Sending a model to field devices (edge) happens on two screens.

  • Deployments (배포) — records which model version went to which device, and when. Rollbacks are recorded here too.
  • Edge Fleet (Edge Fleet) → device → Commands (명령) tab — sends a pull model command to the device, and the device actually fetches the model.

This page is done with an operator or admin account.

1) Create a deployment record

  1. In the sidebar, open Deployments (배포) and press New deployment (새 배포). Enter the model name, version and target device ID, choose a Deployment strategy (배포 전략) (Rolling, Blue-Green, Canary) and press Deployments (배포).
    Trigger a deployment — model, version, target device, strategy
  2. The deployment detail opens. Progress steps (진행 단계) passes through Build → Push → Pull → Health → Activate in order, and the result is kept under Health check (헬스 체크). When the status is ACTIVE, you are done.
    Deployment detail — info, steps and health check

2) Send the model down to the device

  1. In Edge Fleet (Edge Fleet), open the device and go to the Commands (명령) tab. Change the command type ① to Pull model (모델 내려받기), and fields for choosing the model and version appear. Choose them and press Send command (명령 보내기) ②.
    Device Commands tab — ① Pull model with the model and version ② Send command

The command goes into a queue; the device's Edge SDK picks it up at its next poll, downloads the model and reports the result. When the command's status in the list moves Queued (대기 중)Delivered (전달됨)Completed (완료), that version appears as Running (실행 중) in the Local models (로컬 모델) card on the device's Overview (개요) tab.

Next: 9. Monitoring and retraining

Written for the platform as of 2026-09-21.

© Geo-MLOps