Installation

Virtualenvs are always encouraged. I recommend virtualenvwrapper with python 3.6. Python 3.5 is also supported.

$ mkvirtualenv kryptoflow
$ workon kryptoflow

Proceed with the installation

$ pip install kryptoflow

NOTE: you may need some additional dependencies. Namely, librdkafka:

On mac, with homebrew:

$ brew install librdkafka

On linux, with apt:

$ wget -qO - 'https://packages.confluent.io/deb/4.0/archive.key' | apt-key add - 
$ add-apt-repository "deb [arch=amd64] https://packages.confluent.io/deb/4.0 stable main" 
$ apt-get install -y librdkafka-dev librdkafka1

Optionally, you may need NLTK for processing tweets and reddit comments/posts. You may do so by downloading a couple of datasets with NLTK (which will be installed with kryptoflow):

$ python3 -c "import nltk; nltk.download('vader_lexicon'); nltk.download('punkt')"

Kafka, TensorFlow, et. al.

The default installation will install vanilla tensorflow. For installing tensorflow with GPU support, visit the OS-specific installation guide on tensorflow's website

Kafka's set up is pretty much dealt with automatically. If you're running Kafka locally (which is currently the only supported way), you may need to do one more tweak:

Edit your /etc/hosts file by running:

sudo nano /etc/hosts

and adding the following line to it:

127.0.0.1  localhost kafka1

That's it! Head over to the quick start to get to work.