Add GatewayActivityCard: is this gateway actually hearing anything?

[ampledata]

Every pytak gateway plugin can now render the same activity panel — decode rate, trend sparkline, health, and a feed of recent contacts — reading the status file pytak.StatusWriter maintains at /run/<app>/status.json (snstac/pytak#107). Only the columns differ per gateway, because a tail number and an ACARS label matter for one and an MMSI matters for another.

<GatewayActivityCard
  appName="acarscot" title="ACARS Activity" noun="message"
  counterLabels={{ rx: 'received', emitted: 'plotted' }}
  columns={[
    { key: 'tail', label: 'Tail' },
    { key: 'flight', label: 'Flight' },
    { key: 'label', label: 'Label' },
    { key: 'freq', label: 'MHz' },
  ]}
/>

The hard part isn't reading the file

It's deciding whether what it says is still true.

Staleness is measured client-side, as time since we observed the file change. The obvious check — compare wall_t to Date.now() — is wrong: Cockpit runs in a browser that may be on a different machine from the gateway, so those are two different clocks. A laptop ten minutes off would show every healthy gateway as stale, or worse, a dead one as live. wall_t is still read, but only to detect and report that the clocks disagree — itself worth surfacing on a box meant to be GNSS-disciplined.

A timer drives re-render, because the interesting event is the absence of a change. A watcher alone would leave the last document on screen forever — precisely the failure this card exists to prevent.

Five states, because "no contacts" has several very different causes:

State	Meaning
live	file changing, contacts arriving
quiet	writing, but hearing nothing
stale	stopped writing — may be wedged
missing	no file at all — stopped, or pytak too old
degraded	running but reporting write_errors

A missing file says so explicitly rather than rendering as "working but quiet". And the banner always says when — "Receiving" with no timestamp is the claim this card exists to stop being taken on faith.

trendDirection masks pre-start buckets

Buckets older than the gateway's uptime_s are empty because the service wasn't running, not because it heard nothing. Counting them reported a dramatic rise after every restart.

A genuine gap — up for an hour, silent for twenty minutes, now hearing traffic — is still correctly reported as rising, since that's a real change an operator wants to see. Both cases are tested; the restart case was found by a test failing against my first implementation.

Sparkline

Inline SVG, no charting dependency — these plugins ship to a Pi over a field link. Baseline is always zero: scaling to [min..max] would make a flat line at 3 contacts/min look identical to one at 300. All-zero data draws a flat line on the floor rather than an empty box — silence is a real reading, and it's what a disconnected antenna looks like.

Also

Verified every --sns-*/--aos-* variable the new styles reference actually resolves. The first draft referenced four that don't exist (--sns-ink-console, --sns-muted-console, --sns-rule-color-console, --sns-font-narrow), which CSS would have failed silently.

29 tests pass; typecheck and eslint clean. Version → 1.3.0.

🤖 Generated with Claude Code

https://claude.ai/code/session_0197da7dhvcPoHxYKamrYqyM