In summer the WW boost interferes with car charging priority.
A new ⛔/✅ toggle on the WW consumer card persists to
/etc/ems/ww-boost-off (presence = disabled, absence = enabled).
- config: ww_boost_disable_file flag file path
- main: wwBoostDisabled mode, wwBoostToggleHandler, runCycle skips
computeWWBoost when disabled
- status: WWBoostDisabled in pageData, toggle button in WW card
showing ⛔ Boost aus / ✅ Boost ein depending on state
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Compressor:
- metric changed to PromQL expression that returns actual watts:
sensors_power_value(%) × power_value(kW) × 10
e.g. 27% × 10 kW = 2700 W; idle = 0 W
- idle threshold raised 20→500 W (clean gap: 0W idle vs ≥2700W running)
- status page now displays real kW instead of percent
WW boost reset:
- Added 🌡️ Zurücksetzen button to WW consumer card (when configured)
- POST /ww/reset: restores WWBaseC via Viessmann API + blocks re-boost
via engine override until midnight
- IsWW field added to consumerView for template control
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
User enters car SOC + departure deadline; EMS computes when to start
charging and activates the wallbox at the right time (grid import OK
for trip mode). Charging rate is learned from completed PM sessions,
improving over time. Car profiles (Mini Cooper SE 50 kWh, BMW ix2 63
kWh) are config-driven and selectable per trip.
- internal/trip/trip.go: Goal/Session types, Manager with per-cycle
Tick() that accumulates energy, detects session completion, learns
avg charge rate from JSONL session log
- internal/config: CarProfile + Cars map, TripGoalFile, SessionLogFile
- internal/status: trip card (active goal) + trip form (set new goal),
CarOption list, formatDay/formatDur helpers
- main.go: tripMgr lifecycle, /trip + /trip/cancel HTTP handlers,
runCycle integration (ShouldStartNow → ApplyOverride, auto-clear)
- configs/ems-config.yaml: cars section, trip_goal_file, session_log_file
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
New internal/monitor package provides a thread-safe, file-backed toggle.
Flag file presence (/etc/ems/monitor-only) = monitor-only active, survives
reboots. Deletion = resume normal operation.
Behaviour when active:
- EMS continues polling Prometheus, running the decision engine, and
updating the status page every 2 minutes — full visibility maintained
- All actuator calls suppressed (Shelly switches + Viessmann WW writes)
- Suppressed actions logged as [MONITOR-ONLY] for audit trail
- Startup warns if flag file is already present
Web UI:
- Amber banner "⏸ Monitor-Only — Keine Schaltvorgänge" with inline
"▶ Automatik" resume button when active
- Small unobtrusive "⏸ Monitor-Only" button at page bottom when inactive
- POST /monitor endpoint toggles state and redirects back to status page
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Update apiBase: api.viessmann.com → api.viessmann-climatesolutions.com
(Viessmann renamed their IoT API host; old domain is NXDOMAIN)
- Extend --test-viessmann context to 120s
- Add 30s propagation wait before read-back in both Stage 3 and Restore:
cloud accepts writes immediately but gateway polls on its own schedule,
so GET returns the last confirmed device value with ~30s lag
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Adds GetDHWTemperature() to the Viessmann client (GET on the DHW feature
endpoint) and a --test-viessmann CLI flag that runs three stages without
starting the EMS control loop:
Stage 1: read current DHW setpoint (confirms auth + endpoint path)
Stage 2: write same value back, read-back to confirm round-trip
Stage 3: write +1°C, read-back to confirm, restore original
The +1°C delta is below the heat pump's 5°C hysteresis so the compressor
will not fire, but the change is visible in the Viessmann app for visual
confirmation before trusting the WW boost logic in production.
Usage: ./ems --config /etc/ems/ems-config.yaml --test-viessmann
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Override duration (web UI):
- Dropdown on Ein button: 30min / 1h / 2h / 4h
- Engine notified immediately via ApplyOverride() — no waiting for
next SyncHardwareState cycle
- Aus button always uses 1h lockout (keeps consumer off for 1h)
Hard-stop thresholds that cancel active overrides:
- SOC emergency brake: now also clears ManualOverride flag so EMS
resumes full control after the safety shutdown
- override_max_import_w (default 800W): if grid import exceeds this
while an override is active, override is cancelled immediately
(no hysteresis delay — protection is instant)
Config: ems.override_max_import_w (0 = disabled)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Status page:
- Per-phase grid power (L1/L2/L3) with export/import direction
- Shelly PM live power badge on active wallbox rows
- Compressor power shown below consumer list
- SyncDeviceStates() method on Store for PM power readback
Engine:
- Fix wallbox mutex: re-read Active state after P3 so same-cycle
WallboxA activation correctly blocks WallboxB in P4
- Fix RecoverState: set ActivatedAt to zero so recovered consumers
are immediately eligible for shutdown decisions
- Add 3 new unit tests: mutual exclusion, car-not-charging,
compressor-idle early release
Makefile:
- make install no longer overwrites /etc/ems/ems-config.yaml if it
already exists (prevents password loss on updates)
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
