The first ntfy cycle announced "Warning level at P.77" at 3.02 m. That
gauge's 2.85 m threshold sat below its own dry-season baseline (2.6-2.7 m
at 8-14 % channel capacity): P.77 had been "above warning" for 761 of the
last 2 146 hours, at 22 % capacity. Across 2018-2024, 75-85 % capacity
reads 3.35-4.57 m and 95-105 % reads 4.27-5.08 m; set 4.30 / 4.90. P.75
moved 2.75/3.50 -> 3.20/3.65 on the same evidence (2024: 3.45 / 3.72).
The predictor already handles changed thresholds (regression-derived
probabilities until the Oct 1 retrain).
Second line of defence in notify.py: a clear->alert transition is only
announced when RID's discharge_percent for the reading is >= 60 %, so a
re-rated or datum-shifted gauge cannot page subscribers again. P.1 is
exempt (its stages come from the inundation map, not capacity); readings
without a capacity figure fall back to level only; the all-clear edge is
never blocked. 3 tests.
Anyone can now get push alerts on their phone without an account: the
monitor publishes to an ntfy server (one Go binary, ~30 MB RSS) and
subscribers pick topics in the free iOS/Android/web app.
Semantics are transitions, never state. One message when a gauge crosses
its warning or danger threshold, one all-clear when it drops back (0.10 m
hysteresis), nothing while it sits above. A three-day flood is two
messages; a quiet season is zero. Topics: ping-warning / ping-danger
(basin digest), ping-<station>-warning / -danger, ping-p1-outlook (opt-in:
model P(warning within 24 h) at P.1 rises through 50 %, clears below 25 %,
message says it is experimental), ping-status (feed stale >= 3 h /
recovered). Priority 5 on danger so it rings through Do Not Disturb.
src/notify.py runs once per collection cycle in the API process (leader
only, after the forecast precompute, same data the dashboard shows).
Last-sent state lives in a notification_state table so a restart never
re-sends; a failed publish leaves state untouched so the crossing is
retried next cycle instead of lost. Off unless NTFY_SERVER is set.
Dashboard: a "Get alerts" button (only when configured) opens a panel
with the server, per-topic cards, ntfy:// deep links and web links, app
store links and a disclaimer. EN + TH. GET /api/notifications feeds it.
scripts/install_ntfy.sh: .deb install, server.yml (loopback listen,
anonymous read, token-only write scoped to ping-*, 72 h cache, signup/
login/metrics off, tight visitor limits), systemd, user + token, .env.
Verified against ntfy 2.28.0: anon publish 403, token publish 200, token
on foreign topic 403, anon read 200, and a seeded crossing through the
real _notify_transitions path arrived in the topic with priority, tags,
click and action button. docs/NOTIFICATIONS.md has the deployment and
reverse-proxy notes. Tests: 10 for the state machine (159 total).
grabowski