Skip to content

v3.5.0-beta.3

Pre-release
Pre-release

Choose a tag to compare

@btli btli released this 12 Jul 20:23
a91660f

Quick Charge local control + a large-bank canary fix.

Requires pylxpweb 0.9.38b4 (installed automatically).

Added

  • Quick Charge duration now works on all three connection paths (#366): a LOCAL/HYBRID start writes the register 233 activation together with the register 234 duration in one contiguous Modbus frame — the portal-equivalent sequence, live-validated on FlexBOSS21 hardware — so a locally-started charge runs for the requested minutes instead of the firmware default (a lone idle reg-234 write is firmware-rejected, #251). If the paired frame is rejected, the start falls back to the proven activation-bit-only write with a best-effort live duration write; the cloud start endpoint is used only when the local activation itself fails (HYBRID). Notes: a locally-started session is per-inverter, while portal starts remain parallel-group-wide; the EG4_OFFGRID (XP) family is unchanged — quick charge control there goes through the cloud endpoints (#296), so it requires cloud credentials and remains unavailable on pure-LOCAL XP installs.
  • Quick Charge Duration number: setting it while idle now stores the per-serial start preference (previously rejected with an error); while a charge runs it still adjusts the live countdown. The preference persists across restarts via a start_preference state attribute.

Changed

  • Quick Charge Duration idle display (LOCAL/HYBRID): while no charge is running the entity now shows the stored start preference (default 60) instead of mirroring holding register 234, which the firmware zeroes at session end. If an automation used Duration == 0 to detect "charge ended", key it on the Quick Charge switch state instead.

Fixed

  • Bank sensors went stale at solar noon on large battery banks (#367, reported by @Caymanwent): the bank-current corruption canary's flat 500 A cap rejected a 9-battery bank's genuine ~750 A charging current. The bound scales with the effective battery count — the larger of the reported count (register 96) and the batteries actually present in the read — at 150 A per battery, with a 500 A floor and a 2000 A physical ceiling (pylxpweb 0.9.38b4). Reporter confirmation at the next solar peak welcome on #367.
  • Write logs no longer claim "via CLOUD API" for transport-routed control methods.