Architectural Query: Centralized AI Processing with Distributed Edge Recording in Nx Witness

Answered

mery

November 02, 2026 04:47 Edited

We are currently in the planning and design phase for a large-scale, multi-site video surveillance deployment using Nx Witness VMS. Before finalizing our hardware Oklahoma Toll Road Map  procurement and infrastructure layout, we would like to validate whether our proposed hybrid topology is officially supported and how best to configure it.   

0

• 

  Anderson Chang

  • Network Optix team

  November 02, 2026 07:34

  Hi Mery,

  Yes, you can record on the edge server and perform the AI processing on a central server with Nx Witness.
  However, the actual approach depends on which AI tool or software you would like to use for processing the footage.
  Could you please share more information about it?

  Since Nx Witness supports RTSP re-streaming and offers many server APIs, including writing analytics data, there are several possible approaches.

  Here is one example, which is also the simplest approach: retrieve the RTSP stream from the edge server for AI processing, then write the metadata back to Nx Witness using the Create Bookmark API.

  Camera
    │ RTSP stream
    ▼
  Edge Server
    │ RTSP re-stream
    ▼
  Central Server AI Processing
    │ Create Bookmark API
    │ POST /rest/v4/devices/{deviceId}/bookmarks
    ▼
  Edge Server (or any Nx Server within the same Nx Site)
    Write the bookmark into database

   

  There are several APIs that can be used to write analytics metadata to Nx Witness, depending on your needs.
  Please refer to our API Documentation for further information.

  • Local version: https://<server ip>:7001/#/api-tool/ (aligning API version to the server instance)
  • Online version: https://nxvms.com/doc/developers/api-tool

   

  Please provide further information regarding your architecture, including the tools you plan to use and the kind of data that needs to be recorded, so that we can better understand your plan and environment, and provide the most suitable suggestion.

   

  Thank you.

  0

Please sign in to leave a comment.