agentos 0.0.5

AgentOS is a command line interface and python developer API for building, running, and sharing flexible learning agents.

AgentOS is an open source python API and a command line interface for building, running, and sharing learning agents. AgentOS is licensed under the Apache License, Version 2.0.

Requires Python >= 3.5 (because the CLI uses pathlib)

Install and try it out

To install:

git clone [e-mail]:agentos-project/agentos.git
pip install -e agentos # you may want to do this inside a virtualenv or conda env.

Then run a simple agent that comes with AgentOS:

cd agentos
agentos run agentos.agents.RandomAgent gym.envs.classic_control.CartPoleEnv

Then, maybe create your first Agent:

mkdir my_agent
cd my_agent
agentos init
# optionally edit ./conda_env.yaml so that `agentos` dependency points to your install of agentos
# else the command below will use agentos from PyPI.
agentos run

Also, check out the example agents in the example_agents directory.

Learn more and see the docs at agentos.org.

Tests

To run tests:

pip install -r test-requirements.txt
pytest all_tests.py

Building Documentation

To build the docs you’ll need to install and use jekyll:

gem install jekyll bundler twine
cd docs
bundle exec jekyll serve

Pushing to PyPI

To push a release to PyPI, follow these python.org instructions:

pip install setuptools wheel
python setup.py sdist bdist_wheel
python twine upload dist/*