Show HN: RelativeDB – OSS query engine for relational foundation models (github.com)

RelativeDB has launched as an open-source query engine that optimally implements Relational Transformers, using a novel query language called RelQL designed to predict future events based on relational data. For instance, users can run queries to assess probabilities, such as predicting whether a customer will not place an order. The system excels in environments with multiple tables (10-100) and automatically discovers relevant subgraphs without requiring extensive feature engineering. Its reliance on pretrained models makes it effective even in data-scarce situations. This development is significant for the AI/ML community as it bridges the gap between relational databases and predictive analytics, enhancing the ability to derive insights from data without manual preprocessing. RelativeDB’s capabilities include auto-labeling GitHub issues, predicting customer churn scenarios, and more, all while maintaining impressive performance metrics for accuracy and latency across various configurations. With the ease of installation via pip and a user-friendly querying interface, RelativeDB positions itself as a powerful tool for data scientists and engineers looking to harness relational data more efficiently.