Arco Recipes

Amazfit 2 Dynamic Core: Integration with Python


Under the Amazit branch name, there are a lot of devices. Among them, Xiaomi sells ‘smart’ shoes (or sneakers) that have a device inside which measures some variables and also counts the steps. The device could be purchased alone, and is known as Amazfit 2 Smart Chip or Dynamic Core. In this recipe, we will use a Python library to connect to that device, and acquire the provided information.


Of course, to follow this recipe you will need the device itself, but also the following packages (assuming that you are on GNU/Linux):

The easiest way to install all these packages is using the pike Debian repository. To install it:

wget -qO- | sudo sh
sudo apt update

Then, install with apt the needed packages:

sudo apt install python3-bluez python3-gattlib amazfit2-dynamic-core

Discovering and Connecting

The first step you would do in order to check if your device is ready for connection (has enough battery, is awake, is near your Bluetooth transceiver, etc.) is a discovery.

So, first things first, import the Smart Chip library into your script. There is only one interesting class, called Core2, so:

from amazfit import Core2

Now, you can use the method find_all() (which is a @classmethod) to perform the discovery. It will try to find all the BLE (Bluetooth Low Energy) devices near you, and then filter them by its vendor ID. Because this is an endless operation, you can specify the maximum time you want to wait for responses (by default is 5 seconds), using the argument timeout.

find_all() will return a list with the found devices. A similar operation is find(), which only returns the first result in the list:
chip = Core2.find(timeout=10)
if chip is None:
    print("ERROR: no device found!")

Next step is connect to the device. This will create a communication “channel” between your computer and the chip, used to read sensor values, receive notifications, etc. For this purpose, call the method connect(), which does not have any argument:

If it returns normally, the connection has been stablished. If something went wrong (maybe the device is asleep…), it will raise a RuntimeError with a message about the problem.

About pairing

After the connection has been stablished, you can run a bonding or pairing process. It will be used to exchange a cyphering key with the device, which is the way to validate client’s identity. Also, it will reset the step counter to zero.

Although in many devices this is a critical part (otherwise, the device will disconnect or ignore your requests), on the Smart Chip is not required to read sensor values or receive notifications. If you need it, just use the method pair(), which does not accept any parameter:

During this stage, the chip will blink a red led to indicate that you must shake the device in that moment. If you don’t do it, then the pairing will fail. Otherwise, it will be successful, and you can read the exchanged key in the property key. Store it for later usage:
print(f"Pairing key: {chip.key}")

Reading Static Information

The Smart Chip holds some specific static information that barely changes, like the device address, its name, product id, vendor id, and so on. All this data could be read using the method read_info(). This method will contact with the device and retrieve all those fields. Then, you can access each one using its specific property. For a complete list of the fields, please see the API reference. The following is an example of use:
    f"Device Information:\n"
    f"- name: {}\n"
    f"- address: {chip.address}\n"
    f"- serial number: {chip.serial_number}\n"
    f"- hardware version: {chip.hardware_rev}\n"
    f"- software version: {chip.software_rev}"

Reading Step Count

There are some other values that could be read from the device, but are not considered static information because they change along the time: the step counter and the device orientation. Let’s review first the steps.

The Amazfit library exposes the step counter as a property, steps. Each time is accessed, the library will read it from the device (no caching is done). So, as an example:
print(f"- steps: {chip.steps}")

Now, you may put the device inside your shoes and walk. Then, run the program again, which should return a different result.

Of course, it is very cumbersome (and battery draining) to poll the value every few seconds. For this reason, you can use a push-notification system. The library will call a user-configured callback each time the counter changes. To setup this callback, use the method on_steps(), which accepts a single parameter: the callback function with the following signature:
def on_steps_callback(steps)

Of course, this push mechanism is asynchronous, which means that once the callback is set, the on_steps() method will return, allowing you to do other things. If you just want to wait for incoming events, you can create a waiting event loop, or just call the library’s builtin with wait_until_break().

The following example will print every new update on the step counter:
def on_steps_callback(steps):
    print(f"- steps changed, new value: {steps}")


Reading Chip Orientation

Another dynamic value that could be read from the device is the orientation, which are a pair of values representing the angle of (X, Y) axis (it is given as a tuple). Once again, you retrieve it using a property that will be read from the device each time is used:
print(f"- axis: {chip.orientation}")

It will take a little to acquire this value, and could not be read very frecuently.

Again, you can also configure a callback to be called when this value has changed (using the on_orientation() method), but, you are warned, no new notifications will be received unless someone fires a read request, with the method fire_orientation_read(). See the following example:
print(f"- axis: {chip.orientation}")

def on_orientation_callback(axis):
    print(f"- orientation changed, new value: {axis}")


while True:
    except KeyboardInterrupt:


This is a list of useful resources: