esc
Start typing to search the docs
Navigate Open

The Data Layer
for Physical AI

The data primitives to build, understand, and improve your data loop. Designed for multi-rate, multimodal data, from the first recording to massive scale.

Open source SDK.
A single toolchain to log, transform, query, view, and train on multi-rate, multimodal data.

A flexible framework to build with rather than a platform you can't change.

pip install rerun-sdk
rerun

@rerun-io/rerun

Visualize everything

Review datasets, debug detail-level issues, and extend with your own views and tools for every stage of the pipeline.

Query and transform

Full dataframe or SQL queries over any robotics data. Extend with post-processing or annotations as easily as adding columns to a table.

Log, store, and convert

Store multi-rate, multimodal data as column-chunks in rrd files. Either log directly or easily convert from any other format.

Rerun Hub

Infrastructure that powers your data loop

The production backend for the Rerun data layer. Catalog, byte-range indexing, and retrieval that turns your object stores into a queryable, streamable foundation. Run transforms on the edge or close to the data.

Query

Query into your recordings with SQL

Run any SQL or dataframe query across your catalog, down into the columns, time ranges, and values inside your recordings, not just their metadata.

Transform

Refine your data without copies

Add derived columns and evolve schemas without breaking history. You run the transforms with the SDK; Hub keeps the derived data and your raw recordings organized together.

Train

Train without an export step

Express a dataset mix as a query and stream it to your GPUs. The dataloader is column-aware and video-codec-aware, so you train directly on your recordings.

Share

Everyone works from the same data

One viewer, the same recordings, shared across the team. Explore, annotate, and trace a failure back to the data that caused it.

News

A new data layer for robot learning

With the 0.32 SDK release Rerun becomes a unified data layer for physical data — covering visualization, querying, transformation, and training of multi-rate, multimodal robotics data. This post walks through the architecture, tours the new capabilities, and introduces Rerun Hub, our commercial data catalog and storage engine.

The community loves building with Rerun

The Rerun SDK is open source and widely used across the leading physical AI teams.

LeRobot

Hugging Face's state-of-the-art robotics project uses Rerun as an integrated part of their visualization tools to inspect and debug training runs.

Brush

A 3D reconstruction engine using Gaussian splatting from DeepMind. Written in Rust, portable, flexible, fast — with Rerun visualizing training.

PyCuVSLAM

The official Python wrapper for NVIDIA's cuVSLAM library. GPU-accelerated visual SLAM and camera tracking for real-time localization.

Project Aria

Meta Reality Labs Research's egocentric AI platform. Rerun visualizes sequences in the Aria Dataset Explorer.