Skip to content

README.md

Latest commit

 

History

History
55 lines (41 loc) · 1.58 KB

README.md

File metadata and controls

55 lines (41 loc) · 1.58 KB
Raw

hacker_news extension

Fetches top Hacker News stories that match configured AI/ML keywords and exceed a minimum score threshold, then summarises each one with an LLM.

Pipeline

fetch()    — scrapes HN top stories, filters by score + keyword match
process()  — LLM one-sentence summary per story
render()   — wraps in FeedSection

Config

config/sources.yaml — toggle only:

Key Default Notes
enabled true

config/extensions/hacker_news.yaml — all other settings (see hacker_news.yaml.example to restore defaults):

Key Default Notes
min_score 50 Minimum HN points to include
max_items 20 Maximum stories in the digest
keywords [] At least one must appear in the story title
hours_back 24 How far back to look for stories (hours)
search_terms (built-in list) Terms sent to Algolia API to drive the search
hits_per_page 50 Results per Algolia API query
request_timeout 30.0 Seconds for HTTP requests

Output item schema

{
  "id":           int,
  "title":        str,
  "url":          str,   # external link (may be empty for Ask HN / Show HN)
  "score":        int,   # HN points
  "comments_url": str,   # https://news.ycombinator.com/item?id=<id>
  "summary":      str,   # LLM one-liner
}

Underlying collector

  • collectors/hn_collector.py
    • fetch_stories(keywords, min_score, max_items) — scrapes HN top stories

Tests

PYTHONPATH=. pytest tests/test_hn_collector.py -v