Screenpipe daily summary pipe - runs at 6pm every day

gistfile0.txt

---
schedule: "0 18 * * *"
enabled: true
title: Daily Summary
description: "Comprehensive daily summary of your screen activity, delivered at 6pm"
icon: "📋"
---

Analyze my screen and audio recordings from today. Query the screenpipe search API in 2-hour chunks across the workday (8am-6pm).

Make sure to use local timezone.

## Search API

```
GET http://localhost:3030/search?content_type=ocr&start_time=<ISO8601>&end_time=<ISO8601>&limit=200&min_length=50
```

Content types: `ocr` (screen text), `audio` (speech), `input` (clicks/keystrokes/clipboard/app switches), `accessibility` (UI tree), `all`.

Usually accessibility has most of the screen data.

Extra params: `q` (keyword), `app_name`, `speaker_name`, `offset` (pagination).

Query each modality separately per chunk for richer results:
1. `content_type=accessibility&min_length=50` — what was on screen
2. `content_type=audio` — what was said
3. `content_type=input` — what was typed, clicked, copied

## Task

1. Query the search API in 2-hour chunks from 8am to now
2. Use `min_length=50` to skip noisy/short fragments
3. Synthesize activities into a structured daily report
4. Extract action items, key decisions, and follow-ups

## Output Format

```markdown
# Daily Summary — <date>

## One-Liner
What I mainly did today in one sentence.

## Timeline

| Time | Activity | Apps | 
|------|----------|------|
| 9:00-10:00 | Reviewed PR #123 for auth module | GitHub, VSCode |
| 10:00-11:30 | Call with team about roadmap | Zoom, Notion |

## Accomplishments
- Top things I finished today, with timestamps
- Name specific apps, files, PRs, or projects
- Link to screenpipe timeline: [10:30 AM](screenpipe://timeline?timestamp=2025-02-01T10:30:00Z)

## Key Moments
- Important things I saw, said, or heard — with timestamps and deep links
- Decisions made, ideas discussed, feedback received

## Meetings & Conversations
- Summarize any meetings or calls (participants, duration, outcomes)
- Key quotes or decisions from audio transcriptions

## Action Items
- [ ] Tasks mentioned or started but not finished
- [ ] Follow-ups from meetings or messages
- [ ] Deadlines mentioned

## Unfinished Work
- What I should continue tomorrow — name the app, file, and specific task

## Patterns
- Apps I used most and approximate time per app
- Topics/projects that came up repeatedly
- Communication patterns (who I talked to, which channels)

## Summary
2-3 sentence overview of the day. End with: "**Next step:** [most important thing to continue tomorrow]"
```

## Deep Links

- Timeline: `[10:30 AM](screenpipe://timeline?timestamp=2025-02-01T10:30:00Z)`
- Frame: `[screenshot](screenpipe://frame/<frame_id>)` using frame_id from results

## Rules

- Only report what you can verify from the data
- Keep summaries concise, group related activities
- Skip idle periods or duplicates
- Use the user's local timezone for all displayed times
- Never dump raw OCR — synthesize into activity descriptions
- Redact passwords, API keys, tokens, credentials
- Skip banking/financial/medical content — note as "private activity"
- Summarize conversations, don't paste full transcripts
- Keep track of things I could be more productive on