panic — well pump health

Each alert has a state: open (currently active, triggers an email), acknowledged (you clicked Acknowledge — still tracked, but reminder emails stop), or cleared (condition is no longer present, triggers a one-line "resolved" email).

Over-draw warning

An irrigation zone is drawing water noticeably faster than its own historical normal. Possible causes: a broken sprinkler head, a blown fitting, a stuck-open valve, or a new leak in that zone's line.

Under-draw warning

A zone is drawing noticeably slower than its own historical normal. Possible causes: a clog, a partially failed valve, or a supply restriction.

Valve didn't close critical

After a zone's scheduled run ends, pressure should recover to normal within about 10 minutes. It didn't — and nothing else (another zone, in-home use, a hydrant, or a system-wide low-pressure trip) explains why. Most likely cause: that zone's valve is stuck open. A single occurrence is worth investigating on its own.

Switch drift warning

The pressure switch's cut-in/cut-out setpoints have shifted by 2+ psi from their established baseline, in either direction. Flags mechanical wear in the switch itself, and also matters because drift narrows the safe measurement window that the pump-capacity trend below depends on.

Pump capacity drop critical

An abrupt, statistically detected drop in how fast the pump refills the tank (see q_pump_gpm below), separate from the slow long-term trend. This is the earliest warning of a sudden mechanical problem such as a chipped impeller or a partial blockage.

Short-cycling critical

The pump is turning on and off much faster than a healthy tank should allow, during cycles when nothing was actually drawing water. Usually caused by a waterlogged pressure tank or a failing pressure switch — independently damaging to the pump from motor heating and start-current wear.

Unaccounted drawdown warning

Tank pressure dropped as if water were being used, but the Flo meter measured no flow, no irrigation zone was running, and no hydrant was logged open. Recurring instances suggest a leak somewhere that isn't visible to any meter.

Overnight pump starts warning

The pump refilled the tank overnight (roughly 1–5am), a time nothing should be drawing water. Recurring overnight fills suggest water is leaving the system with no accounted-for use — i.e. a leak. Culligan regeneration cycles are automatically excluded, since they legitimately draw water overnight.

Low salt warning

The Culligan softener's brine tank salt level has dropped below the refill threshold (default 20%). Add salt.

Low filter media life warning

The Culligan Calcite filter's remaining media capacity has dropped below its threshold (default 2,000 gallons). The filter media will need replacing soon.

Device error critical

The Culligan softener or filter is reporting an internal fault (e.g. a motor or position error) instead of its normal "System OK" state. Check the device directly.

Temp excursion critical

A freezer's Govee sensor has read above its configured ceiling (default 10°F) continuously for longer than its debounce window (default 60 min) — a brief spike from opening the lid does not trigger this. An active spoilage risk.

Sensor offline warning

A freezer's Govee sensor has been offline, or hasn't reported a reading, for longer than its configured threshold (default 45 min). This does not mean the freezer is warm — it means there's no way to currently confirm it isn't.

A safety interlock independent of the alerts above: if system pressure ever drops dangerously low, it stops irrigation immediately rather than waiting for a detector to confirm a pattern.

Trip trip

System pressure fell below the trip threshold (default 25 psi) — meaning the pump is on and still losing the fight, or something is seriously wrong. All Hydrawise irrigation zones are stopped immediately and an alert email is sent.

Clear clear

Pressure recovered back to the cut-in threshold (default 40 psi, matching the pump's own cut-in for hysteresis). Irrigation resumes automatically.

Action failed action failed

Low pressure was detected but the call to stop irrigation zones itself errored — the most serious failure mode, since zones may still be running.

Alert failed alert failed

An event was detected and handled correctly, but the notification email failed to send.

Disabled disabled

Guardian was disarmed from the dashboard (click the Armed/Disarmed label) — a break-glass override. While disarmed it will not stop irrigation or send alerts on low pressure.

Enabled enabled

Guardian was re-armed from the dashboard after a break-glass disable.

Status cards (top row)

The four cards at the top show the single latest sample: Pressure and Flow (the same two signals charted below), Zone running (which Hydrawise zones, if any, are active right now), and Ingestion errors (blue "none" when the last poll of both Flo and Hydrawise succeeded; the error text if either poll failed, which means the data below may be stale). The Culligan summary cards to their right come from the Water Treatment tab.

Pressure (psi)

Raw system pressure from the Flo meter. Rises while the pump fills the tank, falls during a drawdown (pump off, or a zone draining faster than the pump can fill).

Blue shading (pump running)

A shaded band behind the pressure trace marks when the well pump is inferred to be running. Rising pressure is always a certain "on" — only the pump adds water to the tank. A falling stretch is less obvious: if it's falling slower than the known draw (a running zone, domestic flow, an open hydrant, a water-treatment regen) alone would explain, the pump must still be running and adding water underneath it — this is the "deep recovery" case, when demand outpaces what the pump can deliver. A paler band means the call is low-confidence — usually an unmetered draw (hydrant or regen backwash) or a guardian trip is driving the balance, so treat that stretch as "not sure" rather than a confirmed reading.

Flow (gpm)

The Flo meter's instantaneous flow reading. It only sees the treatment/domestic branch, so it reads zero during irrigation — a nonzero reading during an irrigation cycle marks simultaneous in-home use, a confounding factor for some detectors.

Irrigation zones (swimlane)

One colored row per Hydrawise zone, showing exactly when each zone ran. Only zones active in the selected time range are shown.

Hydrants (swimlane)

Manually-logged hydrant openings (Well Pump tab). Hydrants share the irrigation branch but aren't metered by Flo or tracked by Hydrawise — a hydrant draw only ever shows up as a pressure dip unless it's logged here, so logging it prevents it from being mistaken for a leak.

Pump capacity card

Current pump output as a percent of its established baseline, plus the long-term trend in %/year. Shows "abrupt drop detected" if a pump-changepoint alert is open, and warns when there's under a year of history (seasonal effects can't yet be separated from real wear).

q_pump_gpm

The core pump-health metric: gallons per minute the pump delivers, estimated from how fast the tank refills across a fixed pressure band during a clean "no-draw top-off" (pump running, nothing else drawing water). Replaces the older, less comparable psi/second rate, since psi and gallons aren't linearly related.

q_pump_sigma (error bars)

The statistical uncertainty (±1 standard deviation) on the q_pump_gpm estimate, propagated from Flo's pressure-sensor noise.

Baseline → current (dashed line)

Connects the earliest-window median pump capacity to the most-recent-window median — the same two numbers shown on the pump capacity card.

Q_min floor (dashed brown line)

The minimum acceptable pump output for this site's demand — falling below it means the pump can no longer reliably keep up. Currently a placeholder value pending a proper site-specific calculation.

Changepoint marker

The date an abrupt capacity drop was statistically detected, if any (see "Pump capacity drop" alert above).

Switch card

Estimated cut-in/cut-out pressures (psi), and whether they're currently drifting from baseline (see "Switch drift" alert above).

Zones card

How many zones currently have an open fault alert, and how many have enough run history to be actively monitored ("armed").

Leak card

How many leak-related alerts (unaccounted drawdown or overnight pump starts) are currently open.

Per-zone diagnostics table

Each zone's historical baseline draw rate (gpm), most recent draw rate, how many samples its baseline is built from, and whether it's armed (enough history) or currently faulting.

Each row in the cycle table is one fill (pump running, pressure rising) or drawdown (pressure falling) segment.

Rate (psi/s)

The raw pressure slope. Not comparable cycle-to-cycle, since pressure and volume aren't linearly related across the fill band — kept for reference only, not a health metric.

40→60s

Seconds to recover from cut-in to cut-out pressure — an older, simpler fill-rate metric, superseded by q_pump_gpm as the primary health signal.

Q_pump (gpm)

The same volumetric pump-capacity estimate as the Analytics trend chart, shown per individual cycle.

Class (what else was happening during the cycle)

no draw topoff — nothing else was drawing water; the cleanest possible pump-capacity measurement.

domestic only — in-home (Flo-visible) water use, no irrigation.

domestic confounded — in-home use during a fill; capacity estimate less trustworthy.

single zone — exactly one irrigation zone running, no in-home use.

hydrant only — a manually-logged hydrant open, nothing else.

multi draw confounded — a hydrant open alongside a zone and/or in-home use.

ambiguous — more than one zone somehow active at once (shouldn't normally happen).

Flags

leak? — this cycle is a candidate "unaccounted drawdown" (see leak alert above).

inter-zone gap — a clean no-draw gap deliberately scheduled between two zones; doubles as bonus pump-capacity data.

hydrant — a hydrant was logged open during this cycle.

deep recovery — the fill started with pressure already below cut-in, i.e. demand was outpacing the pump when it kicked on.

System OK

Blue when the device reports no fault. There's no explicit "healthy" flag on the device — this is just the absence of any error.

Salt % (softener only)

How full the brine tank is, 0–100%. Not meaningful on the filter, which has no brine tank.

Media life (gal) (filter only)

Remaining filter media capacity, in gallons, before it needs replacing. Not meaningful on the softener.

Treated today (gal)

Running total of gallons treated by that device today (the device's own daily counter).

Daily average (gal)

Mean of the last 7 days' daily totals.

Last regen

When the device last ran a regeneration (salt/media recharge) cycle.

Clock drift

How far the device's internal clock has drifted from real time. Shown so a clock jump (e.g. after a device restart) is visible instead of silently corrupting timestamps.

Bypass

Water is currently being routed around the device rather than being treated.

Away mode

Vacation/away mode is enabled on the device.

Regeneration sessions (chart)

When each device regenerated, inferred from consecutive polls — Culligan doesn't expose a direct regen event log, so a session is either caught mid-regen or inferred from its lifetime regen counter incrementing between two polls.

Domestic consumption cross-check (chart)

Three views of the same water: Flo's instantaneous spot flow reading, Flo's real hourly totalizer meter, and each Culligan device's own treated-water delta. They sit in series on the same branch, so the lines should track each other and roughly track Flo's — a sustained divergence points to a meter fault or a leak between meter points.

One Govee H5111 sensor per freezer or refrigerator, polled every 15 minutes (thermal mass means a compressor failure takes hours to become dangerous, not seconds). Each card is one device — every device on the account gets a card automatically, no code change needed (adding a sensor does require an ingest redeploy to pick it up). A device's device_type (freezer or refrigerator, set in the `govee_config` Firestore doc) picks its default excursion ceiling.

Temp (°F)

Latest reading. Turns amber once it's at or above the device's configured ceiling (default 10°F for a freezer, 40°F for a refrigerator), and brown if that's been sustained long enough to trip a "Temp excursion" alert (see Alerts above) — a brief reading above the line from opening the door is not itself an alert.

Online / last seen

Whether the sensor is currently reachable, and how long ago it last reported. Turns red if it's been long enough to trip a "Sensor offline" alert — an offline sensor means the device's temperature is unconfirmed, not that it's necessarily warm.

Humidity / battery

Captured in the data model for a future device/API that exposes them, but the H5111's cloud API doesn't report either field for the current sensors — both always show "n/a".

Temperature chart

Every device's temperature over the range selected above, same chart style as Overview's pressure/flow chart. Each device gets a color: a thin line for the raw sensor reading and an optional bold line for the estimated food temp (below). Two shared dashed lines mark the excursion ceiling for each device type present — one for freezers, one for refrigerators — rather than one per device.

Est. food temp (bold line)

The raw sensor reads air temperature, which swings sharply every time the compressor cycles or the lid opens; the food itself has far more thermal mass and barely moves over those same minutes. The bold line estimates that actual food temperature by running the raw air-temp series through an exponential moving average (a low-pass filter): each reading updates the estimate by a fraction that depends on the real elapsed time since the previous sample, so an occasional missed poll doesn't distort it. It's a smoothed proxy, not a second sensor — the excursion alert still fires on the raw reading, not on this line. The "Est. food temp" dropdown above the chart sets the time constant (how much lag/damping the estimate assumes), or hides the line entirely.

Last excursion / Last offline period

The most recent episode of each alert for this device — "ongoing since <time>" if still open (or acknowledged), or the start–end range if it's since cleared. Only the single most recent episode is ever shown, not a full history: reopening a cleared condition starts a fresh one in the underlying alert record. See the matching entries under Alerts above for what each condition means, and the Active alerts table (Analytics tab) to acknowledge one.

A mow-timing advisor ported from the standalone mownow project (docs/ dev_planning/lawn-mow-advisor.md), fed by a WeatherFlow Tempest weather station rather than any of the site's own sensors. Every hour of the selected day is scored 0-100 for two separate tracks: sun (open lawn) and shade, since shaded grass holds dew/rain moisture much longer. Sun is the headline verdict throughout; shade rides alongside as a secondary line.

Verdict

One of four bands, told apart by word and glyph (not color alone): Great time to mow (filled check), OK to mow (open check), Still drying, not yet (hourglass), or Don't mow (barred circle) - the last covers both an active hard veto (rain, near-freezing temp, high wind, lightning) and grass that's simply still too wet to cut.

Best window chip

The longest run of hours scoring 50+ ("OK" or better) today, shown as a start-end range. Greyed "no good window" when no such run exists.

Day picker

Today through three days out, all pulled from the same already-fetched report - switching days just re-renders, it never refetches.

Hourly strip

One row per daylight hour; bar length is the score (0-100), color is a redundant tint from the same verdict palette. A vetoed hour shows "✗" plus the specific reason (e.g. "Rain expected") instead of a bar.

Conditions footnote

The Tempest station's current instrument reading, plus when it was last fetched - a mow report is only refreshed every 15 minutes, so this is how staleness would show up.