Action Plugin Overview

​

Action Plugin Overview

The Action Plugin is a core component of OM1 that enables agents to drive robots to interact with their environment by executing commands. It serves as the output mechanism for the agent, translating high-level decisions from the Large Language Model (LLM) into concrete physical or digital actions (e.g. moving a robot, speech, etc.). This page covers the architecture of the Action Plugin, the available action types, and how actions are connected to different hardware and software platforms.

github codes

​

Action Plugin Architecture Class Diagram

In order to simplify the diagram, we only show the most important classes and their relationships.

Core components of the Action Plugin:

​

Data Flow

​

Action Orchestrator

The Action Orchestrator is the central component that orchestrates the execution of actions. It manages the states, promise queue, and threads for each action.

github codes

​

Orchestrator architecture class diagram

In order to simplify the diagram, we only show the most important classes and their relationships.

​

Example configuration

"agent_actions": [
    {
      "name": "move",
      "llm_label": "move",
      "implementation": "passthrough",
      "connector": "ros2"
    },
    {
      "name": "speak",
      "llm_label": "speak",
      "implementation": "passthrough",
      "connector": "ros2"
    }
  ]