What's wrong with all the others?

Every solution we've seen has at least 2 of these issues.

• Not open source
• Privacy/security issues (they own your data)
• No syncronization/backup of data
  we've not solved this yet, but working on it
• Hard/impossible to extend
  Not built to be extended. Takes a lot of work to collect a little more data
• Low quality of data
  • Limited collection/detail we want to collect everything*
  • Bad resolution to save storage space

Feature comparison in the README

Architecture

• Server
  • Manages all data and supports basic querying
• Clients
  • Watchers
    • A type of client that gathers data and sends it to the server.
    • Examples: Active window, AFK, Web browser
  • Web UI
    • Written with Vue.js, visualizations in D3.js, uses server API for querying
    • Built versions are hosted by the server
  • Custom interfaces/visualization/analysis tools
• Core library datastructures, transforms
• Client library Makes it easy to write clients

Dataformat

• Buckets are containers of events
• Usually one bucket per watcher
• Events are simple JSON

Anatomy of a window event

{
  "timestamp": "2017-05-10T10:26:52Z", # ISO8601 format.
  "duration": 128.12,                  # In seconds.
  "data": {                            # Any JSON object,
      "app": "google-chrome",          # depends on eventtype.
      "title": "ErikBjare on GitHub"
  }
}

Making watcher-writing easy

Heartbeats to the rescue!
idea taken from WakaTime

Merging events with the heartbeat method

e1 = Event(timestamp="...10:26:52", duration=10, ...)
e2 = Event(timestamp="...10:27:12", duration=0, ...)

def merge(e1, e2, pulsetime=10):
  # e1 is the last recorded event
  # e2 is a heartbeat
  if(e1.timestamp + e1.duration + pulsetime <= e2.timestamp \
     and e1.data == e2.data):
    # e2 had identical data and was within pulsetime window
    return [Event(timestamp=e1.timestamp,
                  duration=e2.timestamp - e1.timestamp,
                  data=e1.data)]
  else:
    return [e1, e2]