Lucas Bonatto Miguel

Currently, one is needed to know a convention of parameters names to set when implementing the components of the DASFE pattern. For instance, in the Preparator class, the user needs...

Currently, the engine Dockerfile has a snippet commented to be uncommented by whom wants to use Spark. Requiring user to uncomment code in order to achieve functionality is not a...

It should be possible to run Marvin's toolbox and engine-executor using docker commands. The installation steps could be as simple as: docker run -it

Adding support to build docker images with the engine-httpserver running as default

1

Now we have a CLI to build docker images. This fixes issue #35 This first version is able to build two type of docker images: - base - This is...

In order to be compatible with most PaaS that run APIs it should be possible to build a docker image using Marvin's CLI

It's not possible to install with Anaconda

2

When trying to install Marvin toolbox with Python Anaconda an error is being thrown when running the command mkvirtualenv

It should be possible to deploy a full engine (engine code + engine executor server) to AWS EC2 instances.

There is room for improvement on readability and decoupling of Actors

1

Currently there is no rule on how to define messages and the contract when exchanging these messages between actors. For example, the PredictorFSM send message to OnlineAction which pipes to...

In order to integrate with most cloud computing platforms it is necessary to have the ability to run engines as docker containers. We may have different versions of docker and...

When running the engine-executor from a remote machine, it's not possible to run batch actions (acquisitor, tpreparator, trainer and pipeline). The artifacts are being generated in the machine where the...