GitHub - polyrabbit/hacker-news-digest: :newspaper: Let LLM Summarize Hacker News for You · GitHub
Skip to content

Repository files navigation

Github Pages license PRs Welcome Hacker News

中文版

Hacker News Summary leverages AI technology to extract summaries and illustrations from Hacker News articles, providing a seamless news scanning experience.

Summaries are primarily generated by a remote LLM service, with local Qwen as fallback when remote inference is unavailable.

Features

  • Clear and easily understandable summaries generated by our advanced AI assistant
  • Relevant illustrations make articles easily scannable and visually engaging
  • Common video sites, PDFs, and GitHub gists are seamlessly embedded
  • Flexibility to sort articles based on their points, comment count, or publication time
  • Filter the topN articles based on their points.
  • RSS feeds fully supported (#14, #19)
  • Local translation (Chinese)

Talk is cheap, show me the screenshot!

hn-summary

Emoji explained:

  • ❤️: point - upvotes received from the Hacker News community
  • 👤: user - Hacker News user who submitted this post
  • 🕘: submission time - a human-readable time indicating when the post was submitted
  • 💬: comment count - comments posted by the community, click to visit this comment page
  • 🔗: source of the news - where the news originated
  • 📰: summary model - the model used to generate the summary

How it works

Hacker News Summary is a static site hosted on GitHub Pages. It performs the following periodic actions:

  1. Parsing the Hacker News page to obtain a list of news articles
  2. Extracting the main content from each news article using a score algorithm
  3. Finding the most suitable illustration for each article and making a local copy
  4. Generating summaries of the article's content using OpenAI API or invoking a local model as a fallback when the API is unavailable
  5. Rendering a template that incorporates the illustrations and summaries, and deploying it to GitHub Pages

Localization

Translation is also performed by LLM, with a single extra step in the prompt. Currently supported languages:

TODO

  • A better way to scrap websites (maybe PhantomJS & Selenium)
  • Also summarize comments (see discussions on Hacker News)
  • Switch to Hacker News API
  • A more beautiful home page (maybe in HTML9)
  • Discover an alternative local models for generating summaries
  • Sort articles by points/comments/time
  • Filter topN articles by points
  • RSS
  • Deploy on github pages
  • Have a good sleep !important