ContextQMD
Back to Eliza

Vision Plugin

packages/docs/plugin-registry/vision.md

2.0.11.3 KB
Original Source

The Vision plugin gives Eliza agents the ability to understand and analyze images, enabling visual reasoning in conversations.

Package: @elizaos/plugin-vision

Installation

bash
eliza plugins install @elizaos/plugin-vision

Enable via Features

json
{
  "features": {
    "vision": true
  }
}

Note: This plugin requires the @tensorflow/tfjs-node native addon. On systems without native build tools, set ELIZA_NO_VISION_DEPS=1 to skip installation of optional vision dependencies.

Configuration

ParameterTypeDefaultDescription
CAMERA_NAMEstring—Camera name to search for (lowercase partial match)
PIXEL_CHANGE_THRESHOLDnumber50Percentage of pixels that must change to trigger a VLM update

Set these in your plugin config or as environment variables.

Features

  • Image understanding and description
  • Visual analysis of screenshots and photos
  • Camera integration with change detection
  • Feature-gated — only loaded when explicitly enabled