Fences and traps
Two modes of the OpenCollar Edge with the fence port (phase 32, design docs/FENCE_AND_TRAP_PLAN.md, decisions D263 to D266): a FenceEdge listens on an electric fence wire and reports what it hears; a TrapEdge watches a switch on a trap door. Both devices stand still, so their place is set by hand on the device page (Bluetooth contacts describes that control), and both are ordinary devices on ordinary entities.
What the devices report
A FenceEdge sends a fence measurement every fence_interval seconds (port 12): whether the measurement worked, the number of pulses it counted in its sampling window, the average peak voltage in volts and an energy figure in the device's own units. Protect stores fence_voltage, fence_pulse_count and fence_energy as measurements and a fence_measurement state; a measurement that did not work raises a fence_measurement_failed event and no readings. The panels and the charts show the voltage in kV, which is how a fence person reads a fence; the stored value stays in volts.
A TrapEdge sends the switch's change at once (port 19) and its state every reporting interval (port 20). Protect stores switch_active and raises switch_activated and switch_deactivated events for any device that reports a switch.
A fence line
A fence voltage is measured at one point, but the question is about a stretch of fence. So a fence is a feature of its own, a fence line, drawn on the map or the Features page like a route. What makes it live is the Fence monitor entities on it: a FenceEdge is a device on such an entity, with a fixed place, and the entity's page has a Fence line card that puts it on a line. A line holds any number of monitors.
The line reads its status from them, section by section: it is cut where its monitors stand (each monitor's place projected onto the line), a section reads the worse of the monitors at its two ends, the ends of the line read the nearest monitor, and one monitor colours the whole line. A monitor's level comes from its newest reading:
| Level | When |
|---|---|
| live (green) | the voltage is at or above the line's ok threshold and pulses were counted |
| low (amber) | the voltage is under the ok threshold |
| down (red) | the voltage is under the down threshold, or no pulse was counted in the window: a wire without pulses is dead whatever the noise reads |
| unknown (grey) | nothing was measured, the newest measurement failed, or the newest reading is older than twice the fence interval: a silent monitor says nothing about the wire |
The thresholds and the interval belong to the line, in its attributes under fence (ok_v, down_v, interval_s), with 4 kV, 2 kV and 60 s when nothing is set. A line without a monitor is grey and says so.
On the live map a fence line is drawn in its sections' colours, under Features as its own type. Its panel shows the essentials and nothing more (Tim, 2026-09-19): the level and since when, the line as one bar with each section in its colour and a tick where each monitor stands, one sentence naming what is wrong and where ("2 of 4 sections down, Middle to South corner"), and how many monitors with the newest reading's age. The rest is on the fence line page, opened from the panel, the Features list (the bolt icon) or a monitor's entity page: the bar large with the monitors named, a table of the sections, every monitor with its reading and its voltage chart over a day, a week or a month, and the status history (the FENCE_STATUS events, newest first).
Project admins set a fence line up on the same page, under Setup: the name, the line itself (corrected on a map by dragging its vertices rather than drawn again; the monitors' places along the line and the sections follow at once), the thresholds, and the monitors. The monitors are placed on a map (Tim, 2026-09-19): the line is drawn in its sections' colours, the monitors on it are markers that drag along it, and a free Fence monitor from the list beside the map is dropped anywhere on the line. Where the pointer lets go is moved onto the line, and that point becomes the monitor's device's fixed place — over any place the device had, since a drop says the hardware moved. Nobody types coordinates, and a monitor is on the line rather than beside it. A monitor without a device cannot be placed, because the place is the device's; a new monitor is made beside the map with its name and its FenceEdge and then dragged on. A new line starts from the Features page like any other feature, and its page opens for the setup once it is made. The Features list shows the line's level.
Every time a section changes level the decoder raises a FENCE_STATUS event on the line: information when a stretch comes live, a warning when it reads low or unknown, critical when it is down. The event is placed in the middle of the stretch that changed and names it ("Fence East reads down near North gate to Corner"), so the feed and the automations see what happened where. The history of a line is these events and its monitors' voltage series.
Two shipped rule templates make the fence alert: Fence down (a fence monitor reads under 2 kV) and Fence monitor silent (no data for six hours; scope it to the Fence monitor type). Both create alerts and remind once a day.
A trap
A trap is a Trap entity with a TrapEdge on it. For a device on such an entity, every switch reading becomes a trap_triggered measurement, and a change against the newest one raises TRAP_CLOSED (a warning: somebody has to go and look) or TRAP_OPENED (information).
A switch is a switch: whether "active" means the door is closed depends on how the magnet and the reed contact were mounted. The device page has a Trap card, shown while the device is on a Trap entity, with the answer; the default is that an active switch means closed. A wrong answer reports every catch as a release, which is why the card says so.
The shipped rule template Trap closed alerts when trap_triggered is on, reminds once a day while it stays shut, and makes the trap read critical on the live map.
What it is not
A fence section's status is what its monitors last said, with the thresholds a person set; it is not a measurement of the whole wire. The energiser's own output is one more monitor, not a distinguished one. And a trap's count of activations (port 20) is stored as switch_count and not yet shown as a figure of the trap.