Flaminis/Dalaran

What it solves

Dalaran provides a robotics-first observability and visualization stack for multimodal time-series data. It solves the problem of synchronizing and visualizing complex robot data—such as LiDAR sweeps, camera streams, IMU data, and TF transforms—on a single timeline, while allowing that same data to be retrieved as dataframes for computation.

How it works

Dalaran uses an Arrow-backed columnar chunk store to record and replay data from robots, simulators, or offline pipelines. It includes a 3D/2D viewer that stays in sync across sensors and provides SDKs for Python, Rust, and C++ to log data. It also features a ROS 2 bridge and rosbag2 replay capabilities, with an extensible message registry to handle custom .msg types.

Who it’s for

It is designed for robotics teams who need to visualize and debug robot sensor data and coordinate frames, and those working with ROS 2 ecosystems.

Highlights

• Robotics-specific API: High-level logging for joint states, base pose, and URDF-driven link transforms.
• ROS 2 Integration: Direct support for ROS 2 topics, rosbag2 replay, and MCAP ingestion.
• Axis Convention Helpers: Built-in support for REP-103/REP-105 (e.g., x-forward, z-up) to prevent visualization errors.
• Occupancy Grids and Costmaps: Specialized handling for nav_msgs/OccupancyGrid and nav2_msgs/Costmap with proper origin and resolution.
• Portable Datasets: .dlrpack bundles that package recordings, blueprints, and assets into a single file.
• Self-hostable: Fully open-source (Apache-2.0) with no dependency on hosted services.