add-atomic-action

What problem does it solve?

It helps you add a new atomic action to an EmbodiChain environment without fighting the framework’s ActionCfg / AtomicAction conventions, so your custom observation, event, reward, action, dataset, or randomization functor can be recognized and executed correctly.

Core Features & Use Cases

• ActionCfg scaffolding: Create a configclass that extends the right base config and provides a unique registry name used for discovery.
• AtomicAction implementation: Implement fast validation and an execute method that returns success plus a properly shaped trajectory and joint_ids.
• Engine registration and API export: Register the action with AtomicActionEngine and export it from the public package so it’s usable by downstream code.
• Docs and tests guidance: Update the supported-actions documentation and add minimal tests to prevent regressions.

Quick Start

Ask your AI to “Create an EmbodiChain PushAction using the ActionCfg/AtomicAction pattern, register it under the name push, export it in atomic_actions/init.py, and add a basic test for config defaults and validate().”