physicalai-studio 0.1.0.dev1

Physical AI Studio

Physical AI Studio Application

Studio application for collecting demonstration data, managing datasets, training VLA model policies, and running trained policies on robot environments.

The application provides a graphical interface to:

• Set up robot arms and cameras.
• Create reusable robot-camera environments.
• Record and review demonstration datasets.
• Train policies using the PhysicalAI library.
• Run trained policies in Studio or deploy them with OpenVINO PhysicalAI.

Start Here

Task	Documentation
Install the application	Installation
Update an existing setup	Update Existing Installation
Complete the first workflow	Getting Started

Application Guides

Guide	Description
Installation	Install with Docker or run backend and UI natively.
Update Existing Installation	Refresh Docker images, dependencies, and services after pulling changes.
Getting Started	Create a project, set up hardware, record data, train a model, and run inference.
Environment Setup	Configure robots, cameras, and environments.
Recording Datasets	Record, review, import, and export demonstration datasets.
Training Policies	Train model policies from recorded datasets.
Deploying Model Policies	Run trained policies in Studio or deploy them externally.

Components

Component	Description	Documentation
Backend	FastAPI server for data management and training orchestration.	Backend README
UI	React web application.	UI README
Docker	Containerized application runtime.	Docker README

See Also

• Main Repository - Project overview and library quick start.
• Library - Python SDK for programmatic usage.