Why use WEDA?
Building an Edge AI prototype is only the first step. Moving that prototype into production requires you to connect hardware, organize device data, package applications for different computing platforms, deploy them across a fleet, and operate them over time. Each stage is a separate engineering problem, and most of that work has little to do with the model itself.
WEDA provides a common development and management framework for these stages. The sections below follow that progression, from the first sensor you connect to the ongoing operation of a deployed system.
1. Simplifying Hardware Integration
The Challenge
Industrial devices, sensors, I/O modules, and cameras each expose their own drivers, APIs, and protocols. AI application developers typically work in Python or C#, while the interfaces underneath often assume a different integration approach.
WEDA's Approach
- Device Library provides higher-level Python and C# access to Advantech hardware peripherals, so you can read sensors and control I/O without working directly against device-specific C APIs.
- WEDA SubNode helps you integrate peripheral devices and sensors, and handles their connection into the WEDA architecture. You implement the sensor logic; SubNode publishes the resulting telemetry to WEDA Node over local messaging.
2. Unifying Data Across Devices
The Challenge
A production system's data arrives from sensors, cameras, I/O points, and several devices at once. If every source gets its own integration pipeline, the number of pipelines grows with the number of sources, and the system becomes difficult to maintain.
WEDA's Approach
WEDA Core uses the DataStream as a single, unified data layer: each source a device reports becomes a DataStream that you query the same way. The Data Engine builds on that layer, so derived values, alerting, and downstream integration are configured against DataStreams rather than against each individual source.
3. Bridging the Deployment Gap
The Challenge
Different edge hardware platforms expect different operating systems, drivers, GPU or NPU runtimes, and application dependencies. An application that runs in your development environment does not necessarily move unchanged onto every target device.
WEDA's Approach
Ready-to-Dev Containers provide development environments with the toolchain and dependencies already prepared for the target hardware. You build edge applications through a consistent, container-based workflow, and that workflow stays the same as you move between supported platforms.
4. Managing Deployments at Scale
The Challenge
Once the device count grows, configuring, updating, and troubleshooting each device individually stops being practical. Distributed sites and firewall restrictions make routine remote maintenance harder still.
WEDA's Approach
- WEDA Core provides centralized device, container, and AI model management, so a change is defined once and applied to the devices you target.
- WEDA Node runs on the device, applies the configuration it receives, and executes the assigned workloads.
- Remote Access Tunnel opens an on-demand connection to a service on the device's local network when you need direct access, without inbound firewall rules.
5. Operating and Improving Production Systems
The Challenge
A deployed system needs continuous attention: device health has to be observed, anomalies handled, configuration updated, and models maintained. Devices also go offline temporarily, so neither the cloud nor the edge can assume the connection is always available.
WEDA's Approach
- WEDA Node reports device status and telemetry to WEDA Core, which provides centralized monitoring, configuration, and lifecycle management.
- When the connection drops, WEDA Node keeps running its local work. Changes made in the cloud are queued and synchronized once the device reconnects.
- Model training stays with your own toolchain. WEDA takes over at registration, versioning, and deployment, so a retrained model reaches its devices through the same managed path as any other update.
WEDA addresses these challenges by separating centralized orchestration from device-side execution. The next section, Architecture, explains how WEDA Core and WEDA Node work together, which components sit on each side, and where you extend the framework with your own code.