Architecture

Wedge has three main components:

  • The Wedge App provides the ability to run AI workloads at the edge, as determined by the user through the App Control UI. By using the Wedge AI SDK, you can split AI models based on node parameters and network layout, and then use the App Control UI deploy the split models across multiple devices.
  • The App Sync Server runs locally on an edge device of the site and manages the nodes running the Wedge App, sending them information such as which AI model to run. Every site where you deploy the Wedge will have a sync server deployed.
  • The App Control UI is a web service running in the cloud. It gives you an overview of your sites and nodes that are running the Wedge App, allowing you to remotely set the video input for the app, the model which should be ran, where inferences should be written to, and define various pre- and postprocessing parameters.

The sync servers periodically fetch the configuration of the connected Wedge apps of their site from the App Control UI and distribute this configuration to the applications.

The Wedge App does not require any local user intervention, it starts automatically when the device is powered up, and connects to the local sync server to get the desired AI workload parameters.

Wedge overview Wedge overview