Quick Installation

To install Crew Wildfire, first follow the instructions to install the base Crew framework. If you have already installed Crew and have downloaded the Wildfire Game Build here, skip to the additional instructions.

Base CREW Setup

Here we provide the setup instructions for the minimal essential components of CREW to help you get started. The following should be ran in a base conda environment in a bash shell. For Windows machines, we strongly recommend using git bash.

1) Clone the CREW github repository:

git clone https://github.com/generalroboticslab/CREW

It should have the following structure:

CREW
├── crew-dojo
├── crew-algorithms
├── Experiment_Pipeline
└── README.md

2) Create a `Builds` folder under `crew-dojo`

Download game environments for your OS from drive and move them under crew-dojo/Builds/. Uncompress the zip files so that it has the following structure:

builds

3) Setup crew-dojo:

a) Download and install Docker Desktop

This is essential for networking multiplayer games.

b) Install go v1.18:

Visit the download page go.dev.dl, find version go1.18 under Archived Versions. Download the installation file corresponding to your machine OS and architecture. Install go and make sure it is the correct version:

go version

c) While docker is up and running, install and run the image for networking:

Under crew-dojo/Nakama, run:

bash run.sh

If successful, it should be running like this in Docker:

docker

4) Setup crew-algorithms:

Under crew-algorithms, run:

bash install.sh

This will create a conda environment named crew and install audio-related packages. For the rest of the dependencies we will use poetry. To install it, activate the crew environment and use the official installer:

conda activate crew
curl -sSL https://install.python-poetry.org | python -

A message should appear after the installation indicating poetry's bin directory. Typically for Linux and OSX, it is "$HOME/.local/bin"; for Windows it is $APPDATA\\Python\\Scripts. Add this directory to your PATH environment variable and append this command to your shell configuration file.

For Linux/OSX users, add export PATH="$HOME/.local/bin/:$PATH" to your ~/.bashrc file and source it:

source ~/.bashrc

For Windows users, you may need to edit the PATH variable manually.

Check poetry can be properly called with:

poetry --version

Next, you may need to activate crew again with conda activate crew if it is not in the environment. After that, simply run:

poetry install

Note: The base CREW setup instructions are sourced from the original CREW installation guide.

Additional CREW Wildfire Setup

After completing the base CREW setup above, follow these steps to set up the wildfire-specific components:

1) Navigate to the CREW Wildfire directory. This is where all the CREW Wildfire components are held:

cd crew-algorithms/crew_algorithms/wildfire_alg

It should have the following structure:

wildfire_alg
├── algorithms/           # Main algorithm implementations
│   ├── coela/           # COELA algorithm implementation
│   ├── embodied_agents/ # Embodied agents implementation
│   ├── hmas_2/         # HMAS 2.0 implementation
│   ├── manual/         # Manual control implementation
│   └── template/       # Template for new algorithms
│
├── core/               # Core functionality
│   ├── alg_utils.py   # Common algorithm utilities
│   └── utils.py       # General utility functions
│
├── config/            # Configuration management
│   ├── build_config.py # Configuration builder
│   └── configs.py     # Configuration definitions
│
├── data/             # Data scripts
|
├── libraries/           # Action and option libraries
│   ├── *_action_library.py  # Action definitions for different agents
│   └── *_option_library.py  # Option definitions for different agents
│
└── results/          # All output files
    ├── logs/         # All chat logs, scores, and screenshots
    ├── plots/        # Generated plots
    └── data/         # Output data files

2) Activate the crew conda environment

conda activate crew

3) Install the wildfire-specific requirements:

pip install -r requirements.txt

This will install all the additional dependencies needed for CREW Wildfire.

4) Set up OpenAI API access:

a) Create an OpenAI account at OpenAI's website

b) Generate an API key at API Keys page

c) Set the API key as an environment variable:

For Linux/MacOS (add to your ~/.bashrc or ~/.zshrc):

export OPENAI_API_KEY='your-api-key-here'

For Windows (in Command Prompt):

setx OPENAI_API_KEY "your-api-key-here"

Note: Replace 'your-api-key-here' with your actual OpenAI API key. Keep your API key secure and never share it publicly.

If you encounter any issues, check out the Troubleshooting Guide.