ทันเอไอ

Hugging Face Proposes End-to-End Robotics Development Workflow: Record, Train, and Deploy in One Place

Hugging Face integrates Strands Agents, LeRobot, and Storage Buckets, enabling developers to capture data, train models, and deploy to robotic hardware in a single workflow.

Hugging Face has introduced an end-to-end robotics development workflow via its official blog, integrating three key tools to allow developers to collect data, train models, and deploy them to physical hardware without switching between multiple platforms.

The three tools consist of Strands Agents for robot coordination and control, LeRobot as an open-source library for training and controlling robots, and Hugging Face Storage Buckets, a customizable storage system featuring deduplication to save space when storing large sets of demonstration data.

The workflow is divided into three main phases. The first is the Record phase, capturing robot demonstration data from both simulations and physical hardware, then storing it directly in Storage Buckets. Next is the Train phase, which streams datasets from the Hub to train models immediately without needing to download all data locally, saving time and storage. The final phase is Deploy, where trained model checkpoints can be installed directly onto robot hardware.

This approach reflects the robotics industry's shift from fragmented experiments to a connected end-to-end workflow, leveraging Hugging Face's familiar AI infrastructure for robotics applications. The development was reported by tech media outlets daily.dev and AIGC.NEWS, with publication dates around August 13, 2026.