AI Model Management
Overview
AI Model Management lets you upload, version, and remotely deploy AI model files to WEDA Node devices — without manual file transfers or shell access to the edge. WEDA Core handles upload integrity verification and secure delivery to the device. Depending on the configured deployment action, WEDA can replace the model file, reload a service, or restart the target inference container.
The feature is built around three concepts:
| Concept | What it is |
|---|---|
| Model | A named container for a family of model files (e.g. defect-detector) |
| Edition | A specific version of a model's binary (e.g. 1.0.3). After the file is uploaded and verified, its content cannot be replaced, but its descriptive metadata can still be updated. |
| Deployment | An edition pushed to one or more target devices, tracked per device |
The Full Picture
The flow has two phases:
- Upload — register the model and an edition, upload the binary via TUS, and wait for hash verification. Covered in Upload and Deploy.
- Deploy — trigger deployment to target devices. WEDA Core delivers the verified file to each device and runs the configured deployment action, which can replace the file, reload a service, or restart a container. Also covered in Upload and Deploy.
Getting the deployed file into your inference container's runtime is a separate, one-time setup step on the container side — see Make Your Container Read the Model.
Next Steps
| Guide | What it covers |
|---|---|
| Upload and Deploy | Register a model, upload an edition, deploy it to devices, and check deployment status |
| Make Your Container Read the Model | Mount the secured volume so your inference container can read the deployed file |
Related
- Deploy the container that runs your inference workload: Container Stack
- Track the target device: Device Management
Last updated on Jul-16, 2026 | Version 1.1.0