rerun-sdk

What it is and what it does

Rerun is a Python SDK for recording and visualizing multimodal scientific and computer-vision data. It captures images, tensors, point clouds, text, and other structured data, streaming them to an interactive Viewer for live inspection or saving to disk for later replay. The SDK is built on Rust and compiled to native wheels, giving it performance suitable for real-time logging in data pipelines.

Typical workflows involve initializing a Rerun session, logging data with typed constructors like Points3D or Tensor, and viewing results either in-process (spawn=True) or by connecting to a separate Viewer server running on the same or different machine. The package depends on numpy, pillow, pyarrow, attrs, psutil, and typing-extensions, making it suitable for scientific Python environments.

Use it for:

• Log and inspect 3D point clouds and geometric data during computer-vision algorithm development.
• Record image sequences with annotations for debugging and validating ML model outputs.
• Stream tensor data and metrics from training loops for real-time monitoring.
• Capture and replay multimodal sensor data for robotics or autonomous systems.
• Visualize structured scientific data for exploratory analysis.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Records and streams data like images, tensors, point clouds, and text to the Rerun Viewer for live visualization or file-based replay.

Yes, if you work with visual or multimodal scientific data and need live or replay-based inspection. The SDK is actively maintained, has no known vulnerabilities, and offers a permissive license. Medium install friction is typical for compiled packages; the Rust backend provides performance benefits for real-time logging. Suitable for research, prototyping, and production monitoring in computer-vision and ML workflows.

Install

rerun-sdk on PyPI

pip install rerun-sdk

uv add rerun-sdk

poetry add rerun-sdk

Installing rerun-sdk

Before you install

Medium install friction due to compiled wheels for multiple platforms (arm64, x86_64, Windows). Active maintenance with recent release (4 days old) and 11298 repository stars. Requires Python 3.10 or later.

License in practice

Dual-licensed under MIT or Apache-2.0, both permissive licenses allowing commercial and private use with minimal restrictions.

Quickstart

pip install rerun-sdk

import rerun as rr
import numpy as np

rr.init("app_name", spawn=True)
positions = np.random.rand(10, 3)
rr.log("points", rr.Points3D(positions))

Requires Python 3.10 or later; compiled wheels available for macOS (arm64), Linux (aarch64, x86_64), and Windows (x86_64).

Verify before relying

• Whether the Viewer can be run headless or requires a display for server mode.
• Performance characteristics when logging high-frequency or large-volume data streams.
• Memory overhead of maintaining live connections between SDK and Viewer processes.

Package facts

Evidence: rerun_sdk-0.36.0-cp310-abi3-macosx_11_0_arm64.whl; rerun_sdk-0.36.0-cp310-abi3-manylinux_2_28_aarch64.whl; rerun_sdk-0.36.0-cp310-abi3-manylinux_2_28_x86_64.whl; rerun_sdk-0.36.0-cp310-abi3-win_amd64.whl

Keywords: computer-vision, logging, rerun

Tags

More Artificial Intelligence packages