Added
- New action
foxess_control.backup: Backup mode for a set time (PR #29, thanks @giggles-38a). Solar still charges the battery, but the battery does not discharge to the house, so the house runs on solar and the grid. Useful for free or cheap grid-power windows (issue #32). It takes aduration, an optionalstart_timeandreplace_conflicts, likefeedin, and the inverter returns to self-use when the window ends. - Schedules can be written to inverters that only accept the v1 API, such as the EVO 10-5H (issue #28). These inverters refused every schedule change with "FoxESS API error 40257", so no action or smart session worked. By default the integration still writes via v0, the API every other reported model uses. If the inverter refuses that with 40257, it retries once via v1 and keeps using v1 from then on. A new Schedule write API option (Auto / v0 / v1) lets you pin a version: choose v1 for an EVO to skip the failed v0 attempt after each restart. The diagnostics download shows the version in use. Thanks to @prgarnett for testing both endpoints and posting the requests that worked.
Fixed
force_chargenow charges at full power for its whole window (issue #32). It runs as a full-power smart session, but the "ahead of schedule, return to self-use" check still applied to it. So whenever the pacing estimate found spare time, a forced charge went back to self-use and later restarted at a reduced rate. With Battery Capacity unset, whichforce_chargeallows, it looked as if no energy was needed, so it stayed in self-use for the rest of the window. On a cloudy day this left the battery well short. Forced charges no longer defer or re-defer, and always start at maximum power.- The charge card no longer contradicts itself (issue #32). The "target reachable" flag, the slack figures, the time-remaining estimate and the forecast now use the same inputs as the charging logic: the battery capacity the session started with, house load and the configured headroom. Before, they read the capacity from the options and ignored house load, so the card could call the target unreachable while the session was still waiting. The waiting message also no longer suggests a solar forecast. It now says charging at full power from the stated time still reaches the target.
- No more "WS-DIAG" warnings in the log. Temporary diagnostics added in 1.0.22-beta.8 for the WebSocket idle-leak investigation logged a warning on every 5-minute poll, and a stack trace every time the WebSocket started or stopped. That leak was fixed in 1.0.21, and the diagnostics are removed.
- Frozen cloud data now shows as stale (issue #41). The FoxESS Cloud sometimes keeps answering the 5-minute poll with the same cached sample for 30-55 minutes. One user saw this whenever Additional solar power variable was set to
meterPower2. The integration presented each repeat as fresh data. It now compares the sample time the cloud returns: while the same sample keeps coming back, Data Freshness keeps the time it was first received, so the cards show its real age and turn stale after 15 minutes. A repeat longer than that is recorded in Recent Errors. The diagnostics download has a newrest_freshnesssection listing recent repeats, their duration, and whether the additional PV variable was in the request.
Changed
- Full power now uses the inverter's own declared maximum, even when it's higher than the old estimate (issue #34). The maximum charge/discharge power was the inverter's
capacity× 1050, lowered to the maximum the inverter declares if that was smaller. The FoxESS API reportscapacityas a whole number, so a KH9.9 reported9and was capped at 9450 W, while it declares, and the FoxESS app allows, about 10400 W. The declared maximum is now used whenever the inverter provides one;capacity× 1050 is only the fallback. - A Backup group made in the FoxESS app no longer blocks every action (D-067). Until now, any Backup group in the schedule made the integration refuse all schedule changes until you removed it in the app. Backup is now handled like the other modes the integration manages. A Backup window that doesn't overlap is kept as it is. One that overlaps is reported as a conflict unless you set
replace_conflicts. Callingbackupreplaces it, andclear_overridesremoves it. Nothing removes a Backup group unless you ask from Home Assistant. Modes the integration doesn't manage at all, such asPeakShaving, still make it refuse and name the mode.