Problem

When fetching energy state or forecast data from Home Assistant, a single entity returning unavailable or unknown (e.g. sensor.battery_soc) caused the entire data retrieval to fail and return None. This meant:

• Optimization cycles were skipped entirely, even when most sensor data was valid.
• Frontend real-time updates were blocked due to a single temporarily unavailable sensor.
• The system was fragile against common HA transient states (restarts, integration reloads, Zigbee/Z-Wave timeouts).

The root cause was a has_critical_error flag that was set to True for any configured-but-unavailable entity, triggering return None and discarding all successfully retrieved data — despite the domain model already supporting partial data (EnergyStateSnapshot.battery and .grid are Optional).

This PR closes #69

Solution

Removed the has_critical_error mechanism entirely. Each entity failure is now handled independently with graceful degradation:

The snapshot is now always constructed with whatever data is available. Missing optional components result in None sub-objects, which downstream consumers (policy rules, decisional context, frontend) already handle.

Changes

adapters/domain/energy/monitors/home_assistant_api.py

• Removed has_critical_error flag and the early return None block.
• Grid unavailability: changed from has_critical_error = True to a warning log. GridState will be None.
• Battery unavailability: changed from has_critical_error = True to a warning log. BatteryState will be None.
• Production unavailability: changed from error to warning log. Defaults to 0W.
• Consumption unavailability: logs error but no longer blocks the snapshot.

adapters/domain/energy/monitors/home_assistant_mqtt.py

• Removed has_critical_error flag and the early return None block.
• Solar production, grid power, battery SOC/power: all changed from critical errors to warning logs.
• Consumption: logs error but snapshot is still created.

adapters/domain/forecast/providers/home_assistant_api.py

• Removed has_critical_error flag and the early return None block.
• energy_today and energy_tomorrow: changed from critical errors to warning logs.
• Forecast is always returned with whatever data is available.

Testing

• Verified that when all HA entities respond correctly, the snapshot is unchanged.
• Verified that when a single entity (e.g. battery) returns unavailable, the snapshot is still created with battery=None and all other fields populated.
• Verified that warning/error logs are correctly emitted for each unavailable entity.