Hackerbrief

Daily Hacker News front-page digest. Summaries of linked articles and comments.

How It Works

1. .github/workflows/daily-digest.yml runs every day at 00:00 UTC
2. The workflow runs pnpm generate, which executes scripts/generate-digest.ts.
3. The script:
   • Fetches top story IDs from https://hacker-news.firebaseio.com/v0/topstories.json
   • For each story: fetches the linked article, runs it through @mozilla/readability to extract main text.
   • Calls Gemini (gemini-2.5-flash) to summarize the article. Stories flagged as primarily partisan/political content are skipped.
   • Fetches the comment tree from the Algolia HN API and asks Gemini to summarize
4. The output is written to digests/hn-digest-MM-DD-YYYY.json.
5. The workflow commits and pushes the new digest file to main.
6. The frontend (src/components/digest-view.tsx) fetches today's digest from raw.githubusercontent.com/P0u4a/hackerbrief/main/digests/, falling back to yesterday's if today's hasn't been generated yet. Results are cached in localStorage.

Environment Variables

• GEMINI_API_KEY - Gemini API key (also accepts GOOGLE_GENERATIVE_AI_API_KEY).

Local Development

Run the frontend:

pnpm dev

Tuning

Constants live in scripts/constants.ts:

• TARGET_ITEMS - number of accepted stories per digest.
• FETCH_BUFFER - extra candidates pulled to survive skipped stories.
• MAX_COMMENTS - cap on comments summarized per story.
• GEMINI_DELAY_MS - throttle applied after each Gemini call.
• ARTICLE_FETCH_TIMEOUT_MS, MAX_ARTICLE_CHARS, MAX_COMMENT_CHARS - request and prompt-size limits.