Solarian Base
Deploy, train and run your robots
Build a project with solarian, deploy it to Base, then train it on cloud GPUs against your datasets and drive your robot from the browser.
From recording to a robot that works
The data, the compute and the robot sessions a solarian project needs, from the CLI or the web app.
Datasets you can see
Upload Solarian datasets and recordings. Play every camera in sync with the action charts, with tracked boxes and masks drawn on top.
Curate without losing data
Mirror, merge, split, trim, track objects, depth estimation and more. Modify episodes as versions of a computed dataset. Review an op episode by episode and keep only what you accept.
Deployments
Every deploy is an immutable snapshot of your solarian project, from the CLI or a git repository. Roll back with one click.
Training on cloud GPUs
Kick off a solarian training job and harness scalable, powerful cloud GPUs to train even the largest models dramatically faster.
Experiment tracking
Experiment tracking with charts, logs, and stored configurations for every run.
Control robots in the browser
Register a robot, let Base host its relay, and open a controller session to drive it from anywhere, with teleop devices included.
Built for agents
Every CLI action takes --json, the API is described by OpenAPI, and a skill walks a coding agent from deploy to a trained run.
Your train config, on a cloud GPU
Log in, link the project and deploy it. A training job runs the same config as solarian train, and its metrics and logs stream back to Base while it runs.
# Connect this project to Base
solarian base login
solarian base link --create
# Ship the code, then train on a Base dataset
solarian base deploy
solarian base train train-umi --dataset dataset_path=3f2a9c1e
# Follow the run
solarian base runs logs -f 1a2b3c4dHow it fits with solarian
- 1
Build with solarian
Robots, controllers, skills and train configs live in an open-source project you run anywhere.
- 2
Deploy to Base
A deployment is an immutable snapshot of that project, with its trainers and configs.
- 3
Train, track and operate
Jobs, runs and robot sessions run against a deployment, from the CLI or the web app.
Bring your robotics project to Base
Start with the quick start, or read what Base does and how to run it yourself.