"Absorb" category type similar to ManicTime's

[bgoscinski]

I'm new user of ActivityWatch migrating from ManicTime but I can't find anything that would resemble ManicTime's "Absorb" autotag type.

tldr: It's a way to categorize some activities based on activities that "surround" it.

For example, consider having 2 categories:

• "programming" that matches anything that happens in my IDE
• "googling" that matches visits to google.com

Suppose that I'm working in my IDE, then jump to my web browser to do a quick search on whatever topic and then go back to my IDE. In this case chances are that the googling was related to my programming work so it would be nice to absorb it into surrounding category. Marking a category as absorb type would allow AW to do just that - in this case the quick googling would contribute to my "programming" category.

Is such feature available today in AW?
Is it something that maintainers are interested in supporting?

[ErikBjare]

This very similar to the concept of (heartbeat/event) "flooding" in ActivityWatch, which we use to estimate what happens in gaps that occur due to the nature of polling.

If you apply the event flooding algorithm to say your "Programming" events with a "floodtime"/"breaktime"/"pulsetime" of 5min then any gap less than that would still count as coding. It isn't perfect though, if you do 1s of "programming" every 4min and 55s, then it would count ~5min of coding despite there just being 2 seconds of actual programming done.

I've used this when people use ActivityWatch to invoice me for work done, where we first filter out the work-related events with categorization and then apply flooding with 5min breaktime (usually).

Not quite how the absorb feature works I imagine given your description about "marking a category as absorb type", but it seems pretty close and maybe the 80/20 solution if we were to implement something like it.

[bgoscinski]

Thanks for pointing me in "event flooding" direction. I'll have something to read up on :)