Rerun

Rerun is an open-source data visualization and logging toolkit designed for spatial and embodied AI. It provides a lightweight, fast, and flexible data engine to ingest, store, visualize, and analyze multimodal data (e.g., points, transforms, images) with time-aware, interactive views. It emphasizes easy setup (no sign-up) and aims to help you run, record, view, build, and query AI system data for debugging, evaluation, and deployment workflows.

---

How it works

• Use the Rerun SDK (C++, Python, Rust) to model your data and write it to storage or stream it to a live viewer.
• The data model is a time-aware Entity Component System, enabling simple yet flexible representations for common robotics and ML scenarios.
• Visualize live or recorded data with a multimodal viewer that supports time travel, 3D layouts, images, transforms, and more.
• Build visualizations and dashboards by coding or via interactive UI, and embed the viewer into your tools and apps.
• Query APIs extract time-aligned datasets, returning Apache Arrow data for analysis with your favorite dataframes.

---

Getting started (quick start)

• Install the SDK:
• Python: pip install rerun -sdk
• C++/Rust options are also available (see docs)
• In code (Python example):
• rr.init("my_data_generating_application")
• rr.connect()
• rr.log("points", rr.Points3D(positions))
• rr.log("camera", rr.Transform3D(pos, rot))
• rr.log("camera/image", rr.Image(tensor))
• Open the interactive viewer to observe time-traveling visualizations and inspect signals such as images, 3D transforms, reprojection errors, etc.

---

Use cases

• Run & record data from systems for analysis and training.
• Visualize training/eval progress and extract time-aligned samples from logs.
• Debug prototypes and diagnose issues with a fast, multimodal viewer.
• Build custom visualizations into your tools and workflows via embedding.

---

Community & Ecosystem

• Open source, with examples and documentation.
• Integrations and projects like LeRobot, Aria Dataset Explorer, Brush, and kornia-rs demonstrate Rerun’s role in visualization across robotics and ML pipelines.
• Works natively and in the browser; can be embedded in notebooks and web apps.

---

Safety and Legal Considerations

• Primarily a developer tool for debugging and data analysis; ensure you have appropriate permissions to log and visualize data.

---

Core Features

• Open-source, no sign-up required for immediate data visualization access
• Time-aware Entity Component System data model for flexible, scalable logging
• Multimodal viewer: visualize points, transforms, images, and more in real-time or from recordings
• Time travel and fast interaction for debugging and analysis
• SDKs in C++, Python, and Rust for flexible integration
• Build and customize visualizations programmatically or via the UI
• Embedding capabilities to integrate the viewer into your apps and tools
• Data extraction via query APIs that return Apache Arrow data for downstream analysis
• Desktop and browser compatibility (native and web runtimes)
• Extensible ecosystem with examples and community-driven projects