Changelog

The latest release is 0.5.0, from 2026-11-10. Download it.

Each release’s notes, generated from its commits, are on GitHub Releases. This page records the larger steps.

0.5.0 (2026-11-10)

• The World: a second way to see the Apps page, as your own Silicon Valley, drawn from your real data. A highway comes in from the side and becomes a curving main road, with a campus for each app in front of it, the services they rely on behind them on higher ground, and the watcher’s dish on a hill behind those. Adding an app never moves a service. Everything in it is a fact from the same data as the Apps page: a floor for each thing watched, fog for what isn’t known, frost for no recent data, smoke when an app is down, a crane during a deploy. There is no decoration: nothing is drawn that says nothing. Open it from the Apps page, See on map on an app’s page, or periskopos://world. It opens by coming down through cloud while the picture gets ready.
• Signs and lights: each building has a sign on its roof with the app’s or service’s logo, turning to face you, and a light on a short mast above it: green running, amber needs attention, red down, blue a service’s maintenance. It’s lit only while the state is known, and the screen behind the logo is tinted to match. A thin line between two buildings is a real dependency of that app on that service.
• The key, and how to read the map: What am I looking at? is a key grouped by Needs a look, Under way, Not known and Fine. What’s on the map now is lit up, and clicking a status takes you to the next thing that has it. The full explanation opens from ?.
• Since you last looked: opening the World says what went wrong while you were away, what recovered and which apps were added, each a click away. Only what’s on the record: deploys aren’t kept, so they aren’t counted.
• A valley that grows: a new app’s campus comes up floor by floor, one that’s watching more or getting more traffic visibly grows, and the valley floor widens at a new stage: small up to 3 apps, wider from 4, widest from 10. What changed while you were away plays when you come back.
• Everything on the road is real, and you can click it: a van is one check the watcher really made, in the colour of its result. A truck is a deploy under way, or stopped at the door when a push isn’t live. Cars are a share of an app’s real requests where its Cloudflare domain is connected, which also sets how wide its building is and how full the car park.
• The watcher: a radio dish on its own hill turns only while your apps are being checked. Clicking it opens a radar of everything on the map and the log of its latest checks.
• Figures: Running, Deploying (only while something is), Requests, Lowest uptime with the app it belongs to, and Last check run across the top, and each opens what it’s about. When all is well they step back to one quiet line. The title says “All systems normal” only when the services your apps rely on are fine too, here and on the Apps page.
• Moving round: drag, turn and zoom, smoothly and by how far you scroll, on a mouse or a trackpad. Tour goes round every campus, whatever most needs looking at first. The map’s one setting, its sky (by your clock or your theme), is behind a gear.
• What changed: an app that changes state sends out rings and its label pulses, and a Recent changes card says what happened and when.
• Replay: a slider shows how every app stood at any moment of the last 24 hours, from the incident record, greyed and with nothing on the road, so it can’t pass for now.
• Works by keyboard, and goes easy on the machine: arrow keys, zoom, turn and Tab all work, and the key reaches screen readers. It draws less when another window has the focus and stops when it can’t be seen. The 3D code loads only when the page is opened. On a machine that can’t keep up it lowers its own detail. The guided tour stops at the World button on the Apps page.
• Uptime strips: every live app on the Apps page shows its last 30 days, one small bar a day, in the list and on cards. Hover a day for its uptime and outages, click to open the app’s history. A day nothing was watching stays blank, not green.
• Apps list: every column now starts under its heading and keeps its place, a long name or message is cut short rather than pushing the page sideways, and the state shows once instead of twice. On a wide window the spare room goes to the uptime strip.
• Charts: Traffic and Search now share one chart: a line with a real axis and gridlines, a crosshair and a proper tooltip on hover, and a switch between measures (requests or visitors, clicks or impressions). The period still counting is dashed, so a part-filled hour doesn’t look like a drop, and days when 5% or more of requests failed are marked. Search also gets the table of numbers.
• Choosing your alerts: under Notify me, each kind of alert has its own switch. Pause alerts quiets all of them, or one app’s from its own page, for 1 hour, 4 hours or until 8 am, and when the pause ends one message says what you missed.
• Settings in full: six tabs (General, Appearance, Notifications, AI assistants, Data, About), one row per setting, with switches for what saves at once. New: your name, the interval for new apps (and for all of them), what closing the window does, the page it opens on, 12- or 24-hour times, how long history is kept, and restoring the defaults. periskopos://settings/... links reach every tab.
• Export and import a setup: Export my setup saves your apps, checks, accounts’ names and settings as a file, without keys or history. Import says what it would do before it does anything, only adds, and connects what was waiting once you’ve added the accounts.
• Report an issue: a bug button in the title bar copies the diagnostics, now including how the World has run on this computer, and opens the bug form. Nothing is sent by itself.
• Smaller things: dropdowns are drawn in the app’s own colours with a tick on the current choice. Every disclosure toggle is at least 24px tall. Long paths and figures no longer push pages sideways on a narrow window.
• Safer settings, import and alerts: an exported setup no longer carries where your alerts go (addresses, chat id) or a pause, only the settings an import takes back. An import is all or nothing, and a file over 6 MB is refused. A damaged row in the settings can’t stop the checks any more. “Checks stopped” is dropped during a pause instead of arriving days late. An assistant limited to some apps no longer sees the pauses of the others, and a link with a broken encoding no longer shows the “unreachable” page.

0.4.1 (2026-11-05)

• A failed Dependabot job isn’t failing CI: when Dependabot can’t make an update, the app no longer shows “CI failing”. Real CI failures and other jobs GitHub runs, like a Pages deployment, still count.

0.4.0 (2026-11-03)

• No recent data: an app that looked up when last checked, but whose check is now two rounds late, says so instead of staying green, on the Apps page, the app’s page, the tray and for AI assistants. “All systems normal” is only said on recent checks. The Checked column follows the check that says whether the app is up, not the newest check of any kind.
• When the checks themselves stop: Needs attention says when background checks are off or have stopped running, and checks that stop on their own send a notification, and another when they’re back.
• Checks that can’t read: a CI, deploy, payments or similar check that can’t read for over an hour shows in Needs attention, instead of leaving “All normal” standing over a problem nobody can see.
• Alerts that don’t arrive: a phone alert is retried, and one that couldn’t be delivered (or a desktop notification the system refused) shows in the bell until one gets through. Notifications switched off show there too, without adding to the bell’s count.
• What counts as up: an answer from another host is said to be that, and a 403 or 429 counts as “can’t tell”, not down. If it goes on for over an hour, Needs attention and a notification say so. This PC is only called offline when neither GitHub nor Cloudflare answers.
• Honest numbers: uptime runs from when the app was added and its site first checked, and leaves out stretches when the checks couldn’t tell. A figure resting on under a week says so (“100% of 3 h”). An outage that spanned a sleeping PC reports the time watched, and says part wasn’t.
• The tray and the Apps page: the tray is grey when an app’s state isn’t known, and says so when the server stops answering; the Apps headline does the same when it can’t load. “Accounts and providers fine” only appears when they were really read. A used-up free plan needs attention. Apps with no checks are counted apart from “All normal”.
• AI assistants get the same facts: what’s wrong with the watching itself, how much time uptime rests on, and what couldn’t be read.

0.3.8 (2026-11-02)

• Windows and Linux: the ☰ menu button sits after Reload, before Minimize.

0.3.7 (2026-11-02)

• Windows and Linux: the menu (File, Edit, View, Help) opens from a ☰ button in the window’s top bar, or a tap of Alt. The frameless window had no menu bar, so Help → Report an issue and File → Show log couldn’t be reached.