Structured Storage for When Your Robot Outgrows ROS Bags

What it does

Mosaico is a client-server data platform that ingests multi-modal robot sensor data and stores it as a structured, queryable archive. It replaces monolithic log formats like ROS bags with a system that supports direct random access to specific signals without parsing entire files. A Rust backend daemon handles compression and storage, while a Python SDK lets you query and upload data natively.

The interesting bit

Rather than treating sensor data as generic byte arrays, Mosaico enforces a strict data ontology that gives every object semantic meaning—making the archive validatable and queryable by physical values. It also rejects traditional file versioning in favor of immutable data layers, so query history stays deterministic. The built-in ML module then handles the tedious alignment and flattening of unsynchronized streams into fixed-frequency tensors, sparing you from intermediate CSV dumps.

Key highlights

• Zero-copy architecture allows random access to specific signals without deserializing whole log files.
• Strictly-typed data ontology replaces generic byte blobs with semantically understood, physically queryable objects.
• Immutable data layers guarantee deterministic query history instead of relying on traditional versioning.
• A Rust daemon (mosaicod) manages heavy lifting on the server side; a Python SDK provides the client interface.
• An ML module automatically resamples and aligns asynchronous sensor feeds into dense, tabular training data.

Caveats

• The README makes bullish performance claims but offers no concrete benchmarks or latency numbers to back them up.
• It is unclear how mature non-Python SDKs are; the docs mention “native SDKs (starting with Python),” implying other languages may be future work.
• The project appears to be very early (cited as version 0.0 in December 2025), so storage formats and APIs may still be shifting.

Verdict

Worth a look if you are building data infrastructure for robotics or Physical AI and need to move beyond ad-hoc ROS bag processing. Skip it if you want a drop-in visualization tool or a mature, language-agnostic ecosystem.