github-trending

Before starting: First say "开始抓 GitHub 榜单 🐙" and state today's date and the scraping period (defaults to the weekly list).

GitHub Trending scraping

Scrape GitHub Trending → score and mark with keywords → write an Obsidian note. Pure Python script, zero tokens.

Step 0: Read the shared configuration

The configuration is in ../_shared/user-config.json, and the personal override is in ../_shared/user-config.local.json(if it exists, local takes precedence). Inside the script it is already loaded automatically via ../_shared/user_config.py , so there is no need to pass parameters manually. Relevant fields:

• paths.obsidian_vault — vault root path
• paths.github_trending_folder — output folder (default GitHubTrending）
• daily_papers.keywords / negative_keywords / domain_boost_keywords — reuse them to score projects and mark relevance

Parse the period and language

Parse from the user input --since：

• "周榜", "过去一周", "这周", "本周" → --since weekly(default)
• "日榜", "今天", "今日" → --since daily
• "月榜", "这个月", "本月" → --since monthly

Optional --language(e.g. "只看 Python" → --language python); if not specified, scrape all languages.

Workflow

Phase 1: Scrape + score (fetch_trending.py)

python3 fetch_trending.py --since weekly > /tmp/github_trending.json
# 指定语言示例：
# python3 fetch_trending.py --since weekly --language python > /tmp/github_trending.json

The script automatically completes:

• Scrape github.com/trending?since=<周期>(compatible with SSL interception environments, automatically falls back to not verifying certificates)
• regex parsing: repo, description, language, cumulative star/fork, stars added in this period
• Score with the shared configuration keywords, mark relevant(matches positive/domain words and is not outweighed by negative words),do not discard non-related items
• Sort in descending order by stars added in this period

Check the output: Confirm that /tmp/github_trending.json is a non-empty JSON array. If it is empty, look at the stderr diagnostics.

Phase 2: Write the Obsidian note (write_trending_note.py)

python3 write_trending_note.py /tmp/github_trending.json

• Output to {vault}/{github_trending_folder}/
• The weekly list filename is based on the ISO week number, e.g. 2026-W28 GitHub周榜.md; the daily/monthly lists use the date
• Note structure: summary → "🎯 与研究方向相关" list → "📊 完整榜单" table (including ✅ relevance markers)
• The script prints the final note path to stdout

Output

After completion, inform the user:

• How many projects were scraped, and how many of them are related to the research direction
• The path of the generated note

Notes

• Neither script starts a Task Agent, so there is zero token consumption
• Do not perform git operations(consistent with daily-papers, git automation is off by default)
• When the network is unavailable, refer to the proxy rules in the global CLAUDE.md (port 17890); the script already has an SSL fallback built in
• GitHub Trending only has three tiers, daily/weekly/monthly, and cannot be customized to any number of days