Computer Usage Summary

Computer Usage Summary

Overview

Use local ActivityWatch data to produce an evidence-based, table-first report of what the user did. Report daily focus, specific activities, local time windows, application usage, and observed foreground browser pages.

This installation records foreground applications, window titles, and AFK status. It does not use browser extensions, cloud sync, browser history, or complete browser URLs.

Core Rules

1. Treat the requested dates as local calendar days.
2. Resolve the Mac system IANA timezone and state it with the UTC offset. Only an explicit --timezone value may override the system timezone.
3. Convert every displayed timestamp to that timezone. Do not display raw UTC timestamps in a human report.
4. Merge duplicate and overlapping AFK intervals before calculating time.
5. Keep each day's active, away, and coverage values at or below 24 hours, or below the shorter local day during a daylight-saving transition.
6. Report untracked time separately. Do not relabel missing coverage as away time.
7. Treat foreground_sessions as observed foreground intervals, not process launches.
8. Treat browser rows as observed foreground pages, not a list of every open browser tab.

Workflow

1. Establish the inclusive local date range and state the timezone.
2. Run the bundled script:

python3 scripts/activitywatch_summary.py --date today

The default command returns a complete copyable Markdown report.

For a multi-day report:

python3 scripts/activitywatch_summary.py --start YYYY-MM-DD --end YYYY-MM-DD

For structured model analysis, request JSON explicitly:

python3 scripts/activitywatch_summary.py --start YYYY-MM-DD --end YYYY-MM-DD --format json

For copyable or spreadsheet-ready tables:

python3 scripts/activitywatch_summary.py --start YYYY-MM-DD --end YYYY-MM-DD --format markdown --table daily
python3 scripts/activitywatch_summary.py --start YYYY-MM-DD --end YYYY-MM-DD --format tsv --table apps
python3 scripts/activitywatch_summary.py --start YYYY-MM-DD --end YYYY-MM-DD --format tsv --table browser
python3 scripts/activitywatch_summary.py --start YYYY-MM-DD --end YYYY-MM-DD --format tsv --table timeline

Use --timezone Area/City only when the user explicitly requests a timezone different from the Mac system timezone.

Use --min-tab-seconds 120 for a compact browser table. The default value is 0, which includes every normalized foreground browser page.

3. Use JSON data to reconstruct daily activities from application names and sanitized titles.
4. Mark an activity confirmed when a title directly supports it. Otherwise mark it unknown context.
5. Keep exact time ranges and accumulated foreground duration distinct:
   • first_seen to last_active is an observation span and may contain gaps.
   • active_duration is accumulated confirmed foreground time.
6. If ActivityWatch is unavailable or empty, use the fallback workflow and label unavailable metrics.

Required Output Contract

Return a copyable Markdown report in this order.

1. Scope

State:

• Inclusive local dates
• IANA timezone and UTC offset
• Evidence source
• Coverage limitations

2. Daily Overview Table

Include one row per day with:

Include a total row for multi-day reports.

3. Daily Activity Timeline

Include one copyable table per day with:

Activities may overlap when the user switched repeatedly between related applications. Do not describe first-to-last spans as continuous focus.

4. Application Table

Include application, foreground sessions, numeric active seconds, HH:MM:SS duration, first seen, last active, and source.

For a multi-day report, show the leading applications per day. Include all applications when the user asks for a full export.

5. Browser Foreground Page Table

Include:

Group repeated observations by browser and normalized title. Remove browser-profile suffixes, memory-usage suffixes, invisible formatting characters, and duplicate watcher heartbeats.

For a readable default chat report, list pages with at least two minutes of accumulated foreground time and add a short-visit table containing the omitted page count and total duration. If the user asks for every page or every tab, list all rows with no duration threshold.

Explain that ActivityWatch cannot identify background tabs, tabs never brought to the foreground, or two same-title tabs as separate objects.