mosaico

We build Mosaico, an open-source (Apache 2.0) data platform for robotics. It takes ROS bags, MCAP files and other sensor logs and turns them into a typed, indexed archive running on your own machines. You can then search recordings by what happened in them and stream just the time slice you need into your code, without parsing whole files.

curl -fsSL https://get.mosaico.dev | sh    # mosaicod + mosaicoctl, Linux & macOS

Prefer containers? mosaicod is also published as an image (latest, x.y, x.y.z and nightly tags), and the install guide has a ready-to-use compose file with PostgreSQL:

docker pull ghcr.io/mosaico-labs/mosaicod:latest

Install the Python SDK and the mosaico CLI, then load your first recording:

pip install "mosaicolabs[cli]"                                     # Python SDK + mosaico CLI
mosaicolabs.ros_injector --file run_042.mcap --sequence run_042    # load a recording

Then see the getting started guide, the Python SDK docs and the CLI reference.

We regularly run robotics hackathons and challenges. If you're taking part in one, mosaico-alchemy and mosaico-test-datasets give you real data to practice on, and llms-full.txt has the whole documentation in one file for your coding agent. Events are announced on our Discord, where the team is around every day for questions about the SDK and robotics data.

mosaico
The platform: the Rust daemon mosaicod and the Python SDK.

mosaico-alchemy
Ready-to-use ingestion pipelines for public robotics datasets.

refx
Compile-time safe C++ library for coordinate frames and transformations in mobile robotics.

Bug reports, questions and design proposals are all welcome: use issues for bugs and discussions for ideas and RFCs. Read CONTRIBUTING.md before opening a PR. The C++ and Rust SDKs are still in development, and new sensor adapters are always useful.

_{mosaico.dev · LinkedIn}