European Researcher's Night Software

This Django application was used for the European Researcher's Night 2024 to collect data. The code was cleaned up for public release, but take note that it was not extensively tested after this.

Development installation

Dependencies are handled with Poetry. See make run for the recommend way to run the project. Dutch translations are handled with Django's built-in translation system.

Production deployment

Make sure to change SECRET_KEY to a random string and ALLOWED_HOST to the domain name of the production server in docker-compose.yml before deploying to production.

docker compose run --rm django-ern python manage.py migrate
docker compose run --rm django-ern python manage.py createsuperuser --username admin --email admin@localhost
docker compose up -d

You should also set up a reverse proxy with HTTPS, such as Traefik, to serve the application.

The sqlite3 database is stored in a Docker volume, which is mounted in the container. The db.sqlite3 file is created automatically if it does not exist.

Model deployment

This project assumes that EN_OPENAI_HOST and NL_OPENAI_HOST refer to OpenAI-compatible endpoints, such as vLLM. It also assumes that there is a separate deployment for the Dutch and English models.

For the European Researcher's Night, the CIT provided an SSH tunnel with port forwarding to the GPU machine. Ask them for the details. Basically, you can create a tunnel to the GPU server with the below command on the host machine, and then reach it from Docker containers through host.docker.internal:

ssh -N -L 8890:localhost:8890 -L 8891:localhost:8891 gpu-machine-tunnel

Getting a CSV dump of the data

Using the sqlite3 command line tool, you can dump the collected data to a CSV file with the following commands:

sqlite3 db/db.sqlite3
.headers on
.mode csv
.output db.csv
SELECT * FROM webapp_sample;
.quit