Rerun is an open-source viewer for time-series and multimodal data used in robotics and physical AI (images, point clouds, tensors, transforms, etc.).
In 0.30, you can plot arbitrary scalar data directly in time series views — floats, ints, uints, bools, including nested Arrow data — even without predefined semantics.
This release also introduces a new extension model: you can register custom visualizers directly into existing 2D/3D/Map views, define your own archetypes, and use fully custom shaders — without forking the Viewer. That means domain-specific GPU renderers (e.g., height fields, cost maps, learned fields) can live inside the standard app.
The Viewer now supports on-demand streaming when connected to a Rerun server or Rerun Cloud, fetching only what you’re viewing and evicting stale data as you scrub. This enables working with recordings larger than RAM — including in the web viewer beyond the 4 GiB Wasm limit.
Interested in feedback from people building robotics tooling, telemetry pipelines, or large-scale visualization systems.
In 0.30, you can plot arbitrary scalar data directly in time series views — floats, ints, uints, bools, including nested Arrow data — even without predefined semantics.
This release also introduces a new extension model: you can register custom visualizers directly into existing 2D/3D/Map views, define your own archetypes, and use fully custom shaders — without forking the Viewer. That means domain-specific GPU renderers (e.g., height fields, cost maps, learned fields) can live inside the standard app.
The Viewer now supports on-demand streaming when connected to a Rerun server or Rerun Cloud, fetching only what you’re viewing and evicting stale data as you scrub. This enables working with recordings larger than RAM — including in the web viewer beyond the 4 GiB Wasm limit.
Interested in feedback from people building robotics tooling, telemetry pipelines, or large-scale visualization systems.