Skip to content

Kinds of things

Markdown

A saved link is never just a link. The reader works out what the page is and keeps the facts that belong to that kind: a book has authors and an ISBN, a flight has a departure and a seat, a product has a price and a brand. The kind decides how the card is drawn and which facts appear in the details panel and as properties.

Three things decide it, in order:

  1. The site. For the services Copilka understands directly (below), the kind is known from the address.
  2. The page’s own data. Structured data on the page (schema.org types in JSON-LD or microdata) maps to a kind.
  3. A model. When the page’s data is not enough, a model reads the text and chooses between place, recipe, film, book, product, article, video, event, flight, ticket and a plain link. It can also recognise a receipt.

A link whose kind cannot be told stays a Link, with a title, picture and summary but no typed facts. You can steer the reader with a Where to look note; see How a link reads itself.

Kind What the reader keeps
Place Address, coordinates, phone, website, categories, rating and number of ratings, price level, opening hours, a map link and a photo.
Recipe Ingredients, steps, preparation, cooking and total times, yield, cuisine, category, nutrition, rating, author.
Film Year and release date, runtime, directors, cast, genres, ratings, tagline, content rating, and season and episode counts for a series.
Book Authors, ISBN, publisher, publication date, page count, language, subjects, series, rating, and the price where the page offers one.
Product Brand, category, price and currency, availability, condition, SKU and barcode, seller, rating, colour, material, pictures.
Article Authors, published and modified dates, section, word count and reading time, publication, keywords, an excerpt.
Video Provider, duration, channel, published date, view count, whether it is live, the thumbnail.
Event Start and end, venue and address, whether it is online, status, organiser, lowest and highest price, whether it is free.
Post Author and handle, avatar, the text, when it was posted, attached media, the network, likes, replies and reposts.
Paper Authors, venue, year, DOI or arXiv id, citation count, categories, a link to the PDF.
Package Registry, name, version, licence, downloads, repository, the install command.
Repository Provider, owner and name, language, stars, forks, open issues, topics, licence, homepage, default branch, last push, whether archived or a fork.
Issue Provider, repository, number, state, whether it is a pull request, author, comment count, opened and closed dates.
Music Whether it is a track, album, playlist or artist; artist, album, track count, duration, year, genres, provider, artwork.
Podcast Show, episode number, duration, published date, provider, artwork, the audio address.
Listing Price, property type, rooms, area, locality and address, for a home for sale or rent.
App Developer, platform, rating and count, price, version, category, icon.
Person Handle, where they work, location, followers, the network, avatar, website, phone and email where the page gives them.
Job Company, location, whether remote, employment type, salary, posted and closing dates.
Question Site, score, answer count, whether answered, author, asked date, top tags.
Model A Hugging Face model, dataset or space: downloads, likes, licence, task, size. Drawn as a model card with a pattern generated from the model’s name.
Code Language, filename, the first sixty lines, line count, author. Also what a code file saved from disk becomes.
Palette The colours.
Flight Airline and flight number, class, departure and arrival airports, cities, terminals and times, duration, aircraft, seat, booking reference, passenger, status.
Ticket The event, issuer, what sort (concert, train, bus, ferry, cinema, museum, match), from and to, times, venue, seat, row, section, coach, entrance, who it admits, reference, holder, price.
Receipt Merchant and address, when, total, subtotal, tax, tip, currency, how it was paid (never a card number), reference, category, and every line with quantity and amount.
Image A summary, the colour palette, pixel size and format.
File A summary and the file’s extension.
Link A title, picture and summary, and nothing typed.

Flights, tickets, receipts, events, people and products can also come from a photograph or a scanned code: see Pictures and Saving on iPhone.

For these, Copilka asks the service rather than scraping the page, so the facts are exact and the reading is quick:

  • Places: Google Maps, Apple Maps, OpenStreetMap.
  • Video: YouTube, Vimeo.
  • Code: GitHub, GitLab and Codeberg repositories; GitHub issues and pull requests; Gists.
  • Papers: arXiv, Crossref.
  • Packages: npm, PyPI, crates.io, RubyGems.
  • Apps and podcasts: the App Store, Apple Podcasts.
  • Questions: Stack Exchange sites.
  • Models: Hugging Face.
  • Music: Spotify, Apple Music, SoundCloud, Tidal, Deezer.
  • Games: Steam.
  • Posts: Hacker News, Threads, Bluesky, TikTok, Reddit, Mastodon.
  • Colours: Coolors.

For books (Goodreads, Open Library, Waterstones, Audible and others), events (Ticketmaster, Eventbrite, Piletilevi, DICE, Resident Advisor, Songkick, Bandsintown) and films (IMDb, Letterboxd, TMDB, Rotten Tomatoes, MUBI), Copilka knows the shape from the address and reads the details from the page.

Everything else is read from the page, and from a model where the page’s own data is thin.

The card’s picture is the page’s own: its OpenGraph image, a video’s thumbnail, a track’s artwork, an app’s icon. When a page has none and is read in a browser, the largest picture on screen stands in.

You can always put your own picture on a card. The Cover section of the details panel takes a file or a pasted image, and Reset puts the page’s picture back. A book is drawn as a book, at an angle, with its picture as the jacket, or its title on a plain jacket when it has none.