Modzy V1.7 offers significant improvements to the way customers can use, monitor, and manage AI at the edge.
We are thrilled to announce several major improvements to the way customers can use, monitor, and manage AI at the edge with Modzy V1.7. Most significantly, Modzy now offers secure, bi-directional communication between Modzy and edge devices running Modzy Core.
Customers can more seamlessly deploy models to fleets of single board computers (SBCs), 5G MECs, and on-prem servers all while managing them from a single, secure hub.
With the enhanced bi-directional communication tunnel, V1.7 offers customers more reliable, real-time insight into the health and status of ML models running on their edge devices.
V1.7 introduces bi-directional communication between Modzy and nearly any edge device.* With bi-directional communication, customers can now deploy and update ML models on hundreds or thousands of edge devices centrally from Modzy. This also makes it possible to receive live, real-time information about the health and status of models running at the edge. Customers can view details about edge devices, connection status, hardware specs, operating system, chipset, and more. Finally, the introduction of bi-directional communication generates inference code unique to each device and model.
To facilitate bi-directional communication, V1.7 replaces Kafka with NATS as Modzy’s central event streaming platform. NATS is light, fast, and secure, with end-to-end encryption of event queues, and enables better communication with fleets of edge devices. With this change, Modzy's Python SDK now supports bi-directional streaming to and from any models running on Modzy Core, enabling incredibly low latencies for real-time applications. Additionally, the introduction of NATS improves Modzy’s edge device registration process, allowing for more interactive logging and feedback.
Full release note details can be found on our documentation site.
For more information, check out this video from Modzy Co-founder, Seth Clark, and Head of ML Engineering, Brad Munday.
* See this page for specifics on minimum edge device requirements.