logblog

A Claude Code plugin that turns your Chrome browsing history into polished tech blog posts on your Hugo site.

Overview

logblog reads your Chrome browsing history, classifies URLs by content type (YouTube, GitHub, docs, web pages), fetches enriched content from each source, and uses Claude AI to write and publish a deep-dive blog post to your Hugo blog — all from a single slash command.

Skills

/logblog:post

Creates a tech blog post from your recent browsing history.

What it does:

1. Extracts Chrome browsing history and Claude Code sessions
2. Classifies each URL by type (YouTube, GitHub repo/PR/issue, docs, web page, AI chats)
3. Shows you the classified list and asks for approval
4. Fetches enriched content — transcripts for YouTube, structured data for GitHub, full text for web pages, deep-crawled docs via Firecrawl
5. Writes a deep-dive technical blog post in English, then translates to Korean (bilingual workflow)
6. Generates a cover image and publishes both versions to your Hugo blog repo

Usage:

/logblog:post

Features:

• Automatic URL classification via regex dispatch table
• YouTube transcript extraction (Korean → English → any language fallback)
• GitHub and Bitbucket repo/PR/issue enrichment via gh CLI and REST API
• AI chat fetching from ChatGPT, Claude, Gemini, and Perplexity via Chrome CDP
• Deep docs crawling via Firecrawl — fetches sub-pages for guide-style coverage
• Dev log mode — turns Claude Code sessions into narrative development posts
• Bilingual publishing — writes in English, translates to Korean, publishes both
• Concurrent page fetching with Playwright
• Separate posts per topic (not combined daily logs)
• Cover image generation with PIL gradients
• Hugo frontmatter with tags, categories, and cover image

/logblog:kakao-post

Creates tech blog posts from URLs people have shared in a user-selected subset of KakaoTalk open chats. The chat is the data source — never the subject of the post.

What it does:

1. Lists open chats with unread messages (uses kakao-chat CLI)
2. You pick which chats to mine (1-5 typical, never bulk-publish all)
3. Extracts the unread URLs and filters them by source tier (GitHub repos, arxiv, vendor blogs, etc.)
4. Shows you ~20 candidates and lets you pick 5-15 to write up, with optional groupings (e.g., bundle 5 same-day vendor announcements into one digest post)
5. Multi-agent dispatch for batch writing — one agent per post in parallel
6. Writes Korean+English posts with inline links throughout, mermaid diagrams, and a categorized References section
7. Publishes to your Hugo blog the same way /logblog:post does

Strict source-hiding rule: Published posts never mention KakaoTalk, 카카오톡, 채팅방, 오픈채팅, open chat, or chat threads. The blog reads as if you encountered each announcement directly. The skill includes pre- and post-publish self-check greps to enforce this.

Usage:

/logblog:kakao-post

Prerequisites:

• Run /logblog:kakao-setup once on a new machine to install kakaocli, derive the KAKAOCLI_KEY, and install the wrapper
• config.yaml already configured (run /logblog:setup if not)

/logblog:kakao-setup

One-shot installer for everything /logblog:kakao-post depends on. macOS only.

What it does:

1. Diagnoses environment — Homebrew, full Xcode (not just Command Line Tools), KakaoTalk app, Full Disk Access
2. Installs kakaocli via brew install silver-flight-group/tap/kakaocli
3. Clones the kakaotalk-chat-analyzer wrapper and adds a kakao-chat() shell function so the wrapper is on PATH
4. Derives the KAKAOCLI_KEY (SQLCipher decryption key) — tries kakaocli auth first, falls back to a SHA-512 brute-force of the User ID when auto-auth times out (handles User IDs above ~22M)
5. Persists the key to your ~/.zshrc (or ~/.bashrc)
6. Runs three smoke tests end-to-end (kakaocli chats → kakao-chat unread)

Usage:

/logblog:kakao-setup

Run this once per machine. After it succeeds, /logblog:kakao-post works without further setup.

/logblog:setup

Sets up a GitHub Pages blog with Hugo and the Stack theme from scratch.

What it does:

1. Installs Hugo and initializes a new site (or connects to an existing one)
2. Configures the Stack theme with your preferences
3. Sets up GitHub Pages deployment via GitHub Actions
4. Creates config.yaml to link log-blog to your blog repo
5. Verifies the full pipeline works end-to-end

Usage:

/logblog:setup

Features:

• Zero-to-published blog in one command
• Hugo Stack theme configuration
• GitHub Actions deployment setup
• Multi-account SSH configuration guidance
• Automatic config.yaml generation

Installation

In Claude Code, open the plugin menu and add the marketplace, then install:

/plugin  →  Marketplaces  →  Add  →  ice-ice-bear/log-blog
/plugin  →  Discover  →  logblog  →  Install for you (user scope)

After installation, run /reload-plugins to activate. Verify with:

/logblog:post    # Create a blog post from browsing history
/logblog:setup   # Set up a new Hugo blog

To update after new releases:

/plugin  →  Installed  →  logblog  →  Update now

Python Dependencies

# Requires Python 3.12+ and uv
uv sync
uv run playwright install chromium

Configuration

Copy and edit the config file:

cp config.example.yaml config.yaml

chrome:
  google_accounts: ["you@gmail.com"]
  history_db_base: "~/Library/Application Support/Google/Chrome"

time_range_hours: 24

blog:
  repo_path: "~/path/to/your-blog-repo"
  repo_url: "https://github.com/you/your-blog.git"
  content_dir: "content/posts"
  language_content_dirs:
    ko: "content/ko/posts"
    en: "content/en/posts"
  default_language: "en"
  language: "auto"  # "auto", "ko", "en"

playwright:
  headless: true
  timeout_ms: 15000
  max_concurrent: 5
  cdp_port: 9222

# Optional: deep docs crawling via Firecrawl
firecrawl:
  api_key: "${FIRECRAWL_API_KEY}"
  max_pages: 10

CLI Usage

logblog also provides a standalone CLI for use outside Claude Code:

# Extract browsing history
uv run log-blog extract              # Pretty table of last 24h
uv run log-blog extract --hours 48   # Override time range
uv run log-blog extract --json       # JSON output

# Fetch page content
uv run log-blog fetch https://example.com --json
uv run log-blog fetch --deep https://docs.example.com/guide --json  # Deep-crawl docs via Firecrawl

# Dev log from Claude Code sessions
uv run log-blog sessions --list      # List recent coding projects
uv run log-blog sessions --project my-app --all --json  # Detailed session data

# Publish a markdown post
uv run log-blog publish post.md --push
uv run log-blog publish post.md --language ko  # Publish to Korean content dir

# Launch Chrome with CDP for authenticated AI chat fetching
uv run log-blog chrome-cdp

# Import AI chat exports (offline)
uv run log-blog import-ai ~/Downloads/conversations.json --json

Requirements

Requirement	Purpose
Python 3.12+	Runtime
uv	Package manager
Chromium (via Playwright)	Fetching web page content
gh CLI (authenticated)	Fetching GitHub repo/PR/issue data
Hugo blog repo	Publish destination
Firecrawl API key (optional)	Deep docs crawling

Troubleshooting

Issue: gh commands fail with authentication errors Solution: Run gh auth login to authenticate the GitHub CLI

Issue: Playwright times out on certain pages Solution: Increase playwright.timeout_ms in config.yaml or reduce max_concurrent

Issue: Chrome history database is locked Solution: Close Chrome before running extract, or use a different profile

Issue: Blog push fails with permission denied Solution: Verify SSH keys are configured for the correct GitHub account. See the multi-account SSH setup guide if using multiple accounts

Tips

• Run /logblog:post at the end of a work session to capture what you learned
• Review the classified URL list carefully — remove irrelevant entries before fetching
• Use --hours 48 in the CLI to capture a weekend session
• Each topic gets its own post with a unique filename to avoid overwriting

Author

ice-ice-bear

Version

0.3.0

License

MIT