Tswarm pt 3 – The code for your node

On my github page you can find first draft of the micropython code for ‘slave’ nodes which are basically responsible only for pulling data from sensor every 5 minutes, and sending it to very simple API I have created with Django + REST framework, and run on Heroku.

Let’s talk more about it…

After flashing your ESP, you can get REPL prompt by connecting node through USB UART interface and using, for example, picocom. Just like this:

sudo picocom /dev/ttyUSB0 -b115200

You may need to experiment with baud rate, but 115200 seemed to work with all my modules. After that you are able to execute python code on your device, and just play with it a little bit.

Under the hood, micropython is using two very important files: boot.py (file content is executed on device boot up) and main.py (ran just after booting finishes) that we can override with standard python file operations (just put your code inside string and write it down – or upload it with software of your choose that works with UART). Remember that it’s not fully featured dev environment, so after opening a file, you have to close it. I find little sense in modifying boot.py. You may want to put some setup code there, but we may just do the same in main.py.

Anyway on tswarm_devices/slaves/ on my github page, you will find tree files:

  • main.py – which covers main loop
  • node.py – which helds all node code
  • config_sample.json – in which you can find sample config for our device

Let’s start with configuration

It’s a little bit overconfigurated. Not everything is used (at this time anyway). It should be pretty self explanatory, but just in case:

"modules": {
  "dht11": {
    "desc": "Temperature and humidity meter",
    "gpio": [2]

Modules part contains dict with all modules connected to this particular board, as well as their configuration. In case above, just gpio (in master node I also have screen module which has some more config)

Netconf part contains wifi config. Use static IP to save energy (it’s more efficient and faster than using DHCP). I’m also providing list of SSID, cause I have few different access points with the same password in my home. Depending on localization, module will try to connect with first one, if fail we will go to second one etc.

Svconf part is responsible for webservice configuration. Every node has it’s own URL to which it posts the data. URL_local is for local testing, used during development and troubleshooting. It also contains some credentials – it’s basically standard HTTP basic authentication. You probably don’t want to use it on unencrypted service, cause it’s worth as much as sending plaintext credentials in your post. Or maybe using ROT26 to protect them ;)

Here be dragons – let’s fianlly use those leds and gpios

In node.py you can see few functions that are responsible for board operation. First we are importing some useful stuff, loading config and configuring our thermal sensor:

import machine
import dht
import ujson
import network
import urequests
import esp

# modules initialization
f = open('config.json', 'r')
config = ujson.loads(f.read())

# thermal
d = dht.DHT11(machine.Pin(config['modules']['dht11']['gpio'][0]))

Let’s move on. initialPowersave() function is for well… Saving power. So we are turning all LEDs off. Keep in mind that .low() and .high() states depends on how your board is connected. You may need to play with it a little bit, but it was pretty consistent through all my boards.

rgbLed(color) function is for handling three basic colors, and turning LED off. You have to use three pins to configure the color. I’ve added the table for different colors in previous post. It’s also in code comment.

networkSetup() – again, pretty self explanatory, but few things here. If you want to use your ESP as station instead of AP (default behavior after flashing board with micropython), you have to deactivate AP interface. After that we can pull up our station interface and try to connect with networks from our config. I’m also flashing some RGB LED colors just to know what is happening. Here is some network module documentation, I just want to point out two things: sta.status() == 1 means connecting, sta.status() != 5 means anything else than connected.

getThermalData()  is very simple. It reads DHT11 sensor and returns tuple with temperature and humidity.

uploadData() is function for handling our communication with webservice.

def uploadData():
  headers = {'authorization': ' Basic %s' % config["svconf"]["xauth"], 'content-type': 'application/json'}
  t = getThermalData()
  data = '{"name":"%s","data":"%s %s"}' % (config["svconf"]["login"], t[0], t[1])
  url = config["svconf"]["url"]

  print('webservice address: %s' % url)
  print('data to be sent: %s' % data)

  # PUT the data
  resp = urequests.put(url, data=data, headers=headers)
  if resp.status_code == 202:
    print('SUCCESS PUT: Data uploaded!')
  # if not POST the data
  elif resp.status_code == 405:
    resp = urequests.post(url, data=data, headers=headers)
    if resp.status_code == 202:
      print('SUCCESS POST: Data uploaded!')
      print('ERROR: Unable to connect with webservice! Code returned %s' % resp.status_code)
   elif resp.status_code == 423:
     return False
 return True

Firs we are setting up HTML headers and some JSON content. Prints are for debugging. Then we are using urequests library to send our data and headers to url from our config. We are sending data to RESTful API, so PUT, POST, GET, DELETE is available. If PUT fails, it means that node is not created in webservice, and we will try to POST the data (well, so maybe it’s not fully restfull design, but you can do it differently if you want). Code 423 it’s just my idea of stopping all nodes and breaking their loops (cause sometimes it’s not easy).

goToSleep(seconds) just goes to sleep for provided time period. ESP is using its own deepsleep function, don’t use the generic one. Keep in mind that you need to hook up GPIO16 with REST pin to allow board to wake up.

Where the magic happens

main.py is executed after every device boot. By every I mean also after waking up from deepsleep. I’ve decided to use main.py as orchestrator instead of putting all code inside it. So it’s very simple:

from node import *

wlan = networkSetup()
sleep = uploadData()

# if False, break deepsleep loop
if sleep:

first we are turning off all LEDs. Then we are connecting to our network. After that we are trying to upload some data to webservice. If I set ENV variable on my heroku app to return 423 status, sleep will have False value and board won’t go to sleep. If not, and 202 or anything else is returned, we are sleeping for five minutes.

Reason for this 423 status is, that during development I’ve messed something up, and had some problem with breaking this main.py loop before it went to sleep. It was just happening too fast. Ultimately i was able to kill it from REPL, but I’ve decided to have mechanism to stop all nodes by changing one ENV var on Heroku, just to simplify update process.


I hope this article helps and gives you some directions on how to tackle your board, and force it to do something useful. In the next part I will show the webservice code, and my idea of handling Django configuration on Heroku.