A desktop shell is easy to notice when it gets in the way. Canopia is an attempt to build one that earns its place on the screen.
There is a short interval after a computer starts when the desktop has the whole stage. The wallpaper appears. A panel settles along an edge. The clock catches up, network and sound controls report their state, and the first workspace waits for a window. A few seconds later, the desktop recedes behind a browser, a terminal, a game, or a document. Most of its work from that point onward is quiet.
That quiet work is Canopia's subject.
Canopia is a desktop shell for Linux, built around the Niri Wayland compositor. In plain terms, Niri decides how application windows are placed, moved, and displayed. Canopia supplies much of what surrounds those windows: panels, an application launcher, quick settings, notifications, workspace controls, wallpaper management, a lock screen, and the settings needed to make those parts belong to the person using them.
This division matters. Canopia is not trying to replace Niri, and it is not a decorative skin laid over it. The two projects solve different halves of the same daily experience. Niri manages space. Canopia makes that space legible and operable.
The project is written largely in Rust and uses GTK for its visible interfaces. It talks to Linux services for things such as audio, networking, power, and notifications, and it uses Wayland protocols to place shell surfaces correctly on the desktop. Those are useful facts for a contributor, but they are not the reason to care about Canopia. The real question is simpler: what kind of desktop does this machinery make possible?
A desktop you can inhabit
Many desktop projects begin with a screenshot. That is understandable. A desktop is unusually visual software, and a coherent image can express an idea faster than a page of architecture notes. The trouble starts when the screenshot becomes the whole idea.
A panel can look finished while its controls do nothing. A settings page can present a generous list of choices while only one path reaches the running shell. A carefully blurred surface can still overlap another surface, crop a calendar, hide the active workspace, or lose its contrast when the wallpaper changes. These are not small implementation details hiding beneath a successful design. They are the design.
Canopia has learned that lesson in public, sometimes comfortably and sometimes not. Its development history is full of interfaces that improved, exposed a deeper structural problem, and then had to be rebuilt. The main bar became a configurable panel system. The launcher grew beyond a grid of frequent applications into a broader place for applications, categories, files, and settings. Wallpaper-derived colors stopped being a visual experiment and became part of the theme model. Settings had to become promises backed by runtime behavior rather than an inventory of aspirations.
That process can look untidy from inside a repository. It also reveals what Canopia is trying to become: not a themed demo and not a collection of independent widgets, but a desktop that can be lived in every day.
Daily use is a demanding standard. It includes the pleasant five minutes spent arranging a panel, but it also includes the next five months when that panel should keep working. It includes a laptop waking on a different monitor layout, a network disappearing during a call, a notification arriving while the screen is locked, an application with an unexpected icon, and a user who has changed enough defaults that the developer's preferred layout no longer exists.
The desktop has to remain understandable through all of that.
Configuration as ownership
Canopia treats customization as more than a color picker.
A panel is not supposed to be a permanent object that the code calls "the bar." It is one instance in a panel system. It can have an edge, a thickness, a gap from the screen, an opacity, a shadow, a color source, and an ordered set of modules. It can be docked or floating. Its contents can be changed because different users need different answers from the same screen.
This approach owes something to the long tradition of configurable Linux desktops, especially Xfce. Xfce's panel has remained useful because it regards the panel as a framework for items rather than a fixed composition. Canopia takes that principle seriously while pursuing a different visual and spatial character.
Granularity, however, is not a synonym for quality. A settings screen with a hundred controls can make a desktop feel less personal if those controls are difficult to understand, poorly grouped, or detached from visible results. Good customization has a shape. Common decisions should be close at hand. Less common ones should remain discoverable without confronting every user at once. Related controls should use the same language in settings, configuration files, and the running shell.
This is why the move from a hardcoded bar to modular panels is not complete merely when a data structure accepts a list of modules. The running interface, settings application, saved configuration, and migration path must all agree on what a panel is. Adding a second implementation beside the old one would make the project more configurable on paper and less trustworthy in practice.
Canopia's standard is that exposed settings must be real. If a switch claims that surfaces can derive their colors from the wallpaper, changing the wallpaper must update every surface that participates in that feature. If a panel item can be reordered, the order must survive a restart. If a visual control has no effect on the shell, it is not a harmless placeholder. It is a false statement made by the interface.
That insistence gives customization a moral dimension that is easy to overlook. Software asks users to form expectations. Honest software takes responsibility for the expectations it creates.
Color from the place you chose
Wallpaper is often treated as the least consequential part of a desktop, a picture visible mainly between windows. In Canopia it can play a more active role.
When dynamic theming is enabled, the shell can derive a usable palette from the current wallpaper and apply it to its surfaces. The point is not to tint every object until the desktop becomes monochrome. It is to let the environment respond to a choice the user has already made, then preserve enough contrast and restraint for text, controls, and status to remain clear.
This sounds like a cosmetic feature. Done properly, it reaches through the entire design system. Selected rows, panel surfaces, popups, notifications, folder accents, focus states, and lock-screen controls all need to understand the same palette roles. A component that falls back to a toolkit's default blue does more than introduce the wrong color. It reveals that the component does not yet belong to the system.
Wallpaper-derived color also tests whether the theme is structural. If changing one image can make a label unreadable or turn a button into yellow text on a yellow background, the design was balanced around a sample palette rather than built on meaningful color relationships. Dynamic color forces the project to distinguish between accent, foreground, surface, border, destructive action, and selection. That discipline improves the fixed themes too.
Canopia's visual direction, known within the project as Obsidian, aims for depth without theatricality. Surfaces should have a clear order. Shadows should suggest separation instead of drawing dark rectangles around everything. Corners, borders, spacing, icons, and motion should feel related across the launcher, settings, notifications, calendar, quick settings, and lock screen.
The ambition is contemporary, but "modern" is a slippery word. A desktop does not become modern because it is dark, translucent, or generously rounded. Those fashions can be pleasant. They can also conceal poor information density and awkward composition. Canopia looks modern only when its visual choices help the user see what is active, what belongs together, what can be clicked, and where the available space went.
The shell is a relationship between surfaces
A desktop shell is made of parts that are often developed in isolation and used together.
Open quick settings over the panel. Open the calendar near a screen edge. Launch the full-screen drawer while the panel changes from floating to attached. Show a notification over a maximized application. Lock the screen while media is playing. Each individual surface may be correct and the combined result may still fail.
This is one reason Canopia relies heavily on screenshot-based development. A screenshot makes composition concrete. It shows whether a sidebar has consumed half the useful width, whether a search field dominates the application grid, whether a popup has become a long black strip, or whether text is clipped inside a control that looked fine in code.
Screenshots are evidence, not verdicts. They cannot prove that a setting persists, that keyboard navigation works, that drag and drop has the right semantics, or that an animation feels responsive. They can, however, stop a developer from declaring a visual defect solved when it remains plainly visible. For software as spatial as a desktop shell, that is valuable discipline.
The deeper goal is to test relationships. The panel should reserve or release space in agreement with the compositor. A full-screen shell view should account for the panel rather than covering it accidentally. A popup should be anchored to the control that opened it and sized for its content. The lock screen should use the same design language without exposing information that belongs only in an unlocked session. Motion should explain these changes, not delay them.
When the launcher opens and the panel moves from a floating shape to the screen edge, the transition can preserve orientation: this is still the same desktop, now in a different mode. When an application category expands inside the launcher, the navigation should rearrange rather than teleport into an unrelated menu. Micro-interactions matter most when they maintain a person's mental map.
Built beside Niri, not above it
Niri gives Canopia a distinctive foundation. Its scrollable tiling model arranges windows in columns along a horizontal strip. Opening a new window does not have to resize every existing window. Workspaces are dynamic, and each monitor can carry its own independent sequence.
Canopia should make this model easier to read without flattening it into the conventions of another desktop. Workspace indicators, window titles, overview controls, and panel actions need to reflect Niri's actual state. The shell cannot pretend that the compositor is a generic backdrop and still feel native to it.
At the same time, a user should not need to study compositor vocabulary before changing the volume or finding an application. The desktop must meet people with familiar controls where familiarity helps. Canopia's job is partly translation: preserve what makes Niri unusually capable, then present it through interfaces that explain themselves in use.
This balance is one of the project's most interesting problems. A shell too independent from Niri would feel pasted on. A shell that exposes every internal compositor concept directly would burden ordinary tasks with implementation detail. The right relationship is close but not possessive. Niri remains a complete compositor project with its own community and configuration. Canopia becomes a desktop around it.
What Canopia is not
Canopia is not finished, and pretending otherwise would contradict its own principles.
It has a broad surface: panels, launcher, files, notifications, quick settings, lock screen, wallpaper, theme generation, system services, configuration, and the links between all of them. Breadth creates risk. A feature can exist in three states at once: represented in settings, stored in configuration, and only partly reflected at runtime. A visual migration can leave old selectors or fallback paths alive. A refactor can create cleaner crate boundaries while moving confusion rather than removing it.
The project therefore needs a less glamorous kind of progress alongside new ideas. Dead paths must be removed. Large modules must acquire clear ownership boundaries. Services must report failure rather than invent success. Settings need end-to-end tests. Visual components need states for empty data, long labels, small screens, large text, alternate palettes, keyboard focus, and unavailable hardware.
That work is not a retreat from design. It is how a design becomes a product.
Nor is Canopia an argument that every desktop should work this way. Conventional floating desktops, mature integrated environments, and minimal compositor setups answer real needs. Linux is healthier because those choices exist. Canopia's value will come from developing a coherent answer of its own, not from declaring all the other answers obsolete.
The ordinary ambition
There is a grand version of Canopia's story in which it is the Linux desktop of the future. Grand stories are energizing, but desktops are judged in ordinary moments.
Can the user see all installed applications? Does the selected folder look selected without acquiring two borders? Can the calendar display every number without clipping it? Is the close button actually circular? Does a panel remain configurable after its layout becomes visually polished? Does the same theme reach the lock screen? Does a missing service produce an intelligible state? Can the user recover from a mistake without editing a file by hand?
These questions are almost comically small beside the phrase "desktop environment." Together, they are the environment.
Canopia is an effort to make those details agree: a shell shaped around Niri, a design system that extends beyond a color sheet, a configurable panel model that treats the user's arrangement as real, and a set of Linux services presented as one understandable place.
Its philosophy is not minimalism, although unnecessary noise should be removed. It is not maximal customization, although the desktop should be deeply configurable. It is closer to stewardship. The shell occupies a privileged layer between a person and nearly everything they do on the computer. It should use that position carefully.
When it succeeds, the reward will not always be admiration. Often it will be the moment after login when the panel settles, the first window opens, and the desktop quietly gets out of the way without ceasing to help.
