CondorWonder

First thing - exclude recording of the devices. My method was to use a glob so I name devices/entity IDs specifically and they don't get recorded (in my case I used f_ as in "filtered" so devices become like "F Source Presence"), but you can add specific entities or use your own glob. In configuration.yaml I have this:

recorder:
  exclude:
    entities:
      - sensor.excluded_entity_1
    # AND/OR this (then of course rename entities as needed)
    entity_globs:
      # exclude all sensor entities that start with f_
      - sensor.f_*

Then I created templates for my presence sensors, that just copy the state so I get history (yaml here, but can do through UI now too in the Helpers section, the import part is the template in the state key below):

template:
  - binary_sensor:
      - name: Real presence
        unique_id: my_presence
        state: >-
          {{ states('binary_sensor.f_source_presence}}
        availability: >-
          {{
            not (
              states('binary_sensor.f_source_presence') == 'unknown' or
              states('binary_sensor.f_source_presence') == 'unavailable'
            )
          }}
        device_class: presence

You could also use a statistics sensor to get a moving average for numeric values and get history from them too (and reduce the noise by reducing the precision and having a larger time window). This is also available through the UI - Helpers.

I’m curious too. As I understand it (and based on my observations after playing with it), it doesn’t change how often they send data to the controller, but instead changes how often the controller passes the data on. It doesn’t help the network, just the MQTT/Home Assistant side, but it can mean they flood the database less, if they’re tracking a value (like temperature). If they’re following a state (open or close) then I find they would miss the important messages and just not work well.

In my case I’ve found a few Tuya devices that seem like they have bad firmware and flood the zigbee network - human presence sensors and co/voc/climate sensors. I experimented with denouncing them, but I still ended up retiring most of those devices as they degraded my network performance and other devices couldn’t communicate very well. If it actually prevented the devices from flooding the network it would make a lot more sense to use.

I still have database filters to not record the main entities (for the mmWave presence sensors that I’m still using) and instead use a template on them to record their state as I found my database grew in size very quickly otherwise.

Yes, the packet passes through routers at each stage and they direct the packet to the ‘closest’ path based on its destination, until the final router has the destination on its network. This can happen a few times (for something in your ISP network), or 10-30+ times for something further away.

I'm using project boxes from Amazon, like these: https://a.co/d/4R4Dtv5 before I had a 3d printer to make something bespoke. Some of the boxes have the ESP board glued down, some it’s loose. It works well enough and doesn’t look too bad. I still use them now as it’s easier to throw everything into a box instead of designing something specific to the project.

Then to link the entities together into a device you need to mimic the auto discovery, or you just have two split entities.

I suppose you could create a template entity with the battery as an attribute to see it in the details view, but you still need the entities with the raw data. I’d be more inclined to create the device with auto discovery, seems like a cleaner way to go.

Zigbee2mqtt should do device auto discovery by default (it did for me and I didn’t have to do anything). Maybe you’ve turned something off? The alternative I can think of is to manually create and maintain device auto discovery records like https://stevessmarthomeguide.com/adding-an-mqtt-device-to-home-assistant/ shows (for example).

Try searching for your automation.entity_id - like in my case it’s something like automation.notify_washer_done (the original entity id of my automation, found via the developer - states tab). Then if I search using that in my YAML I’d see entity_id: automation.notify_washer_done, and add the context to see the full service call:

service: automation.turn_on
target:
  entity_id: automation.notify_washer_done
data: {}

Assuming it’s an automation or script your should find it in the related .yaml file and can scroll up to see the actual automation or script source.

Turned off or Turned on is the disable or enable action. If it’s changed by something in HA it should show what the trigger was too (like a user or other automation).

Here’s an example - it shows the automations that enabled our disabled this automation, and their trigger.

To prevent the automation from being changed you can rename it, that should break anything automatic that’s changing it. You can also try to chase down what’s changing it from logs (once renamed you should start seeing errors in your HA log file), or by searching for the entity_id in your yaml configuration files.

Check out the Zooz ZEN15 too if you want something off the shelf.

I use an Emporia Vue device, it uses an ESP32 internally and you can find instructions on how to flash it with esphome code onto it. No cloud dependency, just wifi.

You can get various kits for one/two/three phase mains, and monitor up to 16 individual circuits via passive current clamps.

I have actionable notifications with notify.notify service working so I’m not sure there - something sounds off.

I don’t think there’s issues with long timeouts, but realize that they won’t persist through restarts of Home Assistant, and depending on your automation settings you can control the number of instances. To disconnect the action from the running automation I often add the event as another trigger to the automation and then add logic to handle the normal trigger and the notification trigger separately. No wait needed in the automation then, just fire the notification and another instance of the automation handles the action.

If you set the wait to 0, it will not wait at all and will proceed immediately (or stop the automation if continue is not set). It might be counter intuitive, in some systems a 0 wait means wait forever.