@linkoi/core

URL and HTML metadata extraction. Developer preview, v0.1.1.

import { resolve, fromHtml, type Metadata } from '@linkoi/core'
const page = await resolve('https://example.com', {
  timeoutMs: 5000,
  maxBytes: 2_097_152,
  selfHosts: ['your-api.example.com'],
  // youtubeApiKey: process.env.YOUTUBE_API_KEY,
})

resolve(input, options) accepts a string or URL and returns Promise<Metadata>. Timeout is per fetch, not a total request deadline. HTML fetches follow up to five redirects and retry upstream 429s twice. Direct media documents are not supported.

fromHtml(html, url) always parses supplied HTML offline. The base URL must pass the URL guard. The legacy fallback option is ignored.

Metadata contains url, nullable title, description, image, logo, publisher, author, date, lang, and source: 'fast' | 'fallback'. The fast path limits titles to 200 and descriptions to 500 characters. Images are URLs, not verified image objects. The favicon fallback may point to a nonexistent file. New results always have source: "fast"; the legacy source union is retained for stored data compatibility.

Lower-level exports are available at @linkoi/core/guard, /extract, /normalize, /fetch-html, /fallback, and /youtube. No caching or HTTP server is created by importing core.

See the root SECURITY.md before fetching user-supplied URLs. Licensed under PolyForm Shield 1.0.0. Commercial use is allowed for non-competing purposes. See LICENSE for the exact terms.