[BUG] - Telemetry parsing fails in Service Mode — bow is required, but the controller intentionally omits <BodyOfWater> during the safety lockout

Comienza en models/telemetry.py alrededor de la línea 503, donde se define el campo requerido BodyOfWater, y revisa las convenciones de colecciones opcionales del modelo Telemetry. Usa el XML mínimo de Service Mode del issue o la suite de pruebas existente para verificar que el documento se analiza, que backyard.state sigue siendo Service Mode y que bow es una lista vacía en lugar de provocar un OmniParsingError.

Disclosure / Notes

Claude did all the heavy lifting for me from a claude code session running on my Home Assistant docker host to review logs, code and HA history. Filing the bug here and not in haomnilogic-local since the issue is part of python-omnilogic-local. I've reviewed what claude did and the information provided. Also, of note, my controller is connected via ethernet, not wifi.

Summary

Telemetry.bow is declared as a required field. When the OmniLogic controller is placed in
Service Mode from the front panel, it stops emitting <BodyOfWater> elements entirely — the
telemetry document reduces to just <Backyard>. Telemetry.model_validate then raises
OmniParsingError, and the library returns nothing at all for the whole duration of the lockout.

This is not a malformed or corrupted response. Service Mode is a safety interlock: it locks out
every remote client so that equipment cannot be started while someone is physically working on it.
The Hayward mobile app is equally locked out during Service Mode, which suggests the controller is
deliberately withholding body-of-water state from all clients rather than glitching. The library is
modelling an intentional, documented operating state as a parse failure.

Related but distinct from #155 (that one is an enum range issue on valveActuatorState; this is a
required-field issue on a structurally different response shape). Same downstream symptom.

Environment

python-omnilogic-local	5.0.0
pydantic	2.13.4
Controller firmware (mspVersion)	R0502000, build revision 28706
Telemetry STATUS version	1.12 (statusVersion="12")
Downstream	haomnilogic-local 2.0.1 on Home Assistant 2026.8.3

Evidence

Entering Service Mode from the front panel, per the controller's own USB log export:

[2026/08/25 18:53:23][Warn][UI][Current Link: node://UIFlow/shutdown.node, PreviousScreen: 2, Current Screen: 38, Running Mode: 3]
[2026/08/25 18:53:29][Notify][PoolLogic][[ 115] UI_SET_SYSTEM_STATE]
[2026/08/25 18:53:29][Warn][UI][[UI_ServiceMode_RemoveTimedServiceEvent:613]ServiceModeTimerFindAndRemoveByEvent]
[2026/08/25 18:53:29][Warn][UI][Current Link: node://UIFlow/serviceMode/serviceModeTop.node, PreviousScreen: 38, Current Screen: 41]
[2026/08/25 18:53:29][Warn][Term-Handler][HPN_SystemStateSet():194, Set System State: "Enter Service Mode", NodeId: 65021]

Thirteen seconds later, the next telemetry poll failed, and kept failing until Service Mode was
exited:

2026-08-25 18:53:42 OmniParsingError: Failed to parse Telemetry: 1 validation error for Telemetry
BodyOfWater
  Field required [type=missing, input_value={'@version': '1.12', 'Bac...spVersion': 'R0502000'}}, input_type=dict]
    For further information visit https://errors.pydantic.dev/2.13/v/missing

Note the truncated input_value: the document that arrived contained the Backyard element and
nothing else. Exactly one validation error is reported, because every other equipment
collection is already optional — bow is the only required one, so it is the only thing that
fails.

Timeline, cross-referencing the panel log against the client:

Time	Event
18:53:29	"Enter Service Mode" set on the front panel
18:53:42	First OmniParsingError — BodyOfWater missing
18:53:42 → 18:55:26	Every poll fails; 12 consecutive OmniParsingError in total
18:55:26	Service Mode exited → BodyOfWater returns → parsing resumes immediately

Nothing else was wrong: the controller was healthy throughout, the pool equipment kept running,
and the network path was unaffected.

Minimal reproduction

The XML below is a constructed minimal case, not a captured document — but it reproduces the exact
error above. state="2" is BackyardState.SERVICE_MODE.

from pyomnilogic_local.models.telemetry import Telemetry
from pyomnilogic_local.models.exceptions import OmniParsingError

SERVICE_MODE = """<?xml version="1.0" encoding="UTF-8" ?>
<STATUS version="1.12">
    <Backyard systemId="0" statusVersion="12" airTemp="101" state="2" ConfigChksum="2759708" mspVersion="R0502000" />
</STATUS>"""

try:
    t = Telemetry.load_xml(SERVICE_MODE)
    print("parsed OK, backyard state =", t.backyard.state)
except OmniParsingError as e:
    print(f"{type(e).__name__}: {e}")

Output:

OmniParsingError: Failed to parse Telemetry: 1 validation error for Telemetry
BodyOfWater
  Field required [type=missing, input_value={'@version': '1.12', 'Backyard': {'@systemId': '0', ...}}, input_type=dict]

Root cause

models/telemetry.py L503:

bow: list[TelemetryBoW] = Field(alias="BodyOfWater")

No default, so pydantic treats it as required. The class docstring 24 lines above states the
intent explicitly:

All equipment collections except backyard and bow are optional and will be None

So bow being required is deliberate — it just doesn't account for Service Mode, where the
controller legitimately reports no bodies of water.

Proposed fix

Give bow a default so a Service Mode document parses:

bow: list[TelemetryBoW] = Field(alias="BodyOfWater", default_factory=list)

I've suggested default_factory=list rather than | None = None because existing consumers
iterate telemetry.bow directly and an empty list keeps them working unchanged. That said,
None would match the convention already used for the other optional collections and would let
callers distinguish "Service Mode" from "no bodies of water configured."

Callers that want to detect the lockout explicitly can then read
telemetry.backyard.state == BackyardState.SERVICE_MODE, which already parses correctly
(BackyardState defines SERVICE_MODE = 2 and TIMED_SERVICE_MODE = 4).

I verified this by applying the change in-memory against 5.0.0 and re-validating the same
Service Mode document:

parsed OK ->  backyard.state = Service Mode | is SERVICE_MODE: True | bow = []

So the one-line change is sufficient: nothing else in the Service Mode document fails to
validate, and the resulting object carries exactly the signal a caller needs to report the
lockout rather than an error.

Downstream impact

In haomnilogic-local, OmniLogicEntity.available is:

return super().available and self.equipment._omni.backyard.is_ready

and the integration ships a dedicated binary_sensor for backyard Service Mode. So there is
already purpose-built handling for exactly this state — and none of it can execute, because
telemetry parsing fails before any of it is reached.

The practical result is that the one operating mode the safety interlock exists to make visible is
the mode Home Assistant cannot report. Instead of entities cleanly showing unavailable because the
system is in Service Mode, the coordinator raises UpdateFailed and every entity for the config
entry drops out with a generic fetch error. Measured on my system: 90 seconds of full blackout
across all 73 entities for a ~2 minute Service Mode session.

---

Identifiers (device ID, MSP ID, local IP) redacted from the log excerpts above; happy to supply a
full unredacted log privately if useful.