simulo

The Simulo SDK and CLI — define robotics simulation and training apps in Python, submit them to the Simulo cloud, monitor jobs, and retrieve trained models.

Simulo is a cloud-first robotics simulation and training platform. Write your app against a small, stable Python API; Simulo handles running it.

Install

pip install simulo

Requires Python 3.11+.

Quickstart

The fastest way to start: scaffold a runnable app.

simulo create mybot
# Created training app in mybot/ (app.py + .simuloignore).
cd mybot

--type picks the starter shape (default training): training (a single training job), inference (train → evaluate → rollout, self-contained), or scenario (a scripted scene, no learning). No network call and no login required — simulo create only writes files to your local folder. Edit app.py for your own robot and task, then see "Log in and submit" below to run it.

To download a curated collection of runnable projects, use simulo install samples. It creates ./simulo-samples as a normal Git clone, so cd simulo-samples and cat README.md open the sample index.

Choose a destination with --directory and a Git branch, tag, or commit with --ref. You can use either option on its own or combine them:

simulo install samples --directory robotics-examples --ref main
cd robotics-examples

The destination must not exist, and its parent directory must already exist. The directory appears only after the clone and requested checkout succeed. Use a tag or full commit ID instead of main to select a fixed revision. With --ref, the checkout is detached at the selected commit.

Or write an app by hand

# app.py
import simulo

app = simulo.App("hello")


@app.job(timeout=5 * 60)
def hello(name: str = "world", repeat: int = 3) -> dict:
    for i in range(repeat):
        print(f"[hello] {i + 1}/{repeat}: hello, {name}!")
    return {"greeting": f"hello, {name}", "repeat": repeat}


@app.entrypoint
def main(name: str = "world", repeat: int = 3) -> None:
    handle = hello.spawn(name=name, repeat=repeat)
    print(f"Submitted job: {handle.job_name}")

Log in and submit to the cloud

simulo login
simulo run app.py --name robot --repeat 5

Without logging in, simulo run packages the app locally without submitting it.

Every cloud submit is fresh — running the same command twice trains twice, never a silent no-op. To continue training from an earlier job's checkpoint instead, pass --from <job-ref>[:best|:latest]. <job-ref> is a full job id, matched by exact equality; an unambiguous prefix (>=4 characters, Git-style) works too, but only for the legacy UUID form some jobs from before friendly identifiers still carry.

Monitor jobs and retrieve results

simulo jobs                      # your jobs and their statuses
simulo logs --follow             # stream the latest job's logs live
simulo result                    # the completed job's result JSON
simulo models                    # list / download trained checkpoints (sha256-verified)
simulo recordings                # download MCAP flight recordings
simulo view                      # watch the live 3D scene (jobs submitted with --viewstream)
simulo cancel                    # cancel a queued or running job

Commands that take a job id default to the latest job, announcing what they resolved to on stderr.

Learn more at simulo.ai.