinput keyboard on resume from suspend. To that extent, any core IO exception will cause KMonad to pause for a second and attempt a restart, ad infinitum. This means its fine to unplug the mapped keyboard and plug it back in, without crashing KMonad.
Currently, the only supported operating system is Linux. KMonad uses the
uinput
subsystem to write events to the operating system. If you want to be
able to run KMonad without using sudo (highly recommended to avoid sudo wherever
possible), you will need to ensure that your user is part of the uinput
group.
On most Linux's this can be achieved by:
sudo usermod -aG uinput username
If the uinput
group does not exist, check whether your system has an input
group and try adding your user to that one instead. If this does not work,
create a new group with:
sudo groupadd uinput
You then have to add a udev rule (in either /etc/dev/rules.d
or
/lib/udev/rules.d
) with the following content:
KERNEL=="uinput", MODE="0660", GROUP="uinput", OPTIONS+="static_node=uinput"
Additionally, you might need to ensure that the uinput
drivers are loaded
before starting KMonad, this can be achieved through:
sudo modprobe uinput
This might have to be repeated whenever you restart your computer. There are
various techniques for getting the uinput
subsystem to load automatically, but
I didn't manage to get any of them to work.
Sometimes you can find your keyboard listed under /dev/input/by-id
. If so,
this is by far the best solution, since there is no guarantee that a keyboard
will be assigned the same numbered event-file. If this is not the case, however,
the easiest way to figure out which event-file corresponds to your keyboard is
probably to use the evtest
Linux utility.
Since KMonad only deals in 'raw', primitive keyboard events, there is no such thing at that level as a special symbol. Instead we emit common keyboard sequences that the operating system needs to map to special characters. To that extent, you need to indicate to X11 what key is supposed to trigger a special-character macro.
There are two ways of doing this:
- Manually, after launching KMonad, use either
xmodmap
orsetxkbmap
to indicate to your OS that 'Right-Alt' should be used as the compose key (support for other compose keys is coming in the future). For example:
# Either
xmodmap -e "keysym Alt_R = Multi_key"
# or:
setxkbmap option compose:ralt
It is probably better to use setxkbmap
here, since it resets your config
before applying modifications, whereas repeated calls to xmodmap
can run into
errors because you are trying to map to buttons that have already been remapped.
- Automatically, through the UINPUT_SINK token. If you consult the syntax guide you will see exactly how you can provide KMonad with a shell-command to execute whenever a new uinput sink is created. This has the added benefit that, whenever we need to recreate the uinput sink (this is sometimes necessary after resuming from suspend, for example), the command is automatically called again for you.
Note that there is a small interval between creating a uinput sink and it
actually being registered by the OS, so whether you manually call setxkbmap
or
use the UINPUT_SINK token to pass a shell command, you need to ensure that it
contains a small period of time for the OS to register the keyboard. I have
found that 1 second is more than sufficient, but experiment yourself.
The low-level API to the operating system differs significantly between Windows
and Linux, which means that the Windows version is currently more limited in
what it can do. There is an active issue on this topic over
here, and if you have
experience with Win32
programming, any help would be greatly appreciated. So
if you want to help, or just want a more technical overview of the windows
limitations head on over there.
Currently, we cannot distinguish between different input keyboards, so whereas a Linux version of KMonad can be started for a variety of different keyboards, and handle them all in different ways, the Windows version of KMonad catches all keyboard input signals. The only distinction KMonad makes under Windows is between 'real' keyboard events and simulated keyboard events. Anything simulated is automatically passed on to the OS (that is also how KMonad avoids handling its own simulated output).
Windows does not support the same compose-sequences as X11, meaning that the special-character-emitting macros won't work out of the box. Luckily there is WinCompose, a windows utility that maps compose-key sequences to special characters. They say they have full support for all X11 based compose-sequences, and in my limited test I did not run into any problems. We currently do not support remapping the compose-key internally, but the KMonad default lines up with the X11 default and the WinCompose default (right Alt).
NOTE: For WinCompose to work with KMonad you have to enable support for handling injected key events, which is off by default.
Since KMonad essentially grabs all standard keyboard input and lets through only simulated events, there is no guarantee at all of this playing nice with AutoHotKey at the moment. Additionally, I am not entirely sure how Windows deals with its low-level keyboard hooks and how AHK tries to get at keyboard input, but it might even be the case that different startup-orders could result in different behavior. I have no experience with AHK at all, and rarely use Windows. If you run into any issues, please file them, and I'm sure that in time we can resolve everything.
Many laptops have a Fn key that mimics some of the functionality that KMonad tries to offer: it changes the mapping of certain keys, like creating a numpad in the middle of the laptop keyboard. This remapping happens in the hardware, before any event is ever registered with the operating system, therefore KMonad has no way to 'get' at any of those events. This means that we cannot remap them in any way.