You’re right, and API for managing baseline conditions isn’t available yet. It’s still in the backlog. One of the challenges with baseline conditions that other condition type don’t have is the sensitivity setting. Unlike others in the space with a similar feature they use standard deviations as their control over the confidence band. We don’t don’t do that as it would inhibit and restrict quality of the feature and what we are able to do with it in the future.
Instead the sensitivity is computed based on the quality of our one-minute-ahead predictions, which varies over different data series. The prediction quality is based not on our ability to predict the actual data series, but on our ability to predict a smoothed version of the data series, to lessen the effects of outliers. This smoothing is complicated, it can evolve over time and also data series specific.
Currently we allow users to bypass these complications by allowing the user to see a preview of the baseline and confidence band and set the sensitivity using a UI that tells them an estimate of how many violations would have occurred with those settings over the last week. This human interpretation of a 7 day preview to set the right sensitivity is the challenge to replicate in an API.
How can we allow customers to set a meaningful sensitivity when creating one of these conditions via API?
One thought is that instead of setting the sensitivity directly the API input is the number of violations in the last 7days you would have liked to have triggered and we’ll choose the sensitivity that most closely gets the condition to achieving that.
If you have any feedback or thoughts on this approach please let me know. Also, know that vision for Alerts is to eventually have everything you can do in the UI available via API in a clean well documented way.