DayTrace

Automatically collects daily activities from Linear, GitHub, Slack, and Claude Code, generates a work-focused summary using Claude Agent SDK (Haiku), and saves it as an Obsidian markdown file.

How It Works

Collectors (Linear, GitHub, Slack, Claude Code)
    ↓ Gather raw activity data (read-only APIs)
Claude Agent SDK (Haiku)
    ↓ Filter work-related activities + generate narrative summary
Obsidian Vault
    ↓ Save as YYYY-MM-DD.md (with YAML frontmatter, Dataview compatible)

Runs automatically via macOS launchd on Tue–Sat at 7:00 AM, recording the previous day's activities.

Output Example

---
date: '2026-03-03'
tracker: true
linear_completed: 6
slack_messages: 10
claude_sessions: 32
---
# 2026-03-03 (Tue)

## Linear

### Completed

- **AI-2300: HWP/HWPX File Processing Skill**
  - Implemented HWP/HWPX file parsing and text extraction
  - Completed full investigation of file processing mechanisms

## Slack

### onprem-tepsco

- **Gemma3 12B Model Verification**
  - Verified checksum in Runpod environment
  - Compared against version 0.10.2

## Claude Code

### mally (8 sessions)

- **Dev Workflow Pipeline Design**
  - Worktree creation → planning → Linear ticket → branch → implementation

Installation

git clone https://github.com/jeongsangmin/daytrace.git
cd daytrace
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

Configuration

cp .env.example .env

Edit .env with your tokens and paths:

Variable	Description	Required
LINEAR_API_KEY	Linear → Profile icon → Account → API → Personal API key	Optional
GITHUB_TOKEN	GitHub → Settings → Developer settings → Personal access tokens	Optional
GITHUB_USERNAME	GitHub username	Optional
SLACK_USER_TOKEN	Slack App User OAuth Token (search:read scope)	Optional
OBSIDIAN_VAULT_PATH	Absolute path to your Obsidian Vault	Required
OBSIDIAN_DAYTRACE_DIR	Folder name inside Vault (default: DayTrace)	Optional
CLAUDE_DIR	Claude config directory (default: ~/.claude)	Optional

Collectors with missing tokens are automatically skipped.

Slack App Setup

1. Go to https://api.slack.com/apps → Create New App
2. OAuth & Permissions → Add search:read to User Token Scopes
3. Install to Workspace → Copy User OAuth Token (xoxp-...)

Claude Code Filtering

Add personal project names to PERSONAL_KEYWORDS in collectors/claude.py to exclude them from the summary.

Customizing the Summary Prompt

Edit prompts/daily_summary.md to change the summary tone, format, or filtering rules.

Usage

source .venv/bin/activate

# Record yesterday's activities
python main.py

# Record a specific date
python main.py --date 2026-03-03

# Test individual collectors
python -m collectors.linear 2026-03-04
python -m collectors.slack 2026-03-04
python -m collectors.claude 2026-03-04

Automation (macOS)

chmod +x setup_launchd.sh
./setup_launchd.sh

Auto-wake Mac

sudo pmset repeat wake MTWRF 07:00:00

Uses dark wake — the display stays off, only the system wakes.

Full Flow

Time	Event
07:00	pmset dark-wakes the Mac
07:00	launchd runs caffeinate -i python main.py
~07:01	Collect → Summarize → Save complete
~07:02	caffeinate releases → system sleeps automatically

Backfill

.last_run.json tracks the last successful run date. Missed days are automatically backfilled on the next run.

Project Structure

daytrace/
├── main.py                 # Entry point (backfill + collect + summarize + write)
├── config.py               # .env loading + settings
├── summarizer.py           # Claude Agent SDK summary generation
├── writer.py               # Obsidian markdown generation & file writing
├── collectors/
│   ├── linear.py           # Linear GraphQL API (read-only)
│   ├── github.py           # GitHub Events API (read-only)
│   ├── slack.py            # Slack search.messages (read-only, public channels only)
│   └── claude.py           # ~/.claude JSONL parsing (work projects only)
├── prompts/
│   └── daily_summary.md    # Summary prompt guide
├── setup_launchd.sh        # launchd plist generator & loader
├── .env.example            # Token template
├── .gitignore
└── requirements.txt

License

MIT