Downsample data to a specific datetime (ie: start of day)

I’m looking for a way to downsample a set of points using a specific timestamp, I’d like to know if that’s possible using a CQ or Kapacitor before using a custom solution (scheduled custom script or something similar).

I need to do so in order to reduce the overall amount of data, here is the plan:

  1. I fetch data every 15sec (including some big texts)
  2. I split the gathered data, and store the text and the key (a hash) in a separate RP “temp”, with a duration of 1h
  3. Using a built-in solution like CQ or Kapacitor
  • get data every few minutes from an RP (“temp”)
  • Write them with a fixed timestamp in a different RP

As time passes, in my “temp” RP I’ll have some data like this

time hash text
2020-10-19 11:00 001 “aaa”
2020-10-19 11:00 002 “bbb”
2020-10-19 11:15 001 “aaa”
2020-10-19 11:15 002 “bbb”
2020-10-19 11:30 001 “aaa”
2020-10-19 11:30 003 “ccc”
2020-10-19 11:45 004 “ddd”
2020-10-19 11:45 005 “eee”
2020-10-19 14:00 001 “aaa”
2020-10-19 14:00 002 “bbb”
2020-10-19 14:15 005 “eee”
2020-10-19 14:15 006 “fff”

I want to achieve a final result which stores all the possible values of those hashes, at the same exact time, in order to don’t have the text repeated every 15sec, but once a day, having the same key (time+hash) will just update the value, allowing me to have less points.

time hash text
2020-10-19 00:00 001 “aaa”
2020-10-19 00:00 002 “bbb”
2020-10-19 00:00 003 “ccc”
2020-10-19 00:00 004 “ddd”
2020-10-19 00:00 005 “eee”
2020-10-19 00:00 006 “fff”

Ideally, this upsert (insert/update) is performed every few minutes (1-5min).

Is this possible using a CQ or Kapacitor (no UDF) ?
For what I’ve found so far, the answer is “no” (sadly)