We may only stream alerts that are above a r/b threshold, or we may stream all of them. (Measurements objects only get created for sources that pass prelminary cutouts in any event.)
Things to do:
Determine alert schema
Determine policy for what's included (incl. forced photometry)
Write code to stream alerts as part of the pipeline. This could happen either as measurement objects are created (which is how lensgrinder does it), or it could be done as another step after the measurements objects have been saved to the database. (I prefer the latter; that way, we know we've saved everything that we send out alerts for. We don't want to send alerts, have the pipeline crash, and then have no databse records saved from alerts we've sent out.)
Write a very thin alerts table that just has a link to the meausrements_id that triggered the alert, a sent timestamp, the server and topic to which the alert was sent (perhaps), and a provenance. Provenance should include all the parameters that decide the policy of what goes into the alerts (e.g. r/b cutoff if any, how many days of past detections and past forced photometry of the same object, image cutouts, etc.).
Alerts will be triggered by measurements objects.
We may only stream alerts that are above a r/b threshold, or we may stream all of them. (Measurements objects only get created for sources that pass prelminary cutouts in any event.)
Things to do: