Wording only. DECISIONS/EV-04: a contradiction is a meter that answered and
fell short (an unknown reading is none), and it grades no alarm while the
guard's input keeps the figure. ctl_car.car_cycle: the car-over row keeps the
figure only with the cover off.
Co-Authored-By: Claude <noreply@anthropic.com>
- [Critical] The latch set `w = None` before `_car_quarter`, so with cover
off the guard's input (`car_exclude`) counted the car as house and the
batteries discharged. The latch now changes only the figure graded for the
alarm; the car's energy count, the guard's input and the car-over row keep
the unlatched `car_alarm_w()`. Cover-off N4 twins (energy gap, T9(c) liar,
SoC 60 % and at the reserve) give the same commands as without the latch.
- [Major] `corroborated()` is False for an unknown energy reading, so one
missing reading latched a correct figure for the quarter. New
`CarTracker.contradicted()`: only a meter that answered over a whole window
and fell short. R1 (energy entity unknown 1 s at E+200, T2 and T3 shapes):
red for that cycle only, explained from the next.
- Minors: unit checks that a contradiction in our own pause latches nothing
(MF) and that the latch ends with the quarter (MG).
Co-Authored-By: Claude <noreply@anthropic.com>
- [Critical] After a restart or a gap only the rest of the quarter was graded
without the car (`_carq["whole"]`), which dropped the house's own energy
from before it and hid an honest alarm (review repro: house 18 kW then
12 kW, restart at E+310). The rest-only grading is gone: the car is
credited only with energy counted since the start, the rest of what is
done counts as house (base's side). T17b covers it; T17 T2 at E+300 now
alarms from the restart (the conservative side).
- [Major] Owner 2026-11-19: a figure the energy meter contradicted (fresh,
charging, not corroborated) stays untrusted to the quarter's end, also as
the EV-15 stand-in. T9(c): one alarm edge, red to the end; T9(c) is back
in the N5 check.
- [Major] T2c lagging-meter variant (readings 389 W then 800 W, the meter at
800 W): the house latch grades on the lower reading.
- Unit stub: a run that starts mid-quarter has the car counted since the
quarter's start (no restart).
Co-Authored-By: Claude <noreply@anthropic.com>
EV-15 rig F-A: with both batteries at the peak reserve and the car charging,
PEAK_GUARD_FLOOR_LIMITED and goodwe_peak_soc_floor fired at a quarter's start
and stayed red for the quarter, the excess smaller than the car. Two causes,
both held by the per-quarter latch `_car_need_excl`:
- C2: "without the car" assumed the batteries give 0 W from now on, so what
the loop still gives below the reserve counted as house (T2 red on base);
- C1: one cycle with no trusted car figure latched it for the quarter (T1).
Owner 2026-11-19 (Q1-Q3): without the car the batteries keep giving what they
give now (lower of two readings, every cycle); a house need seen with a
trusted car figure stays latched to the quarter's end (`_car_need_house`); an
unknown car figure keeps the alarm only while it is unknown. A car known to
draw nothing (fresh reading, not charging) still grades the house's own need.
After a restart or a gap only the rest of the quarter is graded without the
car. A 0 W figure explains nothing outside our own pause (EV-15). The guard's
claim, floor, setpoints and charger writes are unchanged (N4: 41 runs, same
command digests as base).
test_ev18.py: the test plan's T1-T18, T2b, T2c, N2-N5 (tools/ev18_peak.py:
the real Controller across one quarter edge, N=2 at the reserve).
Co-Authored-By: Claude <noreply@anthropic.com>
The writer runs beside the cycle and the car can stop before the write
returns, so no sequence existed yet and the latch set. Window now opens at
send (zero_sent_at), cleared on return; ends on another writer in the tail too.
Co-Authored-By: Claude <noreply@anthropic.com>
While the add-on holds the car at 0 A for its own phase change, the car's
figure had no corroboration (not charging), so nothing was explained and the
peak-floor alarm fired once. For grading only, the car's fresh, not stuck
figure now stands in for that sequence (bounded: its own waits plus one
corroboration window), ending at once on Off, Stop, shadow, unplug, another
writer or a refused write. car_exclude, the claim, floor and writes unchanged.
Co-Authored-By: Claude <noreply@anthropic.com>
Minor bump: car Manual mode (EV-04), car-alone peak alarm fixes
(EV-10, EV-11, EV-12, EV-10b). Two new options with defaults
(ev_batt_cover, ev_mode_start); existing installs need no option change.
Co-Authored-By: Claude <noreply@anthropic.com>
Two checks a mutant survived: crediting the car-made export in the done part,
and taking off the car's quarter average instead of its power now.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
The period without the car was clipped at 0 W, so batteries covering the car
were charged to the house and the floor alarm fired (rig F-1). New
car_exclude_alarm keeps the sign of the rest of the period; car_exclude (the
cover-off guard input) is unchanged.
Co-Authored-By: Claude <noreply@anthropic.com>
EV-10: the car is the cause when the period without it (projected minus the car's
fresh, corroborated average) is at or under the aim, also with the projection
still under it. EV-11: car_over stays on until the test has been false for 180 s.
EV-12: the tracker's own-write mute opens before the write and runs 15 s from its
return.
Co-Authored-By: Claude <noreply@anthropic.com>
N1 the undo gives the old current back (clamped) in shadow mode; DECISIONS records it as the one write shadow mode allows.
N2 the corroboration looks back to a stretch with at least 40 Wh expected (60 s at 2.4 kW and up, 104 s at 1.38 kW).
L4 the current comes back 10 s after the phase select returned, also when it was refused.
Co-Authored-By: Claude <noreply@anthropic.com>
M1 the writer ignores a reading taken before its latest confirmed write.
M2 a refused phase select gives the current back and stops after 3 tries (amber row).
M3 Off, shadow mode or a control/safety stop before the select undoes the change.
M4 the restore on shutdown (and the undo) is clamped.
M5 the car's power reaches the alarm path only while the charger's energy meter agrees with it.
L1 fuzz other-writer branch reachable and shown to run. L2 one wording. L3 Batteries-help-off text.
Co-Authored-By: Claude <noreply@anthropic.com>
Off and Manual (runtime state, Home and Home Assistant): the current (6 A to
the clamp) and the phases (1 or 3) are written once per user change through
the safe phase sequence, read back within 10 s, clamped to
min(ev_max_a, station limit, entity range). Shadow mode gates the writes.
The car's fresh power goes to the peak guard's alarm path: no false "Load too
high" or "Battery stopped limiting the peak", one grey notice per session.
ev_batt_cover and ev_mode_start options, peak line "set by the car", HA
entities, sim writes, DOCS, CHANGELOG, test_ev04.py.
Co-Authored-By: Claude <noreply@anthropic.com>
- MAINT-28: say the hold is for two inverters (the code gates it on N >= 2).
- EV-03: "no longer buy energy for the car" (the profile is a 7-day median).
- TEL-19: plain wording for the PV sensor point.
- Known issues: add TEL-20 (DOCS names a car sensor id HA does not create).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Bump version to 0.23.0 and fold "Not in a release yet" into a new
0.23.0 CHANGELOG section (MAINT-28, CI-17, EV-03, CI-18). Four new
options (ev_*), all with defaults; no option change needed.
Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Bdqr7PiaHcBnzPituu2yTC
The selector reads the graph, and run-affected.sh runs the tests, from the
working tree, so a --base/--head range is only right with the head checked
out. The CI image is built from the merged tree, so an ENV CI_ONLY in its
Dockerfile would make the gate partial; test_affected_tests.py now fails on it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A test first reached as text (a pattern reader) was never expanded again when a
later path reached it by import, so a test importing it was missed. Expanded and
reached are now tracked apart; the empty diff selects the safety set. Test added.
Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Bdqr7PiaHcBnzPituu2yTC
- .gitea/ci/affected_tests.py: which test_*.py a diff reaches (static import
graph plus names in strings, transitive), a fixed safety set, SELECT ALL for
anything it cannot map. .gitea/ci/run-affected.sh runs the selection in the
CI image. run-tests.sh takes CI_ONLY for that; checks.yml never sets it and
test_affected_tests.py fails if it does. The merge gate still runs every file.
- test_modes_p4 -> p4/p4b/p4c/p4d, test_safety126 -> +b, test_safety125 -> +b:
same check lines, none dropped (146, 35, 45 before and after).
- run-tests.sh: 10 workers instead of 6, new longest-first order.
Co-Authored-By: Claude <noreply@anthropic.com>
Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Bdqr7PiaHcBnzPituu2yTC
POWER_UNITS is now derived from _W_PER, so the start-up probe and the
reading cannot drift apart. No behaviour change.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- every HA power unit is accepted and converted (mW, W, kW, MW), in the probe
and in the reading
- a power entity that answers in another unit, with no unit, or not as a
number makes the reading unavailable: after 30 s it is "Charger not
reachable" (amber row, WARNING, notification after 5 min), with a Why that
names the cause (EN and NL)
Co-Authored-By: Claude <noreply@anthropic.com>
A charger with no usable power entity is refused as "not set up", like a
missing setpoint, read-back or state entity, so House and the house profile
behave as with no charger. An entity that is merely unavailable at probe time
is still the unreachable path. NL: help text "met rust", chip "Laadpaalstoring".
Co-Authored-By: Claude <noreply@anthropic.com>
- CarTracker.car_w returns None when the current reading carries no power
(power entity unavailable or in a non-power unit); it can no longer raise
into the control cycle, /status or the MQTT values.
- mode3 E and F are the card state "fault" (EN and NL).
- chargersim gains fault().
Co-Authored-By: Claude <noreply@anthropic.com>
test_pv.py part 6 says only the PV reader's named modules may mention its
material, so chargers.py no longer imports pv.py.
Co-Authored-By: Claude <noreply@anthropic.com>
Read-only: nothing in this change writes to the charger.
- app/chargers.py: ChargerDriver, AlfenHaDriver, UnavailableCharger, ChargerReading, CarTracker
- options ev_charger/ev_current_entity/ev_max_a/ev_phases (default off), Settings group, validator, census
- Car card on Home, unreachable row and notification, two-writer info row, MQTT sensors
- car out of the House tile and the house profile
- tools/chargersim.py charger model, test_ev03.py
Co-Authored-By: Claude <noreply@anthropic.com>
The two fixtures ran their cycles at _SRC_NOW but set rejected_since from the
real clock, so the cycle saw inverter-b live while live_units() saw it stale.
Both now use _SRC_NOW, and a check asserts both clocks agree. run_cycle's
comment no longer claims ctl_cycle.py:49 is the only clock read.
Co-Authored-By: Claude <noreply@anthropic.com>
c_src's telegrams were laid up to one datetime.now(); each cycle() then read
the real clock again (ctl_cycle.py:49). Under --cpus 0.25 a cycle takes ~1.3 s,
so by the 23rd sweep cycle the last telegram was older than max_hold_s (30 s),
the guard counted the gap and filed peak@site. The cycles now run AT the
fixture's own instant. Test only; the exact-set checks are unchanged.
Co-Authored-By: Claude <noreply@anthropic.com>
H-1: one site limit while the pause holds (the period's largest floor, never under this cycle's), so the other pack no longer gives for the house. New live-try test (A 11 %, B 25 %, 90 min at 6.5 kW) red on 07d23d2. M-1: a 45-min rig-shape test that goes red when the settled check is forced True. M-2/L-2: DECISIONS. CHANGELOG line. test_safety84 n2_charge_banked re-pinned (quarters 3349 and 3950 W, aim 3950 W).
Co-Authored-By: Claude <noreply@anthropic.com>
The same diff as f3f9483..07d23d2 (cap, settled rule, tests, DECISIONS). The merge took the release/1.0 tree, so the rounds are applied as one patch on top of the restored #247.
Co-Authored-By: Claude <noreply@anthropic.com>
Restores #247's code on this branch after the release/1.0 merge. The fix rounds follow in the next commit.
Co-Authored-By: Claude <noreply@anthropic.com>
- SAFETY-136 row: the gain is at 10 s late; at 15 s a rig test sent more.
- SAFETY-111: the cycle needs more than 4 nights; the 36 h limit ends it.
- MAINT-28: rig counts were 393-468 flips per 45 min, not one figure.
- MAINT-26 / MAINT-25: the next try is the first start hour after the wait.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Takes PR #247 (f3f9483) out of release/1.0 so the 19 Nov release ships
MAINT-24 + MAINT-25 without MAINT-28. The rig failed f3f9483
(reports/MAINT-28-test-report.md) and the fix-forward PR #248 review found
a High regression (reports/MAINT-28-fix-review-report.md). MAINT-28
continues on PR #248 after the release.
Co-Authored-By: Claude <noreply@anthropic.com>