fix(energysite): fill live_status energy_left/total_pack_energy from get_status()

[Bre77]

Intent

Wire aiopowerwall's newly added Fleet-API-aligned client methods (scaled SoC readers, get_battery_soe/get_battery_soe_raw, battery_level/battery_level_raw, set_operation_mode) into the PowerwallEnergySite router compatibility adapter so it's interface-complete for the tesla-fleet-api EnergyRouter. Studied the sibling python-tesla-fleet-api repo (tag v1.5.2) - tesla_fleet_api/tesla/router.py's Router (hasattr-based per-command dispatch/failover) and tesla_fleet_api/tesla/energysite.py's EnergySite/TeslemetryEnergySite classes - to enumerate the full router-invocable surface. Found the adapter already had 100% method-name parity (every command either mapped to a PowerwallClient call or scaffolded as a documented NotImplementedError placeholder consistent with existing conventions). The actual gap was inside the already-implemented live_status(): it left energy_left and total_pack_energy as documented None gaps even though that data is locally derivable - control.systemStatus.nominalEnergyRemainingWh/nominalFullPackEnergyWh are exactly the fields the new battery_level_raw()/battery_level() helpers already read. Fixed by replacing the separate get_battery_soe() call with a single get_status() call plus the new battery_level() helper, so percentage_charged, energy_left, and total_pack_energy all come from one read (same call count as before, two more Fleet API fields populated instead of None). No behavior for any other command changed; the NotImplementedError placeholder convention for genuinely unsupported commands (storm_mode, grid_import_export, time_of_use_settings, history reads, gRPC device commands, etc.) and the deliberate omission of site_info were left untouched since they're still correct. Added/extended tests in tests/test_energysite.py for the new energy_left/total_pack_energy values plus a missing-status-yields-None case, and updated README.md and AGENTS.md to describe the new live_status data source.

What Changed

• PowerwallEnergySite.live_status() now populates energy_left and total_pack_energy (previously always None) by reading control.systemStatus.nominalEnergyRemainingWh / nominalFullPackEnergyWh via a new _system_status_field helper.
• Replaced the separate get_battery_soe() call with a single get_status() read that feeds percentage_charged (via the client battery_level helper), energy_left, and total_pack_energy from one gateway query — same call count, two more Fleet-API fields filled. Remaining non-derivable keys (backup_capable, grid_services_*, storm_mode_active, timestamp, wall_connectors) stay None.
• Extended tests/test_energysite.py with assertions for the new Wh values and a missing-status-yields-None case (40 passing), and updated README.md / AGENTS.md to document the live_status data source.

Risk Assessment

✅ Low: A small, well-bounded change that populates two previously-None Fleet API fields from an existing single read, with matching test coverage and no behavior change to any other command.

Testing

Baseline uv sync --extra dev plus uv run pytest tests/test_energysite.py passed all 40 tests, covering the new populated-values and missing-status-yields-None cases. To show the intent working the way the router experiences it, I ran a small script that feeds realistic raw gateway payloads (nominalEnergyRemainingWh/nominalFullPackEnergyWh, meter aggregates, grid status) through the real PowerwallEnergySite.live_status() and captured the resulting Fleet-API-shaped response: energy_left=52777.8 and total_pack_energy=100000.0 are now populated (previously hard-coded None) and percentage_charged=50.2924 comes from the same single get_status() read via battery_level(). This is a library/adapter change with no UI surface, so the evidence is a CLI transcript of the actual adapter output rather than a screenshot.

Evidence: live_status() end-to-end output showing populated energy_left/total_pack_energy

live_status().response (fields relevant to this change): { "percentage_charged": 50.2924, "energy_left": 52777.8, "total_pack_energy": 100000.0, "solar_power": 3200.0, "battery_power": -1800.0, "load_power": 1400.0, "grid_power": 1500.0, "grid_status": "Active", "island_status": "on_grid" } Before this change, energy_left and total_pack_energy were hard-coded None.

live_status().response (fields relevant to this change):
{
  "percentage_charged": 50.2924,
  "energy_left": 52777.8,
  "total_pack_energy": 100000.0,
  "solar_power": 3200.0,
  "battery_power": -1800.0,
  "load_power": 1400.0,
  "grid_power": 1500.0,
  "grid_status": "Active",
  "island_status": "on_grid"
}

Before this change, energy_left and total_pack_energy were hard-coded None.

Pipeline

Updates from git push no-mistakes

✅ **intent** - passed

✅ No issues found.

✅ **Rebase** - passed

✅ No issues found.

⚠️ **Review** - 1 info

• ℹ️ src/aiopowerwall/energysite.py:251 - Minor semantic inconsistency at empty-battery edge: energy_left comes from _system_status_field (isinstance check, so a legitimate nominalEnergyRemainingWh of 0 yields 0.0), while percentage_charged comes from battery_level → battery_level_raw, whose if not remaining or not full guard treats remaining==0 as unknown and returns None. So a fully-drained pack would report energy_left=0.0 but percentage_charged=None. Real gateways rarely report exactly 0 nominal Wh, and the None-on-falsy logic lives in the unchanged client helper, so this is informational only.

✅ **Test** - passed

✅ No issues found.

• uv run pytest tests/test_energysite.py -q (40 passed), including the new test_live_status_missing_status_yields_none and extended test_live_status_shape_and_available_values
• End-to-end demo driving PowerwallEnergySite.live_status() with a fake local gateway, printing the Fleet-API-shaped response object the tesla-fleet-api EnergyRouter would consume

✅ **Document** - passed

✅ No issues found.

✅ **Lint** - passed

✅ No issues found.

✅ **Push** - passed

✅ No issues found.