Play Codenames against software which uses a word embedding model.

The Codenames application in this repo consists of both a server and client.

The server exposes an API to interact with the Codenames engine (i.e. for generating clues and making guesses). It is written in Python and uses gensim on the engine side and Flask on the web side.

The client renders a UI and manages the game state. It is written in JavaScript using React.

• python3 & pip3
• node & npm

• Clone this repo
• Install server dependencies: codenames/server$ pip3 install -r requirements.txt
• Install client dependencies: codenames/client$ npm install
• Download a pre-trained binary fastText model from https://fasttext.cc/docs/en/crawl-vectors.html to codenames\server\i18n

Note: Only the german binary model is supported for now: cc.de.300.bin.gz

• Start the server: codenames/server$ export FLASK_APP=server.py && python3 -m flask run
• Start the client: codenames/client$ npm start
• Open http://localhost:3000 in your web browser

Note: The server takes a long time to get ready while loading the huge fastText model into memory.

A Dockerfile is provided to create a self-contained image with:

• nginx front server delivering the static client assets and reverse proxying to the app server
• Gunicorn app server running the Python backend

To create a Docker image run: codenames$ docker build -t codenames .

To start the Docker container run: $ docker run -p 8080:80 -it codenames

Note: This will bind the exposed port 80 to your host port 8080.