So, while working with caching and scrapping, I understood the difference between immutable and mutable objects/datatypes very clearly. I had a scenario, where I am webscraping an API, the code looks like this.

from aiocache import cached

@cached(ttl=7200)
async def get_forecast(station_id: str) -> list[dict]:
 data: dict = await scrape_weather(station_id)
 # doing some operation
 return forecasts

and then using this utility tool in the endpoint.

async def get_forecast_by_city(
 param: Annotated[StationIDQuery, Query()],
) -> list[UpcomingForecast]:
 forecasts_dict: list[dict] = await get_forecast(param.station_id)
 forecasts_dict.reversed()

 forecasts: deque[UpcomingForecast] = deque([])
 for forecast in forecasts_dict:
 date_delta: int = (
 date.fromisoformat(forecast["forecast_date"]) - date.today()
 ).days
 if date_delta <= 0:
 break
 forecasts.appendleft(UpcomingForecast.model_validate(forecast))

 return list(forecasts)

But, here is the gotcha, something I was doing inherently wrong. Lists in python are mutable objects. So, reversing the list modifies the list in place, without creating a new reference of the list. My initial approach was to do this

Welcome to a new issue of This Week in Plasma!

This week we put the finishing touches on Plasma 6.6! It’s due to be released in just a few days and it’s a great release with tons of impactful features, UI improvements, and bug fixes. I hope everyone loves it!

Meanwhile, check out what folks were up to this week:

Notable UI Improvements

Plasma 6.7.0

Moved System Settings’ “Remote Desktop” page to the “Security & Privacy” group in System Settings. (Nate Graham, krdp MR #139)

After

Before

Improved the way loop devices are handled in the Disks & Devices widget. (Bogdan Onofriichuk, plasma-workspace MR #6260)

Reduced visual jagginess in the split image effect of wallpaper previews that show both a light and dark version. (Fushan Wen, plasma-workspace MR #6283)

The Kicker Application Menu widget now supports using a non-square icon for its panel button, just like Kickoff does. (Christoph Wolk, plasma-desktop MR #3522)

Added a dedicated global action for un-tiling a quick-tiled window. It doesn’t have a keyboard shortcut by default, but you can assign one yourself. (Kevin Azzam, KDE Bugzilla #500636)

Videos in SDDM login screen themes can now be previewed in System Settings. (Blue Terracotta, sddm-kcm MR #99)

Improved the appearance of various dialogs created by KWin. Read more here! (Kai Uwe Broulik, kwin MR #8702)

You can now configure how long it takes the window switcher to appear after you start holding down Alt+Tab. (Guilherme Soares, KDE Bugzilla #486389)

Frameworks 6.24

In Kirigami-based apps, hovering the pointer over buttons that can be triggered with keyboard shortcuts now shows the shortcuts. (Joshua Goins, kirigami MR #2040)

After

Before

Gear 26.04.0

Setting up a Samba share for one of your folders so people can connect to it over the network now turns on the Samba service (on systemd-based distros) if needed. This completes the project to make Samba sharing relatively painless! (Thomas Duckworth, KDE Bugzilla #466787)

Notable Bug Fixes

Plasma 6.5.6

Monitor names shown in the Brightness & Color widget now update as expected if you connect or disconnect them while the system is asleep. (Xaver Hugl, KDE Bugzilla #495223)

Fixed multiple issues that caused custom-tiled windows on screens that you disconnect to move to the wrong places on any of the remaining screens. (Xaver Hugl, kwin MR #7999)

Plasma 6.6.0

Hardened KWin a bit against crashing when the graphics driver resets unexpectedly. (Vlad Zahorodnii, kwin MR #8769)

Fixed a case where Plasma could crash when used with the i3 tiling window manager. (Tobias Fella, KDE Bugzilla #511428)

Fixed a potential “division by 0” issue in system monitoring widgets and apps that messed up the display percentage of Swap sensors on systems with no swap space. (Kartikeya Tyagi, libksysguard MR #462)

Worked around a Wayland bug (yes, an actual Wayland bug — as in, a bug in one of its protocol definitions!) related to input method key repeat. Why work around the bug instead of fixing it? Because it’s already fixed in a newer version of the protocol, but KWin needs to handle apps that use the old version, too. (Xuetian Weng, kwin MR #8700)

Unified the appearance of HDR content in full-screen windows and windowed windows. (Xaver Hugl, KDE Bugzilla #513895)

Fixed a layout glitch in the System Tray caused by widgets that include line breaks (i.e. \n characters) in their names. (Christoph Wolk, KDE Bugzilla #515699)

The Web Browser widget no longer incorrectly claims that every page you visit tried to open a pop-up window. (Christoph Wolk, kdeplasma-addons MR #1003)

Fixed a layout glitch in the Quick Launch widget’s popup. (Christoph Wolk, kdeplasma-addons MR #1004)

You can now launch an app in your launcher widget’s favorites list right after overriding its .desktop file; no restart of plasmashell is required anymore. (Alexey Rochev, KDE Bugzilla #512332)

Fixed an issue that made inactive windows dragged from their titlebars get raised even when explicitly configured not to raise in this case. (Vlad Zahorodnii, KDE Bugzilla #508151)

Plasma 6.6.1

When a battery-powered device is at a critically low power level, putting it to sleep and charging it to a normal level no longer makes it incorrectly run the “oh no, I’m critically low” action immediately after it wakes up. (Michael Spears, powerdevil MR #607)

The overall app ratings shown in Discover now match a simple average of the individual ratings. (Akseli Lahtinen, KDE Bugzilla #513139)

Searching for Activities using KRunner and KRunner-powered searches now works again. (Simone Checchia, KDE Bugzilla #513761

Frameworks 6.24

Worked around a Qt bug that caused extremely strange cache-related issues throughout Plasma and Kirigami-based apps that would randomly break certain components. (Tobias Fella, kirigami MR #2039)

Notable in Performance & Technical

Plasma 6.6.0

Added support for setting custom modes for virtual screens. (Xaver Hugl, kwin MR #8766)

Added GPU temperature monitoring support for additional GPUs. (Barry Strong, ksystemstats MR #123)

Plasma 6.7.0

Scrolling in scrollable views spawned by KWin (not Plasma, just KWin itself) no longer goes 8 times slower than it ought to. Thankfully there are very few such views, so almost nobody noticed the issue. However fixing it facilitates adding a “scroll to switch virtual desktops” feature to the Overview effect for Plasma 6.7! (Kai Uwe Broulik, kwin MR #8800)

Frameworks

Moving a file to the trash is now up to 50 times faster and more efficient. (Kai Uwe Broulik, kio MR #2147)

How You Can Help

KDE has become important in the world, and your time and contributions have helped us get there. As we grow, we need your support to keep KDE sustainable.

Would you like to help put together this weekly report? Introduce yourself in the Matrix room and join the team!

Beyond that, you can help KDE by directly getting involved in any other projects. Donating time is actually more impactful than donating money. Each contributor makes a huge difference in KDE — you are not a number or a cog in a machine! You don’t have to be a programmer, either; many other opportunities exist.

You can also help out by making a donation! This helps cover operational costs, salaries, travel expenses for contributors, and in general just keep KDE bringing Free Software to the world.

To get a new Plasma feature or a bug fix mentioned here

Push a commit to the relevant merge request on invent.kde.org.

Hey everyone!

I am Siddharth Chopra, a second year engineering student at the Indian Institute of Technology, Roorkee. I'm really excited to be working on Marknote as a part of the Season of KDE program this year, under the mentorship of Carl Schawn.

Marknote, as it is aptly named, is KDE's own markdown based note taking app. The aim of my project is to improve Marknote by adding the much requested source mode, alongside other enhancements.

Progress so far

3 weeks into the project, I have been successful in adding a working source mode functionality to the editor. When source mode is activated, the contents of the source markdown file are allowed to be edited directly, instead of showing the rendered markdown. This is incredibly useful, in cases where the user needs manual control over the contents of the note, or in case there is some glitch in the rendering (which unfortunately still happens often).

Demo Video

Technical Roadmap & Challenges

The main editor of Marknote comes from the file EditPage.qml. As part of my initial approach, I added a global property here to check if source mode was enabled, and then conditionally changed components of this editor. Although this worked, but it brought along some of its own issues. First of all it made the code unnecessarily complex. It also meant that components like the formatting bar now needed to be repurposed to work with source mode, which is a challenge in itself.

So, my mentor suggested to move the raw editor into a new file, to keep the code maintainable. This led to the original EditPage being split into RichEditPage and RawEditPage. Similarly, the respective backends were also split in two, as the needs of both the editors are significantly distinct.

Additionally, I had to consider specially the source mode for images. Because when loaded, image URL's are not kept intact, instead they are replaced with a hash, that maps to the image in memory. Also, the editor internally uses html for rendering images, which I also had to convert back to markdown for source mode.

And someone who is not a designer by any means, deciding the form and placement of the mode toggle button was in itself a mini lesson in UI design ;) Initially I went with a toggle switch. But when I shared that for feedback, I learnt that a checkable button is the ideal UI element here.

Future Plans

My proposal mentions features apart from the source mode, which I plan to complete. While working on the current feature, I noticed multiple bugs in the app, which I intend to fix as well.

Overall experience

It has been a great experience working with the KDE community, and really exciting to be able to contribute to an app that so many users around the world use every day! I would also like to express my gratitude towards my mentor, for being there for whatever issue I faced!

This release brings improvements to generators, better build system integration and several bugfixes.

As always, big thanks to everyone who reported issues and contributed to QCoro. Your help is much appreciated!

Directly Awaiting Qt Types in AsyncGenerator Coroutines

The biggest improvement in this release is that QCoro::AsyncGenerator coroutines now support directly co_awaiting Qt types without the qCoro() wrapper, just like QCoro::Task coroutines already do (#292).

Previously, if you wanted to await a QNetworkReply inside an AsyncGenerator, you had to wrap it with qCoro():

QCoro::AsyncGenerator<QByteArray> fetchPages(QNetworkAccessManager &nam, QStringList urls) {
 for (const auto &url : urls) {
 auto *reply = co_await qCoro(nam.get(QNetworkRequest{QUrl{url}}));
 co_yield reply->readAll();
 }
}

Starting with QCoro 0.13.0, you can co_await directly, just like in QCoro::Task:

QCoro::AsyncGenerator<QByteArray> fetchPages(QNetworkAccessManager &nam, QStringList urls) {
 for (const auto &url : urls) {
 auto *reply = co_await nam.get(QNetworkRequest{QUrl{url}});
 co_yield reply->readAll();
 }
}

Other Features and Changes

• Generator’s .end() method is now const (and constexpr), so it can be called on const generator objects (#294).
• GeneratorIterator can now be constructed in an invalid state, allowing lazy initialization of iterators (#318).
• qcoro.h now only includes QtNetwork and QtDBus headers when those features are actually enabled, resulting in cleaner builds when optional modules are disabled (#280).

Bugfixes

• Fixed memory leak in QFuture coro wrapper when a task is destroyed while awaiting on a QFuture (#312, Daniel Vr√°til)
• Fixed include paths when using QCoro with CMake’s FetchContent (#282, Daniel Vr√°til; #310, Nicolas Fella)
• Fixed QCoroNetworkReply test on Qt 6.10 (#305, Daniel Vr√°til)

Full changelog

See changelog on Github

Support

If you enjoy using QCoro, consider supporting its development on GitHub Sponsors or buy me a coffee on Ko-fi (after all, more coffee means more code, right?).

Going into Week 3, my next task was clear. I had successfully stopped the Lokalize menubar from jumping around, but it left me with top-level menus that were clickable but completely empty depending on which tab you were using. My goal was to figure out how to dynamically grey out those useless menu headers, like the "Sync" menu in the Project Overview, so users wouldn't be tricked into clicking them.

I started by checking the Qt documentation and a very helpful discussion on the Qt forums to see how to properly handle this. The forum thread highlighted a crucial distinction in the Qt API: using setVisible(false) completely hides a menu, while setEnabled(false) keeps it visible but greys it out. This was an important detail. If i had used setVisible(), the menus would have vanished completely, bringing back the jumping ui bug I had just fixed in Week 1. I needed to use setEnabled() to keep the menu structure frozen in place while making the empty headers unclickable.

To implement this, I hooked directly into the application's tab-switching logic. I wrote a new function that gets triggered every single time a user changes tabs. It grabs the main menubar and loops through every top-level action. For each menu it finds, it checks if the menu has any currently visible actions. If the answer is no, it disables that specific menu header using action->setEnabled(false), turning the menu grey so the user knows it cannot be clicked.

void LokalizeMainWindow::updateMenuAvailability()
{
    QMenuBar *bar = menuBar();
    if (!bar)
        return;

    // Refresh top-level menu state based on currently available actions.
    for (QAction *action : bar->actions()) {
        QMenu *menu = action->menu();
        if (!menu)
            continue;

        // Disable top-level menu when it has no visible actions.
        action->setEnabled(menuHasVisibleAction(menu));
    }
}

bool LokalizeMainWindow::menuHasVisibleAction(const QMenu *menu) const
{
    for (QAction *action : menu->actions()) {
        if (action->isSeparator() || !action->isVisible())
            continue;

        if (QMenu *subMenu = action->menu()) {
            // Check submenu actions too.
            if (menuHasVisibleAction(subMenu))
                return true;
            continue;
        }
        return true;
    }
    return false;
}

The tricky part was writing the helper function to accurately figure out if a menu was actually "empty." It wasn't enough to just check if the menu contained items. Inside the helper, I loop through the menu's actions. If an item is just a ui line (a separator) or is already hidden, the code skips it. I also had to account for nested submenus, so if the action is a submenu, the function calls itself recursively to check inside it. The implementation is completely generic-meaning if anyone adds new menus or actions to Lokalize in the future, this code will handle them automatically.

I opened the merge request, and i'm happy to say it was quickly merged into master! The Lokalize menus now update smoothly based on the context of the active tab. They stay exactly where they belong, and empty menus disable themselves as expected. With this menubar issue resolved:)

As explained in one of my previous blog posts where I revamped the unresponsive window dialog, KWin isn’t really designed to show regular desktop windows of its own. It instead relies on helper programs to display messages. In case of the “no border” hint, it just launched kdialog, a small utility for displaying various message boxes from shell scripts. This however came with a couple of caveats that have all been addressed now:

First of all, the dialog wasn’t attached to the window that provoked it. When the window was closed, minimized, or its border manually restored, it remained until manually dismissed. Secondly, it said “KDialog” and used a generic window icon (that would have been an easy fix of course). Further, the user might not even have kdialog installed which actually was the case on KDE Linux until very recently. Ultimately and most importantly, it told you that you were screwed if you didn’t have a keyboard but didn’t offer and help if you really went without one. I therefore added an option to restore the border right from the dialog. Should you have a dedicated global shortcut configured for this action, it will also be mentioned in the dialog.

The dialog when manually setting a window full-screen has similarly been overhauled, including an undo option. While at it, I removed the last remaining code calling kdialog, too: the “Alt+tab switcher is broken” message. It is now a proper KNotification. Something you should never see, of course.

Another dialog that I gave some attention was the prompt when copying a file would overwrite its destination. If you have Kompare installed and copy a plain text file (that includes scripts, source code, and so on), the dialog displays a “Compare Files” button. It already guesstimated whether the files are similar but now you can actually see it for yourself.

KIO’s PreviewJob that asynchronously generates a file preview now provides the result of its internal stat call, too. This means that once you receive the preview, you also get the file’s properties, such as size, modified time, and file type basically for free. The rename dialog then displays this information in case it wasn’t provided by the application already. Dolphin now also makes use of this information while browsing a folder which should improve responsiveness when browsing NFS and similarly mounted network shares. At least when previews are enabled, it no longer determines file types synchronously in most scenarios.

Since the rename dialog is able to fetch file information on demand, Ark, KDE’s archiver tool, rewrites the source URL it displays in the dialog to a zip:/ URL (or tar:/ or whatever supported archive type). This way the dialog can display a preview transparently through the Archive KIO worker which also gained the ability to determine a file’s type from its contents. In case you didn’t know, you can configure Dolphin to open archives like regular folders.

Finally, most labels in KIO that show the mtime/ctime/atime no longer include the time zone unless it is different from the system time zone. Showing “Central European Standard Time” in full after every date was a bit silly. Unfortunately, QLocale isn’t very flexible and only knows “short” (19:00) or “long” (19:00:00 Central European Standard Time) formats. You can’t explicitly tell it to generate a time string with seconds, or with 12/24 hour clock, or a date with weekday but no year, and unfortunately the “long” time format includes the full time zone name.

The final final 5.2.x release has been made, and the first beta for 5.3.0/6.0.0 is out!

Read on for a look at development news and the Krita-Artists forum's featured artwork from last month.

Development Report

5.2.15 Released

5.2.15, another bugfix release, is out, featuring a few more Android, touch input, and general bug fixes. This is really the last 5.2.x release this time.

Check out the 5.2.15 release post and stay up-to-date!

First Beta for 5.3.0/6.0.0 Released

The next major release will be a dual release; 5.3.0 is built on the familiar Qt5 framework, while 6.0.0 is ported to the newer Qt6 framework. Qt6 allows for better Wayland compositor support on Linux among other things, but unfamiliarity and lack of testing means Krita 6.0.0 will be less stable than 5.3.0. Krita 6 is also not yet available on Android.

The first betas for this release are out. Help out by testing the two years' worth of changes since 5.2 and the Qt6 porting work, and report any bugs you find so they can be fixed before the final release! Learn more in the beta1 release post.

Some of Last Month's Bugfixes

The team has been busy polishing new features and fixing old bugs for the beta release.

Many issues with gradients, layer styles, and resource loading were fixed by Dmitry.

• Layer Styles now use their own copy of gradients, preventing the gradients from unexpectedly changing when the global version is modified, as well as related crashes. (bug 1, bug 2, bug 3, bug 4; change)
• The Bevel-and-Emboss Layer Style now saves patterns correctly. (bug; change)
• Brush presets with embedded resources sharing a name with a different resource are now loaded correctly. (CCbug; change)
• Fixed editing or removing resource bundles causing a database error on startup. (bug; change)

More text fixes by Wolthera:

• Made sure new texts are made with the current foreground color. (change)
• Fixed an issue where light-colored text made Krita stall and then the text disappeared (caused by reading uninitialized data as font metrics). (bug; change)
• Fixed selecting fonts whose full name is the same as their family name. (bug; change)
• Fixed freeze when editing text contour while bound rect snapping is enabled. (change)
• Fixed setting subscript and superscript on the entire paragraph. (bug; change)

The G'MIC filter plugin has been updated by Ivan to version 3.6.6. (change)

On macOS, Krita's G'MIC is now faster, being built with OpenMP's multithreading as was already the case on other platforms. (change)

Community Report

January 2026 Monthly Art Challenge Results

The winner of the "Vintage Travel Poster" challenge is…

Titan by Elixiah

Join This Month's Art Challenge!

For February's theme, last month's winner Elixiah passed the choice of topic to runner-up MossGreenEddie, who has chosen "Alien World Building", with the optional challenge of including food in some way. How might members of an extra-terrestrial society interact? Cook up something out of this world!

Featured Artwork

This month's featured forum artwork, as voted in the Best of Krita-Artists - December 2025/January 2026:

Aging by ShangZhou0

Cat's Eye by _CR

Digital Watercolor Landscapes by daroart37

Photo drawing study: hands by pavuk0.2

Rooster 2025 by z586t

Nominate and Vote For Next Month's Featured Artwork!

Participate in next month's nominations and voting to voice your opinion on the Best of Krita-Artists - January/February 2026.

Krita is Free - But You Can Contribute!

Krita is free to use and modify, but it can only exist with the contributions of the community. A small sponsored team alongside volunteer programmers, artists, writers, testers, translators, and more from across the world keep development going.

If this software has value to you, consider donating to the Krita Development Fund. Or Get Involved and put your skills to use making Krita and its community better!

Additional Changes

Krita Plus (Stable, 5.2.15):

• Gradients: Fix an issue where reloading a gradient removes the transparency, by using premultipled alpha in SVG gradients. (bug; change by Dmitry Kazakov)
• Shortcuts: Prevent Transform/Move Tool move shortcuts from triggering when the tools aren't active, which could override other shortcuts (such as next/previous frame). (bug; change by Dmitry Kazakov)
• Overview Docker: Fix touch-dragging to pan, by disabling the non-existent context menu. (bug; change by Carsten Hartenfels)
• Tablets: Add workarounds for Xiaomi devices, which can be toggled in Configure Krita -> Tablet Settings: Translate the stylus's Page Up and Page Down buttons into right mouse and middle mouse buttons. Ignore jagged tablet stylus position history. Use a steeper default pressure curve so that it's possible to reach 100% pressure. (change by Carsten Hartenfels)

Krita Plus (Stable pre-release, 5.3.0/6.0.0-beta1):

• Comics Panel Editing Tool: Fix crash and other issues with remove gutter mode. (bug; change by Agata Cacko)
• Transform Tool: Liquify: Fix broken 'Accurate with Instant Preview' when zoomed out, and issues when liquifying on the edge of a selection. (bug 1, bug 2; change by Agata Cacko)
• Transform Tool: Show feedback that hidden or locked layers cannot be transformed, when attempted. (change by Agata Cacko)
• Move Tool: Disallow moving layers hidden by isolation mode. (change by Ricky Ringler)
• Freehand Path Tool: Fix preview not updating when changing color. (change by Luna Lovecraft)
• General: Fix crash when closing image after deleting a vector layer under some circumstances. (bug; change by Luna Lovecraft)
• General: Avoid Krita becoming unresponsive due to slow layer thumbnail generation. When thumbnail generation is slow, a warning in the statusbar mentions that the thumbnails have been switched to low quality mode and suggests using the new Image->Purge Unused Image Data action to clean up empty spaces of the image. (bug; change by Dmitry Kazakov)
• Layers Docker: Prevent Ctrl+drag-and-drop from selecting the layer. (change by Luna Lovecraft)
• Layers Docker: Fixes for selecting layers with Shift or Ctrl. (bug 1, bug 2; change by Luna Lovecraft)
• Touch Input: Prevent touch scrolling from triggering context menus. (bug; change by Carsten Hartenfels)
• Brush Editor: Fix Texture Options' offset sliders not working after enabling Pattern. (bug; change by Luna Lovecraft)
• Android: Recorder Docker: Remove the default recording folder path to avoid saving to a user-inaccessible appdata directory. The path can be set manually, which will move any existing recordings from the old default folder. (Note that rendering timelapse videos is not possible on Android Krita, but the frames and their document can be copied to another OS and rendered there.) (bug; change by Carsten Hartenfels)
• macOS: Fix lack of standard Qt text translations such as Yes/No buttons and the macOS application menu. (bug 1, bug 2; change by Ivan Yossi)

Krita Plus (Stable pre-release, 5.3.0/6.0.0-beta1+):

• File Formats: HEIF: Fix crash when opening HEIF/AVIF images in YCrBr. (bug; change by Dmitry Kazakov)

Krita Next (Unstable, 5.4.0/6.1.0-prealpha):

• General: Show hi-dpi versions of the recent file thumbnails on the welcome page. (change by Agata Cacko)

Nightly Builds

These pre-release versions of Krita are built every day.

Note that there are currently no Qt6 builds for Android.

Test out the upcoming Stable release in Krita Plus (5.3.0/6.0.0-prealpha): Linux Qt6 Qt5 — Windows Qt6 Qt5 — macOS Qt6 Qt5 — Android arm64 Qt5 – Android arm32 Qt5 – Android x86_64 Qt5

Or test out the latest Experimental features in Krita Next (5.4.0/6.1.0-prealpha). Feedback and bug reports are appreciated!: Linux Qt6 Qt5 — Windows Qt6 Qt5 — macOS Qt6 Qt5 — Android arm64 Qt5 – Android arm32 Qt5 – Android x86_64 Qt5

In my last post, I laid out the vision for Kapsule—a container-based extensibility layer for KDE Linux built on top of Incus. The pitch was simple: give users real, persistent development environments without compromising the immutable base system. At the time, it was a functional proof of concept living in my personal namespace.

Well, things have moved fast.

It's in KDE Linux now

Kapsule is integrated into KDE Linux. It shipped. People are using it. It hasn't caught fire.

The reception in #kde-linux:kde.org has been generally positive—which, if you've ever shipped anything to a community of Linux enthusiasts, you know is about the best outcome you can hope for. No pitchforks, some genuine excitement, and a few good bug reports. I'll take it.

Special shoutout to @aks:mx.scalie.zone, who has been daily-driving Kapsule and—crucially—hasn't had it blow up in their face. Having a real user exercising the system outside of my own testing has been invaluable.

Bug fixes

Before shipping, I spent a full day rigorously testing every main scenario I could think of—and writing integration tests to back them up. (Those tests run on my own machine for now, since the CI/CD pipelines don't exist yet. More on that below.) The result: the first version that landed in KDE Linux was quite stable. There was only one minor issue that isn't blocking anything.

CI/CD

I'm in the process of building out proper CI/CD pipelines. This is the unglamorous-but-essential work that turns a "project some guy hacked together" into something that can be maintained and contributed to by others. Automated builds, automated tests, the whole deal. Not exciting to talk about, but it's what separates hobby projects from real infrastructure.

Konsole integration

This is the one I'm most excited about. In the last post, I mentioned that Konsole gained container integration via !1171, and that I'd need to wire up an IContainerDetector for Kapsule. That work is underway, with two merge requests up:

!1178 (Phase 2) adds containers directly to the New Tab menu. You can see your available containers right there alongside your regular profiles—pick one, and you get a terminal session inside that container. Simple.

A big chunk of this work was refactoring the container listing to be fully async. There are a lot of cases where listing containers can take a while—distrobox list calls docker ps, which pokes docker.socket, which might be waiting on systemd-networkd-wait-online.target—and we absolutely cannot block the UI thread during all of that.

!1179 (Phase 3) takes it a step further: you can associate a container with a Konsole profile. This is what gets us to the "it just works" experience—configure your default profile to open in your container, and every new terminal is automatically inside it.

These lay the groundwork for Konsole to be aware of Kapsule containers. The end goal hasn't changed: open a terminal, you're in your container, you don't have to think about it. We're not at the "it just works" stage yet, but the foundation is being poured.

Future work: configurable host integration

Right now, Kapsule gives you some control over how tightly the container integrates with the host. The --session flag on kap create lets you control whether the host's D-Bus session socket gets mounted into the container. That's a good start, I think I need to go further.

The bigger issue is filesystem mounts. Currently, Kapsule unconditionally mounts / to /.kapsule/host and the user's home directory straight through to /home/user. That means anything running inside the container has full read-write access to your entire home directory.

That's fine when you trust everything you're running, but some tools are less predictable than others. There are horror stories floating around of autonomous coding agents hallucinating paths and rm -rfing directories they had no business touching. Containers are a natural mitigation for this—if something goes sideways, the blast radius is limited to what you explicitly shared. But that only works if you can actually control what gets shared.

The fix is making filesystem mounts configurable. Instead of unconditionally exposing everything, you should be able to say "this container only gets access to ~/src/some-project" and nothing else. Want a fully integrated container that feels like your host? Mount everything. Want a sandboxed environment for running less predictable tools? Expose only what's needed. The trust model should be a dial, not a switch.

A word of caution, though: this is a mitigation for accidents and non-deterministic tools, not a security boundary for running genuinely untrusted workloads. Kapsule containers share the host's Wayland, PulseAudio, and PipeWire sockets—that's a lot of attack surface if you're worried about malicious code. For truly untrusted workloads, VMs would be a much better fit, and Incus already supports those. It's not on my roadmap, but all the building blocks are there if someone wants to explore that direction.

New on the radar: exporting apps

Here's one I didn't see coming. aks asked about running a GUI application installed inside a container and having it show up on the host like a normal app. I realized... I didn't have that functionality at all.

The naive approach is straightforward enough: drop a .desktop file into ~/.local/share/applications that launches the app inside the container. But I really don't like that solution, and here's why:

• Stale entries. If the container gets deleted, or Kapsule itself gets uninstalled, those .desktop files just sit there like ghosts. You click them, nothing happens, and now you're debugging why your app launcher is showing dead entries.
• No ownership tracking. There's no built-in mechanism to say "this .desktop file belongs to this container, managed by Kapsule." If something goes wrong, there's no clean way to find and remove all the artifacts.

What I really want is a way to tie exported application entries to their owning container and to Kapsule itself. That way, when a container goes away, its exported apps go away too. Clean. Automatic. No orphans.

This might require changes to kbuildsycoca—the KDE system configuration cache builder—to support some kind of ownership or provenance metadata for .desktop files. I need to investigate whether that's feasible or if there's a better approach entirely. It's the kind of problem where the quick hack is obvious but the right solution requires some thought.

Taking a break (in theory)

I'm going to do my best to not touch Kapsule until the weekend. I have a day job that I've been somewhat neglecting in favor of hacking on this, and my employer probably expects me to, you know, do the thing they're paying me for.

We'll see how good my self-control is.