Guides

Each guide walks you through one task from start to finish with the simulo CLI. They assume you have installed the client and signed in, as Getting Started shows.

Train and use a policy

Preview before trainingRun the four preview checks, fix a bad robot pose, inspect the recording, and rerun before training.

Troubleshoot a scene with previewFollow a failed preview report to your task code, fix four common scene problems, and rerun it.

Train cartpoleCreate the starter app, preview it, train it on a cloud GPU, and find the policy it made.

Watch a run liveSubmit with --viewstream and open the live 3D scene with simulo view.

Get and verify checkpointsList your policies and download a best or latest checkpoint, checked against its digest.

Continue trainingStart a new training job from a policy's best or latest checkpoint with --from.

Export a policy to ONNXTurn a checkpoint into an ONNX bundle and verify it on your own machine.

Play a policy and record itRun a trained checkpoint back through its task, watch it, and record an MCAP of the run.

Evaluate a trained policyMeasure a checkpoint against a success rule, inspect recorded failures, and gate a pipeline.

Troubleshoot a policy with evaluationUse an evaluation report, episode data, and recordings to find policy, task, reward, and success-rule mistakes.

Get what a job produced

Retrieve job outputsList and download a job's checkpoints, recordings and saved files from the CLI or Simulo Console.

Record and inspect MCAPCapture an MCAP recording when training goes wrong, download it, and open it in Foxglove or Lichtblick.

Robots, assets and setup

Bring your own robotPublish a USD or URDF robot, check what validation measured, and train with it.

Browse and manage assetsSearch both catalogs, inspect versions, download your own packages, and deprecate or delete versions.

Configure environments and runtimesChoose which Simulo cloud the CLI talks to, pick a GPU tier, and add Python packages to a job.