One 64×32 LED show: swarm-assembled, lit with sweeping sheen, and colored entirely by ScrollKit, captured from its pixel-accurate simulator.

ScrollKit

Most LED-matrix libraries get you a scrolling "Hello, World" and call it a day. I built ScrollKit for everything after that, whether you're making a clock for your desk or a product you'll ship to people who never open a code editor. It pushes over-the-air updates to boards already in the field, refreshes live data on a fault-tolerant loop that keeps scrolling when the network hiccups, onboards Wi-Fi with no secrets file to edit, and gives users a built-in web server to change settings from their phone, all while running transitions and effects that look deliberate. None of those features is the hard part. The hard part is running every one of them at once, on a microcontroller with a couple hundred kilobytes of RAM, without the display stuttering. And because none of that is fun to debug on real hardware, I made it run unchanged in a desktop simulator that matches the panel pixel for pixel and exports its own GIFs and videos, like the one above.

Built by Michael Czeiszperger

from scrollkit.app.minimal import MinimalLEDApp

app = MinimalLEDApp()
app.scroll_text("Hello, World!", color=(0, 255, 128))

Why ScrollKit

• One codebase, two targets. A platform-detecting display layer picks the real displayio hardware backend on CircuitPython and the simulator on desktop. Your application code never branches on platform.
• Board-agnostic. On CircuitPython it auto-detects the board (MatrixPortal S3 or Interstate 75 W) and falls back cleanly; adding a board is a small recipe. See Adding New Hardware.
• Async-first. A cooperative event loop keeps the display scrolling while data refreshes and the web server run as background tasks.
• Memory-aware. Built for the tight RAM budgets of embedded boards (the MatrixPortal S3, the RP2350-based Interstate 75 W): a lightweight import surface and graduated feature degradation when memory is low.
• Batteries included. Priority content queue, an effects/transitions engine, a configuration web UI, manifest-based OTA updates from GitHub, WiFi and HTTP helpers, and JSON settings persistence.

What you can build

ScrollKit is the engine behind DIY scrolling-LED projects: clocks, weather boards, crypto/stock tickers, status displays, and bigger apps like ThemeParkWaits (a live theme-park wait-time board). The library ships with graded demos so you can see each capability in isolation:

Demo	Shows
demos/easy/	Scrolling text, no network
demos/medium/	Live temperature from a public API, periodic refresh
demos/hard/	Web config, priority queue, effects, multiple data sources, OTA, chunked fetch

See them all running in the Demo Gallery, animated previews of every demo, recorded from the simulator.

Architecture at a glance

your app  ──▶  scrollkit.app.ScrollKitApp        (async lifecycle: display + data + web)
                 │
                 ├─ scrollkit.display   UnifiedDisplay ─▶ hardware (displayio) | simulator (pygame)
                 │                      DisplayQueue (priority + expiry), DisplayContent
                 ├─ scrollkit.effects   Transition (content swaps), particles, splash reveals
                 ├─ scrollkit.web       ScrollKitWebServer (config UI; adafruit_httpserver | async)
                 ├─ scrollkit.ota       OTAClient + UpdateManifest (GitHub-hosted)
                 ├─ scrollkit.network   WiFiManager, HttpClient
                 ├─ scrollkit.config    SettingsManager (JSON persistence)
                 └─ scrollkit.utils     color, error logging, timing

Head to Getting Started, then work through the tutorials from easy to hard.

How this was built

I wrote the first two shipping versions by hand in 2024, when all of this was still one application. Splitting it into a library and a separate app layer, then documenting the result, is the kind of project that dies quietly in a spare-time backlog. So I used Claude Code and spec-driven development to handle the refactoring and the first drafts, then went back through all of it in my own voice, with my own screenshots. Yes, AI has touched a lot of this code. It was also directed by an engineer who has shipped production software for a living, including time on one of Sun Microsystems' API teams. Both are true.