Because some wrapper CLAMS apps use fairly large models that take a few minutes to fully load in the memory, if an app user sends POST requests during that initiation time of underlying models/tools, the app would most likely behave badly. This already has happened in https://github.com/clamsproject/app-dbpedia-spotlight-wrapper/issues/3, and we'd like to make a SDK-level helper function to make sure apps don't "die" in such synchronous/asynchronous model initiation scenarios.
Done when
SDK has an implementation of
external server timeout checker
internal (in-memory) model readiness check
and more...?
Additional context
This will be more of a problem soon, as vision-related apps are moving toward ViT-based models.
Because
Because some wrapper CLAMS apps use fairly large models that take a few minutes to fully load in the memory, if an app user sends
POST
requests during that initiation time of underlying models/tools, the app would most likely behave badly. This already has happened in https://github.com/clamsproject/app-dbpedia-spotlight-wrapper/issues/3, and we'd like to make a SDK-level helper function to make sure apps don't "die" in such synchronous/asynchronous model initiation scenarios.Done when
SDK has an implementation of
Additional context
This will be more of a problem soon, as vision-related apps are moving toward ViT-based models.