Docs

Сверх Симолёт gives you a base environment (a simulator or runtime — MuJoCo, PyBullet, CUDA Python, OpenFOAM, real PX4) and runs your own code on it, on a compute node contributed to the pool.

Nothing is hardcoded: you decide what to install, what to run, what to feed in and what to get out. No baked-in task, no fixed metrics. See the run contract for the universal shape.

The core loop is always the same: author a run → submit → poll → pull artifacts. Everything else (pipelines, interactive pods, git, quotas) layers on top of that one loop.

1 Open Run and pick a base (e.g. MuJoCo), or start from a ready example.

2 Give it your code — paste it, point at a git repo, or upload a zip.

3 Set the run command (python main.py), optional setup (pip install …), inputs/outputs and resources.

4 Launch. Watch live logs + metrics, then download whatever your code wrote to /outputs.