Interoperability Features In Kùzu: Build a simple product recommendation system

This video showcases a demo on how to leverage Kùzu's seamless integration with the PyData ecosystem and its ability to handle diverse data formats, both upstream and downstream.
✅ We demonstrate Kùzu's interoperability features for a data science workflow, using a graph-based product recommendation system as an example.
✅ We show how to ingest data from various formats like Parquet and CSV into Kùzu, transform it using Pandas DataFrames.
✅ We also showcase another great usability feature: relational database extensions, that allow you to scan and copy data directly from external relational databases. In this example, we will ingest data from a PostgreSQL database that contains tables of customers and the products they purchased, and combine it with the product data from the Parquet file to construct a customer purchase graph.
✅ The resulting graph can then be used to generate recommendations of products to customers. The workflow includes creating node and relationship tables, ingesting data, and augmenting the existing graph with data with useful metadata, such as historical sales of products.
✅ We then define some heuristics to generate recommendations based on added relationships, demonstrating that it's a breeze working with Kùzu when dealing with these various formats!
If you want to follow along and reproduce the workflow, check out this GitHub repo: github.com/kuz...
Kùzu extensions to scan/copy from relational databases (DuckDB, Postgres, SQLite, and more!):
docs.kuzudb.co...
Find us at:
Github: github.com/kuz...
Discord: / discord
Twitter: x.com/kuzudb