home-assistant: fix zigbee sensor battery reporting

WSDCGQ12LM sensors report battery: 0 due to firmware quirk. Override
battery calculation using voltage via homeassistant value_template.

Also adds zigbee_sensor_stale alert for detecting dead sensors regardless
of battery reporting accuracy (1 hour threshold).

Device configuration moved from external devices.yaml to inline NixOS
config for declarative management.

Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>

docs/plans/zigbee-sensor-battery-monitoring.md

@@ -5,11 +5,11 @@
 Three Aqara Zigbee temperature sensors report `battery: 0` in their MQTT payload, making the `hass_sensor_battery_percent` Prometheus metric useless for battery monitoring on these devices.
 
 Affected sensors:
-- **Temp Living Room** (`0x54ef441000a54d3c`) — area: living_room
+- **Temp Living Room** (`0x54ef441000a54d3c`) — WSDCGQ12LM
-- **Temp Office** (`0x54ef441000a547bd`) — area: office
+- **Temp Office** (`0x54ef441000a547bd`) — WSDCGQ12LM
-- **temp_server** — area: server_room
+- **temp_server** (`0x54ef441000a564b6`) — WSDCGQ12LM
 
-The **Temp Bedroom** sensor (`0x00124b0025495463`) is a different model and reports battery correctly (69% at time of investigation).
+The **Temp Bedroom** sensor (`0x00124b0025495463`) is a SONOFF SNZB-02 and reports battery correctly.
 
 ## Findings
 
@@ -17,15 +17,47 @@ The **Temp Bedroom** sensor (`0x00124b0025495463`) is a different model and repo
 - The Zigbee2MQTT payload includes a `voltage` field (e.g., `2707` = 2.707V), which indicates healthy battery levels (~40-60% for a CR2032 coin cell).
 - CR2032 voltage reference: ~3.0V fresh, ~2.7V mid-life, ~2.1V dead.
 - The `voltage` field is not exposed as a Prometheus metric — it exists only in the MQTT payload.
-- This is a known firmware quirk with some Aqara sensors that always report 0% battery.
+- This is a known firmware quirk with some Aqara WSDCGQ12LM sensors that always report 0% battery.
 
-## Possible Solutions
+## Implementation
 
-### 1. Expose voltage as a Prometheus metric
+### Solution 1: Calculate battery from voltage in Zigbee2MQTT (Implemented)
-Enable the voltage sensor entities in Home Assistant (they may exist but be disabled by default). The HA Prometheus integration would then export them automatically.
 
-### 2. Calculate battery from voltage in Zigbee2MQTT
+Override the Home Assistant battery entity's `value_template` in Zigbee2MQTT device configuration to calculate battery percentage from voltage.
-Override the battery calculation using the voltage field. Approximate formula: `(voltage - 2100) / (3000 - 2100) * 100`.
 
-### 3. Alert on sensor staleness instead
+**Formula:** `(voltage - 2100) / 9` (maps 2100-3000mV to 0-100%)
-Create a Prometheus alert based on `hass_last_updated_time_seconds` going stale (e.g., no temperature update in 1 hour). This detects dead sensors regardless of battery reporting accuracy.
+
+**Changes in `services/home-assistant/default.nix`:**
+- Device configuration moved from external `devices.yaml` to inline NixOS config
+- Three affected sensors have `homeassistant.sensor_battery.value_template` override
+
+**Expected battery values based on current voltages:**
+| Sensor | Voltage | Expected Battery |
+|--------|---------|------------------|
+| Temp Living Room | 2710 mV | ~68% |
+| Temp Office | 2658 mV | ~62% |
+| temp_server | 2765 mV | ~74% |
+
+### Solution 2: Alert on sensor staleness (Implemented)
+
+Added Prometheus alert `zigbee_sensor_stale` in `services/monitoring/rules.yml` that fires when a Zigbee temperature sensor hasn't updated in over 1 hour. This provides defense-in-depth for detecting dead sensors regardless of battery reporting accuracy.
+
+**Alert details:**
+- Expression: `(time() - hass_last_updated_time_seconds{entity=~"sensor\\.(0x[0-9a-f]+|temp_server)_temperature"}) > 3600`
+- Severity: warning
+- For: 5m
+
+## Post-Deployment Steps
+
+After deploying to ha1:
+
+1. Restart zigbee2mqtt service (automatic on NixOS rebuild)
+2. In Home Assistant, the battery entities may need to be re-discovered:
+   - Go to Settings → Devices & Services → MQTT
+   - The new `value_template` should take effect after entity re-discovery
+   - If not, try disabling and re-enabling the battery entities
+
+## Notes
+
+- Device configuration is now declarative in NixOS. Future device additions via Zigbee2MQTT frontend will need to be added to the NixOS config to persist.
+- The `devices.yaml` file on ha1 will be overwritten on service start but can be removed after confirming the new config works.

services/home-assistant/default.nix

@@ -69,6 +69,44 @@
       frontend = true;
       permit_join = false;
       serial.port = "/dev/ttyUSB0";
+
+      # Inline device configuration (replaces devices.yaml)
+      # This allows declarative management and homeassistant overrides
+      devices = {
+        # Temperature sensors with battery fix
+        # WSDCGQ12LM sensors report battery: 0 due to firmware quirk
+        # Override battery calculation using voltage (mV): (voltage - 2100) / 9
+        "0x54ef441000a547bd" = {
+          friendly_name = "0x54ef441000a547bd";
+          homeassistant.sensor_battery.value_template = "{{ (((value_json.voltage | float) - 2100) / 9) | round(0) | int | min(100) | max(0) }}";
+        };
+        "0x54ef441000a54d3c" = {
+          friendly_name = "0x54ef441000a54d3c";
+          homeassistant.sensor_battery.value_template = "{{ (((value_json.voltage | float) - 2100) / 9) | round(0) | int | min(100) | max(0) }}";
+        };
+        "0x54ef441000a564b6" = {
+          friendly_name = "temp_server";
+          homeassistant.sensor_battery.value_template = "{{ (((value_json.voltage | float) - 2100) / 9) | round(0) | int | min(100) | max(0) }}";
+        };
+
+        # Other sensors
+        "0x00124b0025495463".friendly_name = "0x00124b0025495463"; # SONOFF temp sensor (battery works)
+        "0x54ef4410009ac117".friendly_name = "0x54ef4410009ac117"; # Water leak sensor
+
+        # Buttons
+        "0x54ef441000a1f907".friendly_name = "btn_livingroom";
+        "0x54ef441000a1ee71".friendly_name = "btn_bedroom";
+
+        # Philips Hue lights
+        "0x001788010d1b599a" = {
+          friendly_name = "0x001788010d1b599a";
+          transition = 5;
+        };
+        "0x001788010d253b99".friendly_name = "0x001788010d253b99";
+        "0x001788010e371aa4".friendly_name = "0x001788010e371aa4";
+        "0x001788010dc5f003".friendly_name = "0x001788010dc5f003";
+        "0x001788010dc35d06".friendly_name = "0x001788010dc35d06";
+      };
     };
   };
 }

services/monitoring/rules.yml

@@ -226,6 +226,14 @@ groups:
         annotations:
           summary: "Mosquitto not running on {{ $labels.instance }}"
           description: "Mosquitto has been down on {{ $labels.instance }} more than 5 minutes."
+      - alert: zigbee_sensor_stale
+        expr: (time() - hass_last_updated_time_seconds{entity=~"sensor\\.(0x[0-9a-f]+|temp_server)_temperature"}) > 3600
+        for: 5m
+        labels:
+          severity: warning
+        annotations:
+          summary: "Zigbee sensor {{ $labels.friendly_name }} is stale"
+          description: "Zigbee temperature sensor {{ $labels.entity }} has not reported data for over 1 hour. The sensor may have a dead battery or connectivity issues."
   - name: smartctl_rules
     rules:
       - alert: smart_critical_warning