Skip to navigation
Kumo RelationalAbout Kumo Relational

Overview

View as Markdown

Kumo Relational is a relational foundation model for predictive AI on structured, multi-table data. It generates predictions from relational context without requiring you to train a task-specific model first.

Kumo Relational learns from your data at query time. Describe a prediction task with Predictive Query Language (PQL), a lightweight SQL-like interface, and receive results as a pandas DataFrame through the Python SDK.

You access Kumo Relational through structured-data-models-client (pip install "structured-data-models-client[relational]", imported as sdm_client), NVIDIA’s open-source Python client for structured-data foundation model NIMs. The same client serves both Kumo Relational (several related tables, joined as a graph) and Kumo Tabular (a single table) — the model you want is an argument to predict(), not a different package.

What you can predict

Classification and Regression

Predict categories, Boolean outcomes, and numeric values.

Forecasting

Predict numeric values over a future time horizon.

Link Prediction

Rank or recommend related entities.

Get Started

Install and Connect

Install structured-data-models-client[relational] and connect to a NIM serving kumo-relational.

Run Your First Prediction

Build a schema and run a prediction with SDMClient.

Examples and Notebooks

Explore end-to-end predictive workflows and datasets.

API Reference

Look up the Python SDK, PQL, and result schema.