After quite a bit of iteration, I’m mostly happy with the way I’ve integrated Dash buttons into my home automation setup. Here’s a demo:

My goals were:

1. Make the buttons as responsive as possible.
2. Make it robust.
3. Setup should survive reboots and power outages without manual intervention.
4. Integrate with SmartThings.

In a previous post, I outlined two different approaches. I went with the approach that had the lowest latency (<1s). This one is quite a bit more work — mostly because it requires a dedicated wireless card.

Here’s the equipment I used:

1. Raspberry Pi 2 (I used the CanaKit starter kit. If you’re buying now, you’d probably want the Pi 3 edition).
2. Edimax EW-7811Un USB WiFi dongle.

Important! to use this approach, you need at least one WiFi dongle that supports monitor mode. The Edimax dongle I suggested doesn’t support monitor mode, but the one that comes with the CanaKit 2 does. Note that the Pi 3’s onboard WiFi device does not support monitor mode, so you’ll want to buy a dongle that does (you can buy the CanaKit dongle separately for $9).

Set up dash buttons

This approach will work with the normal setup process, but with a slight modification, you can ensure that the dash buttons don’t contact Amazon when pressed.

The only thing you need to do differently is set up the buttons on a network you can delete later. I have dd-wrt on my router, so I used a virtual interface. If your router supports a “guest network” or something to that effect, it’s the same thing.

Create the network, set up the dash buttons on it, delete the network. The buttons will still attempt to connect when pressed, but won’t be able to because it doesn’t exist.

Install required packages

Set up the network

If you’re using ethernet + a WiFi dongle, you shouldn’t need to do much of anything. If you’re using two WiFi devices, it’s a little trickier. In order for this to work consistently across reboots, you’ll have to:

1. Make sure that the interfaces (wlan0, etc.) are named consistently. They seemed to randomly swap by default, which obviously caused some problems.
2. Tell the OS which device should be connecting to the network.

(1) is easy enough with ifrename. There’s probably a way to do it with udev, but this is way easier. It allows you to assign names to interfaces based on hardware (MAC) addresses. Open up /etc/iftab  in your favorite editor (just create it if it doesn’t exist). Mine looks like this:

After a reboot, you should see that the devices are named appropriately:

Notice you can name the interfaces whatever you want. monitorwan and mainwan seemed more informative than wlan0 and wlan1. 🙂

(2) is also pretty straightforward. There might be an easier way to do this, but I just did it by editing /etc/network/interfaces to my liking:

The wpa-psk field is a pre-shared key generated from your network SSID and passphrase. You can generate it with the  wpa_passphrase tool (from the wpasupplicant package):

You can apply these settings with a  sudo service networking restart . Probably good to reboot to make sure it works as expected.

Download ha_gateway

This setup uses ha_gateway, which is a small REST gateway I use to bridge a bunch of custom hackery with the rest of my home automation setup (mostly SmartThings). To install it, just check out the project from Github:

While I haven’t tested ha_gateway with anything but ruby 2.3.1, it probably works with 1.9+. If you’re getting errors when running bundle install , post a comment and I’ll help debug.

Create the monitor interface

In order to use monitor mode, we create a virtual monitor interface. We can do this with the iw tool, but I stuffed all of the setup into a script shipped with ha_gateway. It takes two arguments: the interface you’re using for monitor mode, and what you want to name the virtual interface

This should create an interface called DashMonitor :

To test if it’s working, you can try a tcpdump:

If you have basically any WiFi traffic around you, you should see packets pretty much immediately. If you don’t, it either means the monitor device isn’t working, or you’re legitimately not seeing traffic on whatever channel the NIC is tuned to.

To make sure the monitor device survives reboots, you can invoke the same script from /etc/rc.local :

Figure out MAC addresses of your dash button(s)

The easiest way I’ve found to do this is to use the monitor mode NIC and search for packets associated with the network you set them up on. I set my dash buttons up on a network called CMDashButton:

The hardware address is shown after “SA:” prefix.

Configure ha_gateway

ha_gateway is configured using a central YAML config file. You can just copy from the example:

Ignore the stuff at the beginning and skip down to the  listeners: key. You’ll create a listener for each dash button you want to use:

This will fire an HTTP PUT request to http://google.com/some/path with the specified params every time the button is pressed. We can worry about making it do something useful later. First, let’s verify the button presses are getting picked up.

Use the  run_listeners.sh script to fire up the ha_gateway listener process. Note you’ll have to run it with sudo — it won’t be able to listen on the monitor interface otherwise:

After waiting 10-20 seconds, press your dash button. You should see a log message that looks like this:

This means ha_gateway is successfully detecting dash button presses! Now let’s make it do something useful.

Integrating with Smart Things

ha_gateway integrates with SmartThings. We’ll be able to control your existing ST devices and routines with the dash button. Getting this working is a little complicated because SmartThings requires clients to oauth with it. Let’s get that out of the way first.

First, you’ll have to install ha_gateway’s SmartApp. Log into your ST account (https://graph.api.smartthings.com/) and click on “My Smart Apps”. Click on the green “New SmartApp” button on the right near the top. Click on the “From Code” tab and paste in this code:

This should take you to an editor page. Couple of things to do to finalize setup:

1. Publish the newly created app – click on “Publish”, then “For Me”
2. Click on the “App Settings” button, then click on the “OAuth” section.
3. Click on the “Enable OAuth for this SmartApp” button.
4. You should see two text fields containing a  “Client ID” and a “Client Secret”. Make note of ’em.
5. Click on “Update” near the bottom. OAuth settings won’t persist if you skip this!

Copy the client ID and client secret into ha_gateway’s config YAML:

Setting site_location: is important so that the OAuth redirect ends up hitting the Pi again. For now, also make sure that require_hmac_signatures:  is set to false. It’ll make it easier to go through the OAuth process.

Now fire up the ha_gateway web server by running  bin/run.sh . Now navigate to:

http://ip-of-your-pi:8000/smartthings/authorize

This should direct you to an OAuth page on ST’s site. Select a hub, check the switches you want to allow control of, and click “Authorize”. You’ll be redirected to and endpoint that outputs a JSON blob containing information about the devices you authorized, which might look something like this:

You can control each device via RESTful PUT requests. For example:

This would send the “toggle” command to “Xmas Tree”, which would turn it off since its previous status was on.

If you wanted to configure a dash button to switch on and off your Christmas Tree, you’d edit the listener config like so:

Notice we don’t need to provide the full URL, just the path. ha_gateway will assume we want to send the request to its REST server. It’ll fill in the URL specified in the site_location:  key.

You can start both the REST server and the listener process with the included start script. It’ll run the listener process as root, so make sure you’ve got an active sudo session (i.e., make sure it’s not prompting for a password):

Logs are in logs/ha_gateway.log  and logs/listeners.log .

You can also run routines. You can access /smartthings/routines to get a list of routines. To run a routine, send a GET request to /smartthings/routines/<routine_name>. Normalize routine_name to be all lowercase, remove non-alphanumeric characters, and replace spaces with underscores (e.g., “Good Night!” -> “good_night”).

Starting ha_gateway at boot

Obviously we want the REST server and the listener process to survive a reboot. This is pretty easy. I use monit because I already had it set up, but it’s probably more straightforward to just add this line to /etc/rc.local :

Make sure it appears above the exit 0  at the end of the script.

Securing it

If you don’t mind anyone on your network being able to access ha_gateway (and therefore turn off your Christmas Cheer), you can enable HMAC signatures. This will require anyone making a request to sign the request with a shared secret. Just edit the config file:

Conclusions

This works really well for me. It was way more work than I expected when I decided to look into hacking the dash buttons. I have five dash buttons for various uses, and they work very reliably. Adding new buttons is really straightforward.