Bookwise

Bookwise is a Flask web application that serves personalised book recommendations computed offline by an Apache Airflow pipeline. The key design decision is that no ML ever runs on the web request path. A daily Airflow DAG (extract, compute, load, cleanup) runs the Apriori market-basket algorithm across user ratings and writes the resulting association rules into a MongoDB collection. Flask only reads from that collection, so page loads stay fast regardless of how expensive the model gets. Airflow runs on a Celery executor with a Redis broker and PostgreSQL for metadata, while application data (users, books, ratings, and recommendations) lives in MongoDB. The whole stack comes up through Docker Compose.

Technologies Used

Python

Flask

Apache Airflow

MongoDB

Redis

PostgreSQL

Docker