isoboroff / assess-react

An assessment tool

React-Assess

This is a document assessment tool implemented in Python/Flask (backend) and React (frontend). The goals are to keep things simple, with the exception that it will support labeling with highlighting as well as selecting a relevance level for an entire document.

Running

Currently the code is set up for development. First, you need an ElasticSearch server running on localhost:9200. The URL for the ElasticSearch endpoint is in back-end/server.py.

For the back end, first set up a virtual environment and install all the dependencies listed in requirements.txt:

cd back-end
python3 -m venv venv
. venv/bin/activate
pip install -r requirements.txt

Then, start the backend server: python server.py.

For the front end, you'll need Node.js (v14+) and Yarn (v1.22+):

cd front-end
yarn install
yarn build

At this point, you should be able to browse to localhost:5000 and use the tool.

Setup

The back end looks for pools to assess in a directory back-end/relevance. In there you should create a directory for each user. A pool for topic 123 should be in the file topic123, and the topic description should be in topic123.desc.

Deploy

I've included a uwsgi.ini script in back-end/ that is configured to run the server.py application through a domain socket uwsgi.sock. You'll need a clause in your nginx.conf like so:

        location = /assess { rewrite ^ /assess/; }
        location /assess { try_files $uri @assess; }
        location @assess {
                include uwsgi_params;
                uwsgi_pass unix:/home/ubuntu/react-assess/back-end/uwsgi.sock;
        }

In front-end/package.json, the homepage setting needs to be ".". The full path to uwsgi.sock and the relevance directories in backend need to be readable and executable by the nginx server process.