DCI Agent Python Framework

Contents

• 1   About
• 2   Runtime Requirements
• 3   Development
  • 3.1   Quickstart
  • 3.2   Test suite
  • 3.3   Writing agents
• 4   dci-agent-ctl

1   About

This is a python framework to write agents: An agent is a control script that will drive the execution of programs, specially for the DCI ecosystem.

2   Runtime Requirements

The framework itself tries to have as few dependencies as possible:

• Python 3.6+
• Argparse (part of the python stdlib starting with 3.x)
• Importlib metadata (part of the python stdlib starting with 3.8)
• `DCI client libraries<https://pypi.org/project/dciclient/>`_ for interacting with the DCI control server

3   Development

This tool is written with python 3.6 in mind, as that's what is currently provided for RHEL 8.4.

For development you will have to create your own virtual environment with whatever tool you choose, activate it, then install through the dev-requirements.txt file. You should have a virtual environment with the package installed and ready to run in editable mode, this means that your changes should be reflected instantly (unless you rename/move packages around)

Testing automation is done using the tox automation framework, so installing it is highly encouraged although not strictly necessary.

A highly suggested dependency is the `pre-commit framework<https://pre-commit.com>`_, which will help you speed up the feedback loop when contributing code. Internally, this will check code style using:

• python black
• flake8
• isort
• pydocstyle

`PyTest<https://docs.pytest.org>`_ is used for unit testing on top of the quality of code testing mentioned above.

Development can be done without these tools, but the checks may fail if you don't follow the appropriate guidelines, it is encouraged to configure your IDE to make things easier. If you don't have them that's fine, the checks can be performed either via pre-commit or via tox

3.1   Quickstart

In order to get you started with development you can:

1. sudo dnf install python3-virtualenv tox pre-commit # install development tools
2. pre-commit install # install the pre-commit hooks
3. tox # optionally, run the full test suite
4. virtualenv /tmp/dciagent # create the virtual environment
5. source /tmp/dciagent/bin/activate # activate the virtual environment
6. pip install -r dev-requirements.txt # install in development mode

3.2   Test suite

TBD

3.3   Writing agents

You can find an example of an agent in samples/ansible.py: the agent will execute a very simple playbook against your configured hosts as per /etc/ansible/hosts inventory (or localhost if not configured/default).

In order to write an agent you basically need to:

• import dciagent.core.agent.<agent>
• Create your Agent class inheriting from one of the agent classes
• Scaffold your entrypoint in the way you see fit as long as the entrypoint maps to your Agent.run() method

4   dci-agent-ctl

This framework also provides a single entrypoint that can leverage python namespace packages to auto-discover agents. If you develop your agent and register it in the dciagent.agents namespace, it should be automatically picked by dci-agent-ctl.

Agents successfully registered in the namespace will appear as sub-commands of the main entrypoint, you should be able to run them that way. There is an example command in dciagent/agents/example.py to showcase a sub-command registered in the namespace. Using this example, you should be able to run this like dci-agent-ctl example [opts].