Input Plugins
Input Plugin Overview
Input plugin overview
Input Plugin Overview
The Input Plugins in OM1 provide the sensory capabilities that allow AI agents to perceive and interact with their environment. These plugins capture, process, and format various types of input data from different sources, making them available to the agent’s runtime core for decision-making.
Input Plugin Architecture
Classes Diagram
In order to simplify the diagram, we only show the most important classes and their relationships.
Data Flow Diagram
Key relationships and notes:
Inheritance Hierarchy
Sensor<T>is the base abstract class that defines the core interfaceFuserInput<T>extendsSensor<T>and implements the polling mechanismVLM_COCO_LocalextendsFuserInput<T>and implements the specific VLM functionality
Key Components
- Uses PyTorch’s
FasterRCNN_MobileNet_V3_Large_320_FPNmodel - Processes images from webcam or other sources
- Detects objects using COCO dataset classes
- Provides spatial awareness (left/right/front)
Key Functionality
- Real-time object detection
- Spatial object localization
- Message buffering and formatting
- Webcam integration