VietSpeakOJ

A module that gives pronunciation score based on the audio transcript produced by Kaldi/Slack and the model transcript. Powered by the Carnegie Mellon University Pronouncing Dictionary and VOSK API

Development installation

python3 -m venv env
source env/bin/activate
pip3 install black isort pytest pytest-cov sqlalchemy wheel gunicorn flask vosk slack_bolt scikit-learn aiohttp alembic pytz flake8 flask-login Flask-Limiter pymemcache markdown
sudo apt-get install ffmpeg unzip memcached
mkdir database
mkdir tmp
python3 installer.py

Add the following lines to the end of your ~/.bashrc file

export SLACK_BOT_TOKEN=<your Slack bot token>
export SLACK_BOT_USER_TOKEN=<your Slack user token>
export SLACK_SIGNING_SECRET=<your Slack signing secret>
export PORT=<port>
export SECRET_KEY=<secret_key>
export ADMIN_PORT=<admin_port>
export ADMIN_PASSWORD=<admin_password>
export USER_PORT=<user_port>

Run the following command

. ~/.bashrc

Download a model from https://alphacephei.com/vosk/models, and unzip the model. Then, rename the new folder vosk_model

Testing

coverage run -m pytest tests && coverage report --show-missing

Database migrations

alembic revision --autogenerate -m "Your commit message"
alembic upgrade heads

Deployment

sudo ufw allow $PORT
sudo ufw allow 5000
gunicorn --bind 0.0.0.0:$PORT wsgi:flask_app
gunicorn --bind 0.0.0.0:$ADMIN_PORT add_task_app:app
gunicorn --bind 0.0.0.0:5000 wsgi2:app

Load remote database

scp $REMOTE_SERVER:~/VietSpeakOJ/database/db.sqlite ./database/db.sqlite