Subscribed! Our team builds DCIM, BMS, and EPMS systems for data centers in Korea, so we’re really looking forward to what 2027 brings. We’d love to see continued improvements in Gateway Redundancy and failover behavior, as well as broader support for power-system protocols like IEC 61850. Happy to test the beta and share feedback from the field!
In general, it's not a concern.
The key point is that we are not running the actual LLM inference, nor are your gateways. They are calling out to whatever provider you choose to connect to - most likely, one in a remote cloud (though self-hosted options are available, in which case providing the model and the hardware to run it becomes your responsibility).
The gateway diagnostic assistants are carefully bound to minimize impact on the active runtime while they're collecting diagnostics - this is pretty much the only area where we've seen resource usage that could be a problem at scale, but you're more than welcome to turn off the diagnostic assistants on your production systems and still enjoy the other benefits of the Catalyst modules.
1 Like