Trace every model back to its source.
Every run roar captures becomes a graph you can walk — from a published model back to the code, config, and data that produced it.
Here's a real run in the registry.
Open any job or artifact to see what GLaaS recorded.
Install roar and register to GLaaS
% uv tool install roar-cli % roar init # wrap any script — no code changes % roar run python train.py % roar register model.pt
Commands, file reads and writes, git state, timing, and runtime context with no pipeline rewrite.
Hash-addressable DAGs, jobs, and artifacts so lineage stays queryable after the run is over.
Which code, data, and environment produced a model, and where that artifact flows next.
Samples in the Registry
DAG from a Nanochat training run.
A step in the execution of a Nanochat training DAG.
Model output artifact from a Nanochat training DAG.
Example Workflows
This guided example walks through a complete roar + GLaaS workflow.
More examples
coming soon
Ready to track your lineage?
Track every run, collaborate on insights, and bring structure to your ML workflows.
Sign up to get started.
GLaaS helps you understand what happened.
TReqs helps you safely re-run, review, and operationalize it.