LeLab
Official Hugging Face browser GUI for the LeRobot robot-learning workflow
License
Apache-2.0
Overview
LeLab is the official graphical interface for LeRobot, Hugging Face's open robot-learning library. It puts the full LeRobot workflow—calibration, teleoperation, dataset recording, training, inference, replay, and one-click upload to the Hugging Face Hub—into a single browser UI, removing the multi-CLI and keyboard-prompt friction that slows newcomers. You install and launch everything with a single command; a page opens in your browser and you can go from unboxing a low-cost robot arm to training your first policy in minutes. LeLab is maintained by Hugging Face, released under Apache-2.0, and also available to try as a hosted Hugging Face Space.
LeLab is a web-native front door to the LeRobot ecosystem. Each step of the robot-learning loop becomes a guided screen: a calibration flow for leader/follower arms with no keyboard prompts, live teleoperation with joint streaming, episode recording into a LeRobotDataset (with cameras), launching and monitoring a training job, running inference with a trained policy, replaying recorded episodes, and pushing datasets to the Hub. The goal is to make the powerful-but-CLI-heavy LeRobot workflow approachable for newcomers while staying fully compatible with the underlying library and hardware.
Why it matters
LeRobot is one of the most active open robot-learning ecosystems, but its power has been gated behind several command-line tools and manual steps that intimidate beginners. LeLab dramatically lowers that barrier with a guided web UI for the entire pipeline, widening access to low-cost robot learning on platforms like SO-ARM/SO-101—directly complementing the LeRobot and SO-ARM content already on the platform.
Problem it solves
LeRobot's multi-CLI, keyboard-prompt workflow is a steep barrier for newcomers who want to calibrate hardware, record demonstration data, and train policies.
Key takeaways
- Brings the entire LeRobot workflow into one browser UI
- Guided calibration, teleoperation, recording, training, inference, replay and Hub upload
- One-command install, plus a hosted Hugging Face Space to try it
- Official Hugging Face project, Apache-2.0 licensed
- Designed to take newcomers from unboxing to a first trained policy in minutes
Who it is for
- LeRobot and SO-ARM/SO-101 beginners
- Robotics educators and workshops
- Rapid prototypers and hobbyists
- Anyone collecting teleoperation datasets
Limitations
- A front-end over LeRobot—requires the LeRobot library and supported hardware
- Young, fast-moving project under active development
- Advanced or custom workflows still benefit from the LeRobot CLI/library directly