DayFlow - Personal Activity Analyzer

You have access to the user's DayFlow database via the

sqlite-dayflow

IMPORTANT: Before executing any queries, you MUST load the MCP tools first:

• Use ToolSearch with query

  +sqlite-dayflow

  or

  sqlite dayflow

  to load the tools
• The main tool you need is

  mcp__sqlite-dayflow__read_query

  for SELECT queries
• NEVER use write/create/delete tools — this database is read-only for us

If the tools are named differently (e.g.

mcp__dayflow__read_query

Database location

~/Library/Application Support/Dayflow/chunks.sqlite

Schema

timeline_cards (main table — activity summaries)

appSites

distractions

WHERE is_deleted = 0

observations (raw AI observations per screenshot batch)

timeline_review_ratings (user's self-assessment)

journal_entries (daily journal — may be empty)

analysis_batches (processing metadata — rarely needed)

Known categories

• 2GIS Works

  — work tasks (meetings, code, Mattermost, Confluence, GitLab)

• Agents & Blog

  — personal AI/agent projects, Claude Code, blogging

• Distraction

  — entertainment, YouTube, series, social media

• Planning, reflections, mindset

  — calendar, planning, journaling

Known subcategories

Calendar

Communication

Development

Documentation

Entertainment

Research

Setup

metadata JSON structure

{
  "appSites": {
    "primary": "mm.2gis.one",
    "secondary": "gitlab.2gis.ru"
  },
  "distractions": [
    {
      "id": "uuid",
      "title": "Distraction title",
      "summary": "What happened",
      "startTime": "2:30 PM",
      "endTime": "2:35 PM"
    }
  ]
}

Key apps/sites:

• mm.2gis.one

  — Mattermost (work chat)

• 2gis.ktalk.ru

  /

  2gis.talk.ru

  — KTalk (work video calls)

• gitlab.2gis.ru

  — GitLab (work code)

• calendar.google.com

  — Google Calendar

• claude.ai

  — Claude web

• telegram.org

  — Telegram

• youtube.com

  — YouTube

How to answer user questions

"What did I do today/yesterday?"

SELECT start, end, title, summary, category
FROM timeline_cards
WHERE day = '2026-02-17' AND is_deleted = 0
ORDER BY start_ts ASC

"How much time did I spend on X category?"

SELECT category, SUM(end_ts - start_ts) / 60 as minutes, COUNT(*) as sessions
FROM timeline_cards
WHERE day = '2026-02-17' AND is_deleted = 0
GROUP BY category
ORDER BY minutes DESC

"What was I doing on project X?" (search by keyword in title/summary)

SELECT day, start, end, title, summary
FROM timeline_cards
WHERE is_deleted = 0
  AND (title LIKE '%товарн%' OR summary LIKE '%товарн%' OR detailed_summary LIKE '%товарн%')
ORDER BY start_ts DESC

"Where did I stop yesterday on work?"

SELECT start, end, title, summary, detailed_summary
FROM timeline_cards
WHERE day = 'YYYY-MM-DD' AND category = '2GIS Works' AND is_deleted = 0
ORDER BY start_ts DESC
LIMIT 3

"What distractions did I have?"

SELECT day, start, end, title, summary,
  json_extract(metadata, '$.distractions') as distractions
FROM timeline_cards
WHERE is_deleted = 0
  AND (category = 'Distraction' OR json_array_length(json_extract(metadata, '$.distractions')) > 0)
  AND day = 'YYYY-MM-DD'
ORDER BY start_ts ASC

"What's my most productive time?"

SELECT
  CASE
    WHEN cast(strftime('%H', start_ts, 'unixepoch', 'localtime') as integer) BETWEEN 6 AND 11 THEN 'Morning (6-12)'
    WHEN cast(strftime('%H', start_ts, 'unixepoch', 'localtime') as integer) BETWEEN 12 AND 17 THEN 'Afternoon (12-18)'
    ELSE 'Evening (18+)'
  END as time_of_day,
  category,
  SUM(end_ts - start_ts) / 60 as minutes
FROM timeline_cards
WHERE is_deleted = 0
GROUP BY time_of_day, category
ORDER BY time_of_day, minutes DESC

"What meetings did I have?"

Look for cards where

subcategory = 'Communication'

detailed_summary

SELECT day, start, end, title, summary
FROM timeline_cards
WHERE is_deleted = 0
  AND (subcategory = 'Communication'
    OR title LIKE '%встреч%' OR title LIKE '%синк%' OR title LIKE '%sync%'
    OR summary LIKE '%звонк%' OR summary LIKE '%call%'
    OR json_extract(metadata, '$.appSites.primary') LIKE '%ktalk%')
ORDER BY start_ts DESC

"Progress on X since date Y"

SELECT day, start, end, title, summary
FROM timeline_cards
WHERE is_deleted = 0
  AND day >= 'YYYY-MM-DD'
  AND (title LIKE '%keyword%' OR summary LIKE '%keyword%')
ORDER BY start_ts ASC

Deeper details — use observations table

When the user needs very detailed info (what exactly was on screen, what code was open, what messages were visible):

SELECT datetime(start_ts, 'unixepoch', 'localtime') as time, observation
FROM observations
WHERE start_ts >= unixepoch('YYYY-MM-DD', 'localtime')
  AND start_ts < unixepoch('YYYY-MM-DD', '+1 day', 'localtime')
  AND observation LIKE '%keyword%'
ORDER BY start_ts ASC

Important notes

1. Always filter

   is_deleted = 0

   on timeline_cards
2. Dates are in "YYYY-MM-DD" format in the

   day

   column
3. Timestamps are Unix seconds (not milliseconds)
4. User speaks Russian — titles and summaries are mostly in Russian, search with Russian keywords
5. For keyword search use

   LIKE '%keyword%'

   on title, summary, and detailed_summary
6. Time zone: the user is in Moscow time (UTC+3). Use

   'localtime'

   modifier in datetime functions
7. Start with timeline_cards for most questions. Only go to observations when you need screen-level detail
8. Present results as a readable narrative, not raw data. Summarize, highlight key activities, calculate totals
9. When user asks about "work" — filter by

   category = '2GIS Works'

10. When user asks about personal projects — filter by

    category = 'Agents & Blog'

11. If user mentions a person (e.g. "Sergey", "Seryozha") — search in summary/detailed_summary with Russian name forms
12. Calculate time in minutes/hours using

    (end_ts - start_ts) / 60

    for minutes
13. The query examples above are templates, not the only options. Use your knowledge of the schema to compose any SQL query that fits the user's question. Combine tables, use JOINs, aggregations, subqueries — whatever is needed