pybela

pybela enables seamless interfacing with Bela, the embedded audio platform, using python. It offers a convenient way to stream data between Bela and python in both directions. In addition to data streaming, pybela supports data logging, as well as variable monitoring and control functionalities.

Below, you can find instructions to install pybela. You can find code examples at tutorials/ and test/. The docs are available at https://belaplatform.github.io/pybela/.

pybela was developed with a machine learning use-case in mind. For a complete pipeline including data acquisition, processing, model training, and deployment (including rapid cross-compilation) check the pybela-pytorch-xc-tutorial.

Installation and set up

You will need to (1) install the python package in your laptop, (2) set the Bela branch to dev and (3) add the watcher library to your Bela project.

1. Installing the python package

You can install this library using pip:

pip install pybela

2. Set the Bela branch to dev

pybela is relies on the watcher library, which currently only works with the Bela dev branch. To set your Bela to the dev branch, you can follow the instructions below.

Note: if you just flashed the Bela image, the date and time on the Bela board might be wrong, and the Bela libraries might not build correctly after changing the Bela branch. To set the correct date, you can either run (in the host)

ssh [email protected] "date -s \"`date '+%Y%m%d %T %z'`\""

or just open the IDE in your browser (type bela.local in the address bar).

Option A: Bela connected to internet

If your Bela is connected to internet, you can ssh into your Bela (ssh [email protected]) and change the branch:

# in Bela
cd Bela
git checkout dev
make -f Makefile.libraries cleanall && make coreclean

Option B: Bela not connected to internet

If your Bela is not connected to internet, you can change the branch by cloning the Bela repository into your laptop and then pushing the dev branch to your Bela. To do that, first clone the Bela repository into your laptop:

# in laptop
git clone  --recurse-submodules https://github.com/belaPlatform/bela
cd Bela

Then add your Bela as a remote and push the dev branch to your Bela:

# in laptop
git remote add board [email protected]:Bela/
git checkout dev
git push -f board dev:tmp

Then ssh into your Bela (ssh [email protected]) and change the branch:

# in Bela
cd Bela
git checkout tmp
make -f Makefile.libraries cleanall && make coreclean

You can check the commit hash by running git rev-parse --short HEAD either on Bela or your laptop.

3. Add the watcher library to your project

For pybela to be able to communicate with your Bela device, you will need to add the watcher library to your Bela project. To do so, you will need to add the files Watcher.h and Watcher.cpp to your Bela project. You can do this by copying the files from the watcher repository into your Bela project. To do so, you can run:

scp watcher/Watcher.h watcher/Watcher.cpp [email protected]:Bela/projects/your-project/

Getting started

Modes of operation

pybela has three different modes of operation:

• Streaming: continuously send data from Bela to python (NEW: and from python to Bela! check the tutorial).
• Logging: log data in a file in Bela and then retrieve it in python.
• Monitoring: monitor the value of variables in the Bela code from python.
• Controlling: control the value of variables in the Bela code from python.

You can check the tutorials at tutorials/for more detailed information and usage of each of the modes. You can also checktest/test.py` for a quick overview of the library.

Running the examples

The quickest way to get started is to start a jupyter notebook server and run the examples. If you haven't done it yet, install the python package as explained in the Installation section. If you don't have the jupyter notebook package installed, you can install it by running (replace pip with pipenv if you are using a pipenv environment):

pip install notebook

Once installed, start a jupyter notebook server by running:

jupyter notebook # or `pipenv run jupyter notebook` if you are using a pipenv environment

This should open a window in your browser from which you can look for the tutorials/notebooks folder and open the examples.

Basic usage

pybela allows you to access variables defined in your Bela code from python. To do so, you need to define the variables you want to access in your Bela code using the Watcher library.

Bela side

For example, if you want to access the variable myvar from python, you need to define the variable in your Bela code as follows:

#include <Watcher.h>
Watcher<float> myvar("myvar");

You will also need to add the following lines to your setup loop:

bool setup(BelaContext *context, void *userData)
{
	Bela_getDefaultWatcherManager()->getGui().setup(context->projectName);
	Bela_getDefaultWatcherManager()->setup(context->audioSampleRate);
    // your code here...
}

You will also need to add the following lines to your render loop:

void render(BelaContext *context, void *userData)
{
	for(unsigned int n = 0; n < context->audioFrames; n++) {
		uint64_t frames = context->audioFramesElapsed + n;
		Bela_getDefaultWatcherManager()->tick(frames);
        // your code here...
    }
}

you can see an example here.

Python side

Once the variable is defined "in the watcher", you can stream, log and monitor its value from python. For example, to stream the value of myvar from python, you can do:

from pybela import Streamer
streamer = Streamer()
streamer.connect()
streamer.start_streaming("myvar")

to terminate the streaming, you can run:

streamer.stop_streaming()

Testing

This library has been tested with Bela at dev branch commit 69cdf75a and watcher at main commit 903573a.

To run pybela's tests first copy the bela-test code into your Bela, compile and run it:

rsync -rvL  test/bela-test [email protected]:Bela/projects/
ssh [email protected] "make -C Bela stop Bela PROJECT=bela-test run"

you can run the python tests by running:

python test/test.py # or `pipenv run python test/test.py` if you are using a pipenv environment

Building

You can build pybela using pipenv:

pipenv install -d # installs all dependencies including dev dependencies
pipenv lock && pipenv sync # updates packages hashes
pipenv run python -m build --sdist # builds the .tar.gz file

To do and known issues

Long term

• Design: remove nest_asyncio?
• Add: example projects
• Issue: Monitor and streamer/controller can't be used simultaneously –  This is due to both monitor and streamer both using the same websocket connection and message format. This could be fixed by having a different message format for the monitor and the streamer (e.g., adding a header to the message)
• Issue: The plotting routine does not work when variables are updated at different rates.
• Issue: The plotting routine does not work for the monitor (it only works for the streamer)
• Code refactor: There are two routines for generating filenames (for Streamer and for Logger). This should be unified.
• Possible feature: Flexible backend buffer size for streaming: if the assign rate of variables is too slow, the buffers might not be filled and hence not sent (since the data flushed is not collected in the frontend), and there will be long delays between the variable assign and the data being sent to the frontend.
• Issue: Flushed buffers are not collected after stop_streaming in the frontend.
• Bug: OSError: [Errno 12] Cannot allocate memory

License

This library is distributed under LGPL, the GNU Lesser General Public License (LGPL 3.0), available here.