src/feeds.js

import { Feed } from 'feed'
import { JSDOM } from 'jsdom'
import { Readability } from '@mozilla/readability'

const PORT = process.env.BEACON_PORT || 3700

export async function pageToRSS(page, sourceUrl) {
  const [html, title, url] = await Promise.all([
    page.content(),
    page.title(),
    Promise.resolve(page.url()),
  ])

  const dom = new JSDOM(html, { url })
  const doc = dom.window.document

  const feed = new Feed({
    title: title || sourceUrl,
    id: url,
    link: url,
    feedLinks: {
      rss: `http://localhost:${PORT}/feed?url=${encodeURIComponent(sourceUrl)}`,
    },
    generator: 'beacon',
    updated: new Date(),
  })

  // Try Readability — works for article pages
  const reader = new Readability(doc.cloneNode(true))
  const article = reader.parse()

  if (article && article.length > 500) {
    feed.addItem({
      title: article.title || title,
      id: url,
      link: url,
      description: article.excerpt,
      content: article.content,
      date: article.publishedTime ? new Date(article.publishedTime) : new Date(),
      author: article.byline ? [{ name: article.byline }] : [],
    })
  } else {
    // Link-list page (HN, Reddit, news front pages) — harvest significant links
    const seen = new Set()
    const links = [...doc.querySelectorAll('a[href]')]
      .filter(a => {
        const text = a.textContent.trim()
        const href = a.href
        if (!href || !text || text.length < 12) return false
        if (href.startsWith('#') || href.startsWith('javascript:')) return false
        if (seen.has(href)) return false
        seen.add(href)
        return true
      })
      .slice(0, 40)

    for (const link of links) {
      feed.addItem({
        title: link.textContent.trim(),
        id: link.href,
        link: link.href,
        description: link.title || link.textContent.trim(),
        date: new Date(),
      })
    }
  }

  return feed.rss2()
}