Stop killing your plants

Lately I’ve been in an automation spree around the house. Since COVID-19 we’ve been working from home and spending most of our time indoors, so it makes sense that all the projects we had gathering dust are becoming important again.

In future posts I’ll talk more about what I’m using to drive my IoT devices, but the core of it all is a Home Assistant instance running inside my FreeNAS (now TrueNAS) server, in a Virtual Machine with Home Assistant Operating System.

This gives me access to the official Add-on Store and runs everything seamlessly, without fear of breaking everything up on an upgrade.

I’ve got a few wall switches, some plugs, LED strips and bulbs… everything running the latest Tasmota. Tasmota is an Open Source firmware that can be installed on almost all Chinese IoT devices that are running inside an ESP8266 (think of it like an Arduino with integrated WiFi).

Tasmota works great, and in latest version it integrates directly with Home Assistant. Older versions needed an MQTT server to broker messages to and from the devices. If you’ve never used MQTT you’ll find it similar to an email inbox and outbox, where your device sends messages to Home Assistant (ie: I turned myself on) and Home Assistant leaves messages for your device (ie: Turn yourself on).

Home Assistant integrates with MQTT as a way to communicate with a lot of stuff, not only with Tasmota devices, but with basically anything that can send messages in a format HA understands.

And thats the intro, now the more interesting part. I bought a couple of soil moisture, nutrient and light detection off Aliexpress. They’re based on the Xiaomi MiFlora and you talk with them through Bluetooth LE with a phone. They’ll tell you if your plant needs watering or if there’s too much or too little light for it to grow.

MiFlora plant sensor

My original intention was to have it talk with HA directly, since there’s a custom component that integrates this and more Xiaomi Bluetooth sensors. I couldn’t go this route because my HA instance runs in a bhyve Virtual Machine inside FreeNAS, and I can’t pass it USB hardware. That leaves me without BLE support, plus the server is too far away from the plants.

My next strategy was to setup a HA remote instance on a Raspberry Pi 3 and link it with my main Home Assistant. That proved too complicated, and Home Assistant is a bit too bulky for a Raspberry Pi 3.

Then I found out about miflora-mqtt-daemon. It’s a Python application that will talk with MiFlora devices and push messages to an MQTT server. That’s a perfect way to talk with Home Assistant, and I already had an MQTT server set up.

MiFlora MQTT Daemon debug output

I won’t go into detail of how to set it up, since the README in the official repository is very comprehensive.

Raspberry Pi Zero W

I decided to install it on a Raspberry Pi Zero W, running the latest Raspbian. I put it near my plants and power it with one of my router’s USB ports. The Zero doesn’t need as much power as a Pi 3 and that makes it perfect for this.

It took less than 10 minutes to get the decide IDs, write up a config and have them show up in Home Assistant.

Home Assistant MiFlora plant sensor integration

Next step is to add automations and a Telegram bot, so the plants can tell me when they need watering.

Rechargeable AAA batteries, with a simple little trick

I recently found out about the existence of rechargeable 1.5V AAA batteries. Most people don’t know that normal NiMH AA or AAA batteries don’t output 1.5V like a normal Duracell battery would.

NiMH AA/AAA batteries are rated to 1.2V but they’ll give you 1.4V when fully charged and start dropping their voltage when they start to deplete.

The chinese company Kentli sells 14500 format Li-ion batteries that include a buck converter that will drop the 3.7V a Li-ion battery gives to 1.5V. The form factor is the same as a normal AA/AAA battery and they’ll work in most appliances.

Medtronic 722 insulin pump

This is specially useful for insulin pump users. I just moved away from disposable batteries, that last at best 4 days, to a pair of Kentli AAA cells.

There are two main caveats to this type of batteries in pumps:

  1. They show 1.5V for the full battery life, so the pump will always show 100% battery capacity remaining and die unexpectedly.
  2. They use a special charger

The positive end of the battery has a ring, protected by a piece of plastic, that will output 3.7V. This is directly connected to the Li-ion cell and it’s where the special charger makes contact.

AAA Kentli battery
Kentli special charger

The round lead on top, goes to the buck converter, giving you exactly 1.5V for the whole duration of the cell.

I haven’t found information anywhere about their use on insulin pumps, so I’ll report back once the battery depletes or if it gives me any troubles.

You can grab a pair for 15€ on Aliexpress.

Connect Simplify3D with OctoPrint

I recently moved from Cura to Simplify3D and one of the things I really missed was the ability to upload and start a printing job directly from the slicer UI.

This is not directly supported, but Simplify3D can execute a script to post process the generated gcode file. There you can tell Simplify3D to upload and print the genearted gcode file.

You need to have cURL installed in your system. This already comes out of the box on macOS and most Linux distributions. Check this guide out to install cURL on Windows.

Once installed, edit your printed profile by editing a Process, and on the Scripts tab, under “Post processing”, input the following:

curl -k -H "X-Api-Key: YOUR_API_KEY" -F "select=true" -F "print=true" -F "file=@[output_filepath]" "http://OCTOPRINT_IP/api/files/local" {STRIP "; postProcessing"}

Replace YOUR_API_KEY with your OctoPrint API Key (under Settings > API) and OCTOPRINT_IP with the IP address where OctoPrint it’s installed.

If you just want the gcode to be uploaded but to start the print manually you can change print=true to print=false in the previous script.

DIY bias lighting for macOS

A few weeks ago, while I was reorganizing my workspace I found a 10 meter reel of NeoPixels.

Originally I wanted to put them behind a Ikea Kallax shelf where I put random junk, but since I’m working from home a lot lately it seemed a better idea to put them behind my screen!

You probably now “bias lighting” as “Ambilight“, which is the commercial name Philips uses on their TVs. It consists on placing RGB LED strips along the edges of a screen and illuminating them depending on the colors of the edges of the image.

The system has two working pieces: a client that captures screen contents and calculates the colors for each LED and a server that receives that information and actually turns each LED to the given color.


I like to start project on the hardware side, and then figure out the software. This sometimes bites me in the ass at a later time, but where’s the fun otherwise?

The finished product

To craft this project you’ll need a few things. You don’t have to use actual NeoPixels, any WS2812 RGB LEDs will do. You can also use any microcontroller that supports FastLED and can be programmed using the Arduino IDE toolchain.

Material list

  • RGB LED strip with integrated controller (WS2812)
  • Microcontroller (Arduino, ESP8266/ESP32)
  • Multi-core cable (26 AWG)
  • 1000uF capacitor
  • 3A 5V power supply
  • Double sided tape

I used an ESP32 I had lying around and I’m not sure if my LEDs are NeoPixels or random WS2812 from AliExpress, but they work.

The first thing you need to know is how many pixels are you going to wire in each side. Write this down, since you’ll need to know this number later on when configuring the client app.

If you check out the previous image you’ll see that the LEDs aren’t taped directly to the screen. Since the 2017 iMac has a curved back I designed a “holder” to angle the LEDs parallel (or almost parallel) to the wall. You probably don’t need this, but it’s available on Thingiverse.

I used 32 pixels on top and 24 on each side. You can also add them to the bottom of the screen, but I choose not to to avoid complicating the wiring.

The wiring is fairly simple. You need to place the capacitor between the 5V/3A power supply and the first LED on the first strip. Ground and VCC for the LED strip goes directly to this power supply.

Connect the DIN of the LED strip to any GPIO on the microcontroller and make a note of the pin number, you’ll need that later on.

Ground together the external power supply and the microcontroller and you’re good to go.

The microcontroller will get it’s power from the computer’s USB port and also communicate with the client application with that same USB cable.

Server software

We’ll use two pieces of software to make this work. The first is the server that will run on the microcontroller. In my case, since I’m using Prismatik, it needs to support the Adalight protocol

I can’t figure out where did I get the code I’m using. I’m sorry for not crediting the person who wrote it. If you’re reading this, do let me know!

Click on “expand source” to copy the source code for the microcontroller. You’ll need to edit the NUM_LEDS variable to reflect the total number of LEDs your strip has.

You may also want to edit the PIN variable to match the pin you wired your strip to.

// Arduino interface for the use of ws2812 operated LEDs
// Uses Adalight protocol and is compatible with Boblight, Prismatik etc
// "Magic Word" for synchronisation is 'Ada' followed by LED High, Low and Checksum
#include <FastLED.h>

///// User definitions /////

// Define the number of LEDs
#define NUM_LEDS 80

// Define SPI Pin
#define PIN 2

// Baudrate, higher rate allows faster refresh rate and more LEDs (defined in /etc/boblight.conf)
#define serialRate 115200

// Adalight sends a "Magic Word" (defined in /etc/boblight.conf) before sending the pixel data
uint8_t prefix[] = {'A', 'd', 'a'}, hi, lo, chk, i;

// initialise LED-array

void setup()
  FastLED.addLeds<WS2812, PIN, GRB>(leds, NUM_LEDS);
  // initial RGB flash
  LEDS.showColor(CRGB(255, 0, 0));
  LEDS.showColor(CRGB(0, 255, 0));
  LEDS.showColor(CRGB(0, 0, 255));
  LEDS.showColor(CRGB(0, 0, 0));
  Serial.print("Ada\n"); // Send "Magic Word" string to host

void loop() { 
  // wait for first byte of Magic Word
  for(i = 0; i < sizeof prefix; ++i) {
    waitLoop: while (!Serial.available()) ;;
    // Check next byte in Magic Word
    if(prefix[i] == continue;
    // otherwise, start over
    i = 0;
    goto waitLoop;
  // Hi, Lo, Checksum
  while (!Serial.available()) ;;;
  while (!Serial.available()) ;;;
  while (!Serial.available()) ;;;
  // if checksum does not match go back to wait
  if (chk != (hi ^ lo ^ 0x55))
    goto waitLoop;
  memset(leds, 0, NUM_LEDS * sizeof(struct CRGB));
  // read the transmission data and set LED values
  for (uint8_t i = 0; i < NUM_LEDS; i++) {
    byte r, g, b;    
    r =;
    g =;
    b =;
    leds[i].r = r;
    leds[i].g = g;
    leds[i].b = b;
  // shows new values;

Use the Arduino IDE to compile and burn this code into your microcontroller. You’ll probably need to install the latest FastLED library for it to work.

If this worked you’ll see a color flash sequence when powering the microcontroller and the LEDs.

With this done we’re ready to install and configure the client and make some LEDs light up!

Client software

I’ve tried a lot of client applications, but the best I found that works fast on macOS is Prismatik. I’m using psieg’s fork, release

On first run the configuration wizard will try to detect a Lightpack (a commercial solution to do bias lighting) and fail, letting you configure another device.

On the device selection check “Adalight”, and continue. Then you’ll need to select the serial port where the microcontroller is connected, in my case is /dev/tty.SLAB_USBtoUART, but this will vary depending on the setup. Leave everything else as is.

Next up is the Zone Placement. This maps the corners of your screen to each led. Fill out the total number of LEDs, how many of them are on top and on the sides. You can also change the thickness and width of the capture zone, this will affect how much of screen color is averaged to set the LED color.

Once you’re finished you’ll start to see the color change with the screen contents, but we’re not done yet!

Enable “Expert mode” on the profile section of Prismatik, this will let you fine tune the screen grabber. This is the most important part to get good performance in movies and fast-changing multimedia content.

Once Expert mode is enabled you’ll get a new option on the menu, click on Experimental and change the Capture source from Mac CG to Mac AV, this is the most performant screen grabber on newer versions of macOS.

On the “Mode” option you can swap between “Screen grabbing”, “Mood lamp” and “Sound visualization”.

Check “Screen grabbing” and set the Grab interval so something you feel confortable enough for your system. This can be resource intensive if set too low, so find a value that changes LEDs fast enough but doesn’t take too much CPU cycles away from you.

I have it set at 20 fps, and it usually keeps up, depending on my computer load. Prismatik sits at 8% CPU and WindowServer at 6-7%. This is in a single core, since I have a 4,2Ghz i7 it’s not noticeable at all.

Wrapping up

Last thing to do is tidy everything up. Use the double sided tape if your strips doesn’t already come with it or if you’re using the strip holders.

You can also place the microcontroller in a box to avoid desk clutter. I used an enclosure generator written in OpenSCAD and 3D printed a box to keep it all away from view.

If you have any questions don’t hesitate to comment!