Python agent

Introduction

Python agent are simple agent that allows you to write python code that takes a Resource as input and return a Resource. It only supports python code and already installed package. It can be useful to manipulate an existing resource.

When you write the code of a python agent, there are 2 important variables available in your code context :

• sources : type List[Resource]
• variable already initialized with the input Resources (an array containing a Table for example)
• targets : type List[Resource]
• variable not initialized, the outputs Resources must be assigned to this variable. This will be the outputs of the task.

The context of this a python agent is the same as a Task run method . This means that you can do the same things as you would in a normal run method of a Task. You can import bricks or libraries and call methods of the Task, like self.log_info_message.

Here is a simple example of a agent that takes a Table as input/output and transpose it. It takes just 1 input and 1 output Table.

# Import modules
from gws_core import Table
# access task method to log a messages
self.log_info_message('Transposing table')
# Transpose the input table
table: Table = sources[0].transpose()
# set the new table a output or the agent
targets = [table]

How to develop an agent ?

The agent config integrate a light IDE to develop the agent. If you know what you are doing, this might be enough, but you can also develop the agent in a real IDE before integrating it to the lab. The agent is just a script with some constraints on input/output, you can easily code it outside the lab and import the code in your lab task. This will save you some time and frustration.

You can develop it in the Codelab, here is how we recommend to do it.

Ready to develop your agent ? Let's go 💪

Let's say we want to develop a agent that takes a Table as input, transpose it and return the transposed Table.

To do this we will develop and test our agent in a Jupyter notebook in Codelab. Open you Codelab, upload your data and create a Jupyter notebook file with the gws environment activated. You can find how to do this here : Jupyter notebook

Here is the code to do it.

# Initialize notebook to test agent
import env     
cwd = env.activate()    #activate GWS environment -> cwd will contain the current working dir as abolute path
# The import of any brick after the env.activate
from gws_core import File, TableImporter
# create a File resource
file = File('/lab/user/data/iris.csv')
# Import the File to Table
source = TableImporter.call(file, {"header": 0})
# convert to array
sources = [source]
#Start agent code
from gws_core import Table
targets = [sources[0].transpose()]
#End agent code
# print result Dataframe for notebook
print(targets[0].get_data())

Let's break it down

The first part until the #Start agent code comment, is the part that configure the notebook to be close to the configuration of a real agent. First we import and activate the gws environment. By using the File and TableImporter we create a Table resource from a file and assigne this Table to the source variable. We use the source variable to imitate the real agent. With this, the following code will be executed in a similar environment that the agent.

The second part is the agent code : from #Start agent code to #End agent code . This is the code we are testing and once we are happy with it, we will be able to copy it to the agent.

the Third part after the #End agent code is for testing purpose in the notebook. This is useful to check the result. This part does not need to be copied in the agent..

Log message

The python agent has the same context as a normal task, you can call the log methods. The logs will be available in the task messages. Theses methods can be useful to debug a agent.

To log a simple methods you can do : self.log_info_message('Hello world')

To log progresss you can do : self.update_progress_value(10, 'Reached 10%')

Here is a more detail information about the log methods : Log messages

Working directory

Beside sources and targets variable, a third variable named working_dir is available. This variable contains the path of an empty temp directory if you need to create file or folder .After the execution of the agent, this directory is deleted. If a resource was referencing this directory (like a File or Folder), don't worry, it will be transferred before the deletion.