39 releases 39 tags

RSS feed

• v0.2.0-rc.12 d98bda885a

  Compare

  v0.2.0-rc.12 Pre-release

  Cody Bryant released this 2026-08-12 16:36:20 -07:00 | 132 commits to main since this release

  Thanks to @CryingPecan, whose LitterHopper robot
  on ESP 1.4.4 is behind much of what's below. Protocol detail lives in
  docs/devices/litter-robot-4/.

  Added

  • The clean cycle and reset buttons are back, and this time they work: they
    synthesise the same button press the panel sends. Proven on ESP 1.1.75 and 1.4.4.
  • Empty cycle and Power buttons, disabled by default and named (danger) — an
    empty cycle dumps the globe into the drawer, and Power takes the robot off the
    network. The CLI gains empty-cycle and power, which prompt first.
  • Pet weight actually works. Recent rc builds doubled the reading; weights now
    match the household scale (raw ÷ 100, the cloud's own units).
  • LitterHopper support: connected, fill gauge, and an out-of-litter alert the
    firmware itself never raises. The hopper entities switch themselves on the first
    time a hopper reports.
  • Litter level as a percentage. It calibrates itself over time; one button press
    with the globe filled to the line pins it immediately.
  • New entities: last cat visit, last visit duration, waste drawer last moved,
    and panel brightness for bright and dark rooms.
  • Activity-derived entities survive a restart instead of reading unknown until
    the next cat visit.
  • New install channels: brew install sisyphusmd/tap/whiskerless, .deb and
    .rpm for amd64 and arm64, and standalone Linux binaries for both. None of them
    need a system Python — the audience is someone provisioning a robot from a laptop
    that has none.
  • Excess weight detection. The robot refuses to cycle while it thinks something
    is sitting on the scale, raises the condition itself after 30 minutes, and shows it
    on the panel — but says nothing about it over MQTT. One robot here sat like that for
    over two hours after a bonnet was reseated slightly off, with its clean cycle stuck
    the whole time and nothing on the dashboard to explain it. Now a sensor. Pressing
    Reset zeroes the scale and clears it.

  Fixed

  • The hopper entities no longer disappear from a robot that has one. An upgrade
    sweep retired hopper detections recorded from the link register, which is right —
    that register proves nothing. But it also cleared them on robots whose hopper was
    genuinely proven, and the replacement evidence is a dispense, which only happens
    when the litter is actually low. A well-fed robot could go weeks without one. The
    sweep now recognises a previously recorded fill gauge as the proof it is, since
    only a dispense can produce that number.
  • Handling the robot no longer shows up as a cat visit. A Reset press closes a
    visit on the same register a cat does; two of them were published as genuine
    four-minute and three-minute visits. A visit now needs something to have actually
    broken the beam, which a hand on the bonnet does not.
  • The globe motor fault sensor could sit at off through a real fault. It read
    the state document, and the state document does not carry the fault: a robot raised
    one on its activity stream, held it for fifty minutes and cleared it, while
    globeMotorFaultStatus reported no fault in every single state document it published
    in that window. The sensor now watches both channels, and either one raising a fault
    is a fault.
  • The LitterHopper is now detected by watching it deliver litter. The link
    register 0x57 looked like the answer, but a narrated session produced healthy
    readings from it with the hopper sitting on a bench, and its "disconnected" code
    from merely opening the hopper's drawer to refill it — so a refill could park the
    hopper sensor on disconnected with nothing on the wire to ever clear it. Nothing
    is derived from that register any more. Hopper reports connected once litter
    has actually been dispensed and never reports disconnected, because no signal for
    that exists.
  • A robot that dispenses but rarely reports its link no longer loses its hopper
    telemetry. Requiring 0x57 to corroborate a dispense discarded every fill
    reading on such a robot and left its four hopper entities disabled indefinitely.
  • The hopper level survives a restart. Dispensing only happens when the litter
    bed is actually low, so a well-fed robot can go days without one; the last gauge
    is now remembered instead of the level reading unknown until it next runs low.
  • Event sensors now appear only once their fact has actually been reported. Pet
    weight, last cat visit, and waste drawer last moved start hidden and switch on at
    their first real report — some firmware never emits the drawer event or a weight,
    and those sensors read unknown forever there. A one-time sweep on upgrade applies
    the same standard to existing installs: sensors whose values were real stay, hopper
    and visit-duration detections re-prove themselves at the next report (a robot with
    a real hopper re-enables within one visit), and phantom entities disappear.
  • Last cat visit now updates on every robot. It stamps from the occupancy signal
    itself, not only from weight events — one robot has visits but has never weighed
    anything, and its visit sensor stayed empty.
  • Cat detection no longer mistakes weight on the scale for a cat. The occupancy
    field is two bits — one for what the robot can see, one for what it can feel — and
    a bonnet reseated slightly off held the second bit for over two hours, which the
    robot itself reports as an "excess weight" fault. Occupancy and litter calibration
    now use the bit that tracks the animal.
  • Fewer unknowns while calibration settles. The litter calibration reference
    shows the built-in default (marked source: default) instead of unknown, and the
    hopper level shows a labelled estimate until the empty floor has actually been
    learned.
  • The weekday sleep schedule now arms every day. It is a per-day bitmask, not a
    switch, and turning it on armed Sunday alone — so it looked fine if you tested on
    a Sunday and did nothing all week.
  • The panel sleep schedule can actually be set. The sleep and wake time entities
    wrote a register the firmware only computes, so they never did anything.
  • Litter readings are suppressed while the globe is not level. Mid-cycle the
    sensors read the globe rather than the litter, and that was published as a level.
  • robotStatus 10 is the clean cycle on every firmware; the old map called it a
    cat pause. The boot cycle and the filter-change wizard are mapped too — both move
    the globe, and while unmapped their readings published as real litter levels.
  • whiskerless set night-light-mode auto works. Every spelling the command
    accepts used to crash — and it is the command in the README quickstart.
  • The declared Home Assistant minimum was wrong (2025.2.0), so a user on 2025.2
    could install this and watch it fail. It is 2025.3.0.
  • Every settings write is verified by reading it back, a multi-register write no
    longer loses one of its parts, and broker failures are reported rather than raised
    as a traceback.
  • The hopper no longer drops to unknown on a link code that is not a disconnect, and
    one dispense can no longer prove an empty hopper on its own.

  Changed

  • Breaking: binary_sensor.<robot>_waste_drawer_removed is replaced by
    sensor.<robot>_waste_drawer_last_moved. The robot reports that the drawer moved
    and never which way — nine codes turned up across removals and insertions alike.
  • Breaking: clean cycle wait time is a number (3–30 minutes), not a select. Move
    automations from select.<robot>_clean_cycle_wait_time to
    number.<robot>_clean_cycle_wait_time.
  • Breaking (library): Hazard.MOTOR, MotorCommandError and allow_motor are
    gone. A written press is the same event as a physical one, so the robot's own
    interlocks apply either way. Power still requires allow_dangerous.
  • Last visit duration is not reported by every robot, so it ships disabled and
    switches on the first time yours reports one, rather than reading unknown for the
    life of a robot that never will. This was thought to be a firmware split; it is not.
    Two robots on the same ESP build sit either side of it.

  Downloads

  • Source code (ZIP)
    0 downloads
  • Source code (TAR.GZ)
    0 downloads

  ---

  • whiskerless-0.2.0~rc.12.aarch64.rpm
    0 downloads · 2026-08-15 18:21:26 -07:00 · 26 MiB
  • whiskerless-0.2.0~rc.12.x86_64.rpm
    0 downloads · 2026-08-15 18:21:27 -07:00 · 26 MiB
  • whiskerless-linux-arm64
    0 downloads · 2026-08-15 18:21:29 -07:00 · 26 MiB
  • whiskerless-linux-x86_64
    0 downloads · 2026-08-15 18:21:29 -07:00 · 26 MiB
  • whiskerless-macos-arm64.pkg
    1 download · 2026-08-12 16:39:33 -07:00 · 11 MiB
  • whiskerless-macos-x86_64.pkg
    1 download · 2026-08-12 16:39:35 -07:00 · 12 MiB
  • whiskerless_0.2.0~rc.12_amd64.deb
    0 downloads · 2026-08-15 18:21:27 -07:00 · 26 MiB
  • whiskerless_0.2.0~rc.12_arm64.deb
    0 downloads · 2026-08-15 18:21:28 -07:00 · 26 MiB