PyCommandCenter: Library for making bots for Starcraft II

PyCommandCenter is a slimmed down version of the popular Starcraft AI CommandCenter connected to Python.

Key differences:

• Library made for use with Python, made possible by binding existing C++ code to Python using pybind11
• No decision-making, only perception
• Improved TechTree, by importing data from the JSON files provided by sc2-gamedata
• Everything is built using cmake, allowing for one Visual studio project to build PyCommandCenter together with all its dependencies

Download page for Windows

Wiki page describing how to use it

Code example for making a bot

import os

from library import *


class MyAgent(IDABot):
    def __init__(self):
        IDABot.__init__(self)

    def on_game_start(self):
        IDABot.on_game_start(self)

    def on_step(self):
        IDABot.on_step(self)


def main():
    coordinator = Coordinator()
    bot1 = MyAgent()

    participant_1 = create_participants(Race.Terran, bot1)
    participant_2 = create_computer(Race.Random, Difficulty.Easy)

    coordinator.set_participants([participant_1, participant_2])

    coordinator.launch_starcraft()

    path = os.path.join(os.getcwd(), "maps", "InterloperTest.SC2Map")
    coordinator.start_game(path)

    while coordinator.update():
        pass


if __name__ == "__main__":
    main()

How to build (Windows)

First you need to make sure you got all the build dependencies:

• cmake
• Visual Studio 2017 or later (earlier might work, but untested)
• git
• python 3.7 or later (earlier might work, but untested)

If you ever want to make any changes to the library, e.g. adding new features or making bug fixes, it's easier to fork the project before cloning the repository. For instructions on how to fork, see the gitlab documentation.

Now, you are ready to build the python library:

1. Open up a terminal, download the source code using the command: git clone --recurse-submodules https://gitlab.liu.se/course-starcraft-ai/pycommandcenter.git
2. Next, open the repository in your file viewer and run the batch script called create-visual-studio-solution.bat in order to use cmake to create a Visual studio solution
3. Open the Visual Studio solution located in the newly created directory build/
4. The project called library should be selected as the default StartUp project in the Solution Explorer (on the right side by default)
5. Change settings to Release and x64
6. Click "Local Windows Debugger" and it will start compiling
7. Visual Studio will open an error message telling you it cannot open the resulting library file, which means it successfully created the library file. The file will be located at build\python-api-src\Release and its name will depend on the python version used.

How to build (Linux, untested)

Same dependencies applies as for Windows, although you don't need Visual Studio.

1. Open up a terminal, download the source code using the command: git clone --recurse-submodules https://gitlab.liu.se/course-starcraft-ai/pycommandcenter.git
2. Next, enter the directory and run the command mkdir build followed by cd build and cmake .. in order to create the makefiles needed for building the library.
3. Run make to build the project (use make -j N if you want to use N threads)

How to use the library with PyCharm

See separate page.

Building the documentation

1. Build the library binary as described above, the documentation uses the binary to automate some parts
2. Install Sphinx
3. Go to the docs subfolder,
4. If you are not running Visual Studio and building in Release mode you have to change row 17 of conf.py to match the location of the resulting pyd file
5. Build the documentation:
   • If you are using Windows, open a terminal, navigate to the docs/ folder and run the command make.bat html.
   • If you are using Linux, open a terminal, navigate to the docs/ folder and run the command make html

Credits

CommandCenter is written by David Churchill, Assistant Professor of Computer Science at Memorial University, and organizer of the AIIDE StarCraft AI Competition.

CommandCenter is in turn based on the Blizzard's StarCraft II AI API and the architecture of UAlbertaBot.