You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
So from the UI(console) the user can set up different parameters to watch out for anomaly
eg: to train a model for z-score, the user could pass in a JSON like the following:
{"start_time": "2023-09-20T08:41:11.398Z",// training interval start timestamp"end_time": "2023-09-20T13:41:11.398Z",// // training interval end timestamp"stream": "frontend","interval_window": '1m',// string with number and a unit of (h,m,s)"algorithm": "z-score"// the model to be used}
We could save this in parseable server and then an anomaly detection server or plugin could just get this from parseable whenever it requires (eg: say when it restarts).
Hence, an API for config eg: /api/v1/anomaly/config (both GET and PUT methods to get and update configs) would be great.
As soon as the config is set, parseable can then call the anomaly detection server to start training based on the this new config (or some other way based on a good architecture)
We could also have another api to list out models/algorithms etc in future maybe.
P.S. This plan is in early stages and we can ideate and discuss on better solutions too.
The text was updated successfully, but these errors were encountered:
So from the UI(console) the user can set up different parameters to watch out for anomaly
eg: to train a model for z-score, the user could pass in a JSON like the following:
We could save this in parseable server and then an anomaly detection server or plugin could just get this from parseable whenever it requires (eg: say when it restarts).
Hence, an API for config eg: /api/v1/anomaly/config (both GET and PUT methods to get and update configs) would be great.
As soon as the config is set, parseable can then call the anomaly detection server to start training based on the this new config (or some other way based on a good architecture)
We could also have another api to list out models/algorithms etc in future maybe.
P.S. This plan is in early stages and we can ideate and discuss on better solutions too.
The text was updated successfully, but these errors were encountered: