Monitor your plants with Home Assistant

I love to have a lot of plants in my apartment, but unfortunately I sometimes forget to water them. So why not using modern technology to make sure, I don’t ever forget it again.

What you need

  • A Raspberry Pi or a Linux-PC with a supported Bluetooth LE interface.
  • Bluez
  • Mi Plant flower sensor for every plant
  • Home assistant

Setup

Home assistant

First you should install Home Assistant and make sure everything works well. Home assistant provides good documentation how to setup the platform.

Bluetooth software

The Linux bluetooth software might not yet be installed on your system.

Check communication with the plant sensor

The easiest way to make sure your PC can read data from the sensor is using the hcitool command line tool.

hcitool lescan
LE Scan ...
XX:XX:XX:XX:XX:XX (unknown)
XX:XX:XX:XX:XX:XX Flower mate

It might list a lot of other devices. Just make sure, it can also see the “Flower mate” device. If you can’t see this device, the distance between the flower sensor and your PC might be too large. Try moving the plant near your PC and check if it works then.

Integrate polling into Home Assistant

First, you need to poll the data from the sensor in Home Assistant:

sensor
  platform: miflora
  mac: xx:xx:xx:xx:xx:xx
  name: Flower 1
  force_update: false
  monitored_conditions:
   - moisture
   - light
   - temperature
   - conductivity

As the MAC address, you need to use the address that is shown in the hcitool lescan output.

Notify yourself when you come home

It doesn’t make much sense to send you a notification during the day that you should water the flowers. You might have forgotten it already when you come home. So let’s just send a notification when you come home.

Setup presence detection

Home Assistant documentation how to do this. There are a lot of different device trackers. Have a look and decide what works best for you.

Setup notifications

The next step is sending a notification to you. For this guide I use Pushetta, but Home Assistant provides a large number of notification mechanisms. Have a look at them and select the one that fits best for you.

An example notifier configuration might look like this:

notify
  name: pushetta
  platform: pushetta
  channel_name: mychannel
  api_key: xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Define the automation

Now comes the new rule:

automation:
  alias: Alarm me to water plants
  trigger:
    platform: state
    entity_id: device_tracker.myname
    from: 'not_home'
    to: 'home'
  condition:
    condition: numeric_state
    entity_id: sensor.flower1_moisture
    below: 20
  action:
    service: notify.pushetta
    data:
      message: 'Water Flower 1!'

What does this do? First it checks it check, if you came from (device tracker changes from not_home to home) and check then if the moisture is below a well-defined threshold. The value might be a bit different for your plant, you need to adjust this. If the moisture is too low, it sends a notification via Pushetta.

You can combine multiple plants in the condition statement or create an automation for every single plant. I personally prefer the latter as I then know exactly which plant I should water.

Long-term monitoring

It often helps to have long-term data available to optimize not only watering, but also fertilize in the right intervals. While you can store the whole history within Home assistant, it isn’t really optimized for this use case. Therefore I recommend to use InfluxDB for this. This is a time-series database specifically designed to store sensor data over a long period. To visualize the data, I use Grafana. With this, you can create nice reports like this:
TODO

Reverse engineering the Mi flora plant sensor

The Xiaomi Mi plant sensor is a well-made plant sensor that does not only measure the humidity, but also fertility, temperature and environment light. Pretty cool for a small device like this. And even cooler is that it uses Bluetooth Low energy for communication with your mobile device. This should make it easily hackable.

Ok, let’s start and check what characteristics are available:

handle = 0x0002, char properties = 0x02, char value handle = 0x0003, uuid = 00002a00-0000-1000-8000-00805f9b34fb
handle = 0x0004, char properties = 0x02, char value handle = 0x0005, uuid = 00002a01-0000-1000-8000-00805f9b34fb
handle = 0x0006, char properties = 0x0a, char value handle = 0x0007, uuid = 00002a02-0000-1000-8000-00805f9b34fb
handle = 0x0008, char properties = 0x02, char value handle = 0x0009, uuid = 00002a04-0000-1000-8000-00805f9b34fb
handle = 0x000d, char properties = 0x22, char value handle = 0x000e, uuid = 00002a05-0000-1000-8000-00805f9b34fb
handle = 0x0011, char properties = 0x1a, char value handle = 0x0012, uuid = 00000001-0000-1000-8000-00805f9b34fb
handle = 0x0014, char properties = 0x02, char value handle = 0x0015, uuid = 00000002-0000-1000-8000-00805f9b34fb
handle = 0x0016, char properties = 0x12, char value handle = 0x0017, uuid = 00000004-0000-1000-8000-00805f9b34fb
handle = 0x0018, char properties = 0x08, char value handle = 0x0019, uuid = 00000007-0000-1000-8000-00805f9b34fb
handle = 0x001a, char properties = 0x08, char value handle = 0x001b, uuid = 00000010-0000-1000-8000-00805f9b34fb
handle = 0x001c, char properties = 0x0a, char value handle = 0x001d, uuid = 00000013-0000-1000-8000-00805f9b34fb
handle = 0x001e, char properties = 0x02, char value handle = 0x001f, uuid = 00000014-0000-1000-8000-00805f9b34fb
handle = 0x0020, char properties = 0x10, char value handle = 0x0021, uuid = 00001001-0000-1000-8000-00805f9b34fb
handle = 0x0024, char properties = 0x0a, char value handle = 0x0025, uuid = 8082caa8-41a6-4021-91c6-56f9b954cc34
handle = 0x0026, char properties = 0x0a, char value handle = 0x0027, uuid = 724249f0-5ec3-4b5f-8804-42345af08651
handle = 0x0028, char properties = 0x02, char value handle = 0x0029, uuid = 6c53db25-47a1-45fe-a022-7c92fb334fd4
handle = 0x002a, char properties = 0x0a, char value handle = 0x002b, uuid = 9d84b9a3-000c-49d8-9183-855b673fda31
handle = 0x002c, char properties = 0x0e, char value handle = 0x002d, uuid = 457871e8-d516-4ca1-9116-57d0b17b9cb2
handle = 0x002e, char properties = 0x12, char value handle = 0x002f, uuid = 5f78df94-798c-46f5-990a-b3eb6a065c88
handle = 0x0032, char properties = 0x0a, char value handle = 0x0033, uuid = 00001a00-0000-1000-8000-00805f9b34fb
handle = 0x0034, char properties = 0x1a, char value handle = 0x0035, uuid = 00001a01-0000-1000-8000-00805f9b34fb
handle = 0x0037, char properties = 0x02, char value handle = 0x0038, uuid = 00001a02-0000-1000-8000-00805f9b34fb
handle = 0x003b, char properties = 0x02, char value handle = 0x003c, uuid = 00001a11-0000-1000-8000-00805f9b34fb
handle = 0x003d, char properties = 0x1a, char value handle = 0x003e, uuid = 00001a10-0000-1000-8000-00805f9b34fb
handle = 0x0040, char properties = 0x02, char value handle = 0x0041, uuid = 00001a12-0000-1000-8000-00805f9b34fb

Looks like some standard characteristics (the uuid’s starting with 00002) and a lot of non-standard. Let’s check the standard characteristics first. UUID 00002a00-0000-1000-8000-00805f9b34fb should contain the device name:

# gatttool --device=C4:7C:8D:60:8F:E6 --char-read -a 0x03
Characteristic value/descriptor: 46 6c 6f 77 65 72 20 6d 61 74 65 

“46 6c 6f 77 65 72 20 6d 61 74 65” is “Flower mate” in ASCII. Cool. Seems that Xiaomi is using the standards – not something we see from every Chinese company.

How do you know that this UUID contains the name? Check out the list of GATT characteristics on the Bluetooth web site.

Ok, let’s have a look at all characteristics that return values:

0x03 : 160823-164329 : 46 6c 6f 77 65 72 20 6d 61 74 65
0x05 : 160823-164329 : 00 00
0x07 : 160823-164329 : 00
0x09 : 160823-164329 : 0a 00 14 00 00 00 f4 01
0x0e : 160823-164329 : 01 00 ff ff
0x12 : 160823-164329 : c3 c8 fd 30
0x15 : 160823-164329 : 98 00
0x17 : 160823-164329 : 1f 8e 8e 13 86 dd a0 d8 52 66
0x1d : 160823-164329 : 4c cc ca 13 fc fa f7 a6 51 50 c6 85 ee 61 85 47 7f 3d d6 6b
0x1f : 160823-164329 : 2e a1 bc 3e b3 87 94 e9 68 6f ff b7
0x2f : 160823-164329 : 00
0x33 : 160823-164329 : aa bb
0x35 : 160823-164329 : f5 00 00 00 00 00 00 10 61 00 00 00 00 00 00 00
0x38 : 160823-164329 : 64 10 32 2e 36 2e 32
0x3c : 160823-164329 : aa bb cc dd ee ff 99 88 77 66 55 44 33 22 11 10
0x3e : 160823-164329 : aa bb cc
0x41 : 160823-164329 : 81 8d 23 00

Hmm, look pretty wild. Looking at the data you have no idea what they might encode.

Let’s try something different and dump all values every 10 minutes.

We see a lot of values that do not change at all like this:

0x38 : 160823-073844 : 64 10 32 2e 36 2e 32
0x38 : 160823-074852 : 64 10 32 2e 36 2e 32
0x38 : 160823-075856 : 64 10 32 2e 36 2e 32
0x38 : 160823-080859 : 64 10 32 2e 36 2e 32
0x38 : 160823-081909 : 64 10 32 2e 36 2e 32
0x38 : 160823-082914 : 64 10 32 2e 36 2e 32
0x38 : 160823-083918 : 64 10 32 2e 36 2e 32
0x38 : 160823-084922 : 64 10 32 2e 36 2e 32
0x38 : 160823-085925 : 64 10 32 2e 36 2e 32
0x38 : 160823-090931 : 64 10 32 2e 36 2e 32
0x38 : 160823-091936 : 64 10 32 2e 36 2e 32
0x38 : 160823-092940 : 64 10 32 2e 36 2e 32
0x38 : 160823-093945 : 64 10 32 2e 36 2e 32
0x38 : 160823-094951 : 64 10 32 2e 36 2e 32
0x38 : 160823-100006 : 64 10 32 2e 36 2e 32
0x38 : 160823-101010 : 64 10 32 2e 36 2e 32
0x38 : 160823-102014 : 64 10 32 2e 36 2e 32
0x38 : 160823-103018 : 64 10 32 2e 36 2e 32
0x38 : 160823-104022 : 64 10 32 2e 36 2e 32

Others seem to change randomly like this

0x41 : 160823-073844 : d4 0d 23 00
0x41 : 160823-074852 : 31 10 23 00
0x41 : 160823-075856 : 8e 12 23 00
0x41 : 160823-080859 : f0 14 23 00
0x41 : 160823-081909 : 4d 17 23 00
0x41 : 160823-082914 : aa 19 23 00
0x41 : 160823-083918 : 02 1c 23 00
0x41 : 160823-084922 : 5f 1e 23 00
0x41 : 160823-085925 : bc 20 23 00
0x41 : 160823-090931 : 19 23 23 00
0x41 : 160823-091936 : 76 25 23 00
0x41 : 160823-092940 : d3 27 23 00
0x41 : 160823-093945 : 30 2a 23 00
0x41 : 160823-094951 : 97 2c 23 00
0x41 : 160823-100006 : f4 2e 23 00
0x41 : 160823-101010 : 51 31 23 00
0x41 : 160823-102014 : ae 33 23 00
0x41 : 160823-103018 : 06 36 23 00
0x41 : 160823-104022 : 63 38 23 00
0x41 : 160823-105025 : c0 3a 23 00

But then there is one characteristic with very controlled changes:

0x35 : 160823-073844 : f2 00 00 68 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-074852 : f2 00 00 54 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-075856 : f1 00 00 54 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-080859 : f2 00 00 54 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-081909 : f2 00 00 68 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-082914 : f2 00 00 54 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-083918 : f2 00 00 68 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-084922 : f2 00 00 68 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-085925 : f2 00 00 57 00 00 00 10 66 00 00 00 00 00 00 00
0x35 : 160823-090931 : f2 00 00 57 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-091936 : f2 00 00 57 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-092940 : f2 00 00 57 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-093945 : f2 00 00 57 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-094951 : f2 00 00 6b 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-100006 : f2 00 00 6b 00 00 00 10 63 00 00 00 00 00 00 00
0x35 : 160823-101010 : f2 00 00 6b 00 00 00 10 63 00 00 00 00 00 00 00
0x35 : 160823-102014 : f2 00 00 79 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-103018 : f2 00 00 79 00 00 00 10 63 00 00 00 00 00 00 00
0x35 : 160823-104022 : f2 00 00 6b 00 00 00 10 65 00 00 00 00 00 00 00
0x35 : 160823-105025 : f2 00 00 79 00 00 00 10 65 00 00 00 00 00 00 00

Seems like there are 4 bytes of values and a lot of zeros. Could these 4 byte be the sensor values? They are!
Now let’s check the values in the Mi app on the iPhone:

  • 24.2 degree celcius
  • 121 Lux
  • 16% moisture
  • 101 us/cm fertility (whatever this means)

Except for the temperature these values map nicely to the data:

  • 16 = 0x10 – byte 8
  • 121 = 0x79 – byte 4
  • 101 = 0x65 – byte 9

The only thing that isn’t yet clear is the encoding of the temperature.

Also the sensor should be able to measure up to 10000 lux. This won’t work with a single byte. So, let’s put another plant it into the bright sunlight:

25 01 00 f7 26 00 00 28 0e 01 00 00 00 00 00 00
  • Fertility is now 270, which is bytes 9 and 10 (MSB byte 9, LSB byte 10)
  • Sunlight is 9975 lux which is stored in bytes 4 and 4 (MSB byte 4, LSB byte 5)
  • And now it is also getting clear how the temperature is encoded. It is in 0.1 degree Celcius steps in bytes 1 and 2, which gives us 29.3 degree celcius in this example (which fits the data displayed by the app).