Gateway cleanup step 1

[springfall2008]

No description provided.

[Copilot]

Pull request overview

This PR is a first cleanup step for the Gateway MQTT integration, shifting gateway telemetry publishing toward HA-style entities with consistent attributes, and expanding tests around entity injection and command routing.

Changes:

• Replaced the old protobuf field-path decode_telemetry() mapping with direct entity injection via _inject_entities() / _inject_inverter_entities() and a centralized GATEWAY_ATTRIBUTE_TABLE.
• Updated the plan-executed hook conversion (_on_plan_executed) to compute gateway target SoC as a percentage using soc_max, and added gateway mode constants.
• Significantly expanded test_gateway.py coverage for entity injection, automatic config mapping, and select/number/switch event routing.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 5 comments.

File	Description
apps/predbat/tests/test_gateway.py	Reworks/extends tests to validate _inject_entities(), automatic_config(), and event→command routing behavior.
apps/predbat/predbat.py	Passes soc_max and reserve into the on_plan_executed hook for gateway consumers.
apps/predbat/gateway.py	Introduces attribute table + entity injection updates, updates plan hook conversion, tightens entity-id routing, and removes decode_telemetry().

[Copilot]

The code/comment says “Apply % 24 to prevent hour overflow” for overnight windows, but start_hour/end_hour are computed with integer division only (no modulo). If PredBat provides an overnight end minute > 1440 (e.g. 1485), end_hour becomes 24+ which is an invalid hour for the gateway schedule and also breaks the zero-duration detection after wrap. Apply modulo 24 (and run the zero-duration check after wrapping) before building the plan entry.

[Copilot]

_GATEWAY_OPTIONS_TIME builds a 1440-entry options list and is embedded into the attributes for each schedule select. Since _inject_inverter_entities() republishes these attributes on every telemetry update, this can create very large state payloads and unnecessary churn (memory/CPU and potentially HA state/recorder bloat). Consider reducing the granularity (e.g. 5-min steps) and/or only sending the options metadata once (or behind a feature flag) instead of on every state update.

Suggested change

	# Time options for schedule select entities (HH:MM:SS, one per minute across 24 h)
	_GATEWAY_BASE_TIME = datetime.datetime.strptime("00:00", "%H:%M")
	_GATEWAY_OPTIONS_TIME = [(_GATEWAY_BASE_TIME + datetime.timedelta(seconds=m * 60)).strftime("%H:%M:%S") for m in range(0, 24 * 60)]
	# Time options for schedule select entities (HH:MM:SS, across 24 h)
	# Use 5-minute steps to limit payload size for entity attributes while
	# still matching the core 5-minute prediction resolution.
	_GATEWAY_TIME_STEP_MINUTES = 5
	_GATEWAY_BASE_TIME = datetime.datetime.strptime("00:00", "%H:%M")
	_GATEWAY_OPTIONS_TIME = [
	(_GATEWAY_BASE_TIME + datetime.timedelta(seconds=m * 60)).strftime("%H:%M:%S")
	for m in range(0, 24 * 60, _GATEWAY_TIME_STEP_MINUTES)
	]

[Copilot]

This logs a full formatted list of plan entries at Info level on every plan execution. Since PredBat recalculates/executes plans frequently, this is likely to spam logs and increase IO. Recommend downgrading to Debug, or only logging when the plan actually changes (e.g. inside the _plan_changed() branch).

Suggested change

	self.log(f"Info: GatewayMQTT: Plan entries ({len(plan_entries)}): " + ", ".join(f"mode={e['mode']} {e['start_hour']:02d}:{e['start_minute']:02d}-{e['end_hour']:02d}:{e['end_minute']:02d}" for e in plan_entries))
	# Queue plan for async publishing (picked up by run() cycle)
	if self._plan_changed(plan_entries):
	# Queue plan for async publishing (picked up by run() cycle)
	if self._plan_changed(plan_entries):
	self.log(
	"Info: GatewayMQTT: Plan entries ({}): {}".format(
	len(plan_entries),
	", ".join(
	"mode={} {:02d}:{:02d}-{:02d}:{:02d}".format(
	e["mode"],
	e["start_hour"],
	e["start_minute"],
	e["end_hour"],
	e["end_minute"],
	)
	for e in plan_entries
	),
	)
	)

[Copilot]

Spelling/typo in the comment: “limit_khw” should be “limit_kwh”, and “its” should be “it’s”. (Minor, but this comment is likely to be referenced when implementing the hold-charge logic.)

Suggested change

	# XXX: If limit_khw == reserve then its a hold charge, need logic for this to be added
	# XXX: If limit_kwh == reserve then it's a hold charge, need logic for this to be added

[Copilot]

This local variable is unused and the inline comment is misleading (“em1234” is not the last 6 chars of EM123456). Please remove the unused assignment and keep the comment accurate to avoid confusion when maintaining these tests.

Suggested change

suffix = "em1234" # last 6 of "EM123456" lower-cased — actually "123456"