Media

Status: Built 2026-09-30 (API and seed shipped; web and mobile screens in the same change set; posters wait on the media CDN)

The Trick Book has The Couch, a Netflix-style library of action-sports films: a rotating hero, shelves, portrait cards, a detail page with credits and comments, and an admin that imports films from batch files. Vegan Grove's Media section is the same grammar for the cause, and then more: every title ends in an action, carries a content warning when it needs one, lists every place to watch it with a free-first rule, and never records who watched what.

Parity with The Couch

The table is the standard this feature is held to. "Couch" is what The Trick Book ships today (surveyed 2026-09-30 across its backend, web and mobile repos); "Vegan Grove" is what this repo set ships; "Extension" is what the cause needed on top.

Feature	The Couch	Vegan Grove	Extension
Catalogue record	one document per film with slug, type, tags, year, producer, directors, riders, duration, poster and backdrop, watch options with an access level, provenance, rights, SEO	same shape as media_items with kind, tags as topics, directors and featuring, runtime, poster and backdrop, watch links with access, provenance, external ids	content warnings, an action list, an official site, TMDB rating
Stable public URL	slug or id, both resolved	slug only	none needed
Watch providers	array stored, only the first free one shown on web, ignored on mobile	every provider rendered on both clients with an access badge, JustWatch attribution when providers came from TMDB	curated official free stream stays first; providers never overwrite it
Rights	"external only" plus a notice that the film is not rehosted	never hosted, same notice on every item page	none
Hosted streaming	Bunny HLS with signed URLs for owned edits	not built; the library is links and trailers	Vegan Grove may host its own recordings later through the feed's video pipeline
Trailer	YouTube iframe loaded on page view	click to load through youtube-nocookie.com, nothing fetched before the click	none
Posters	hot-linked YouTube frames	poster and backdrop from TMDB, re-hosted on the media CDN; a generated poster when there is none	none
Hero	rotates every 8 seconds through featured plus recent items with an image	featured items in a deterministic daily order from /api/media/home, identical on web and mobile, 8 second rotation, paused on hover and under reduced motion	none
Shelves	one collection per film, ten per row, built with one query per row	many-to-many collections with an editor-set order, plus automatic rows (new, free, under 30 minutes, one per topic) in one call	topic rows come from the vocabulary in the seed
Browse and sort	sport pills and six sorts in the URL	kind pills, search, and five sorts in the URL (featured, release, title, rating, runtime), cursor pagination	free-only and runtime filters
Search	unindexed regex over five fields	indexed filters plus a regex over title, tagline, synopsis, directors, featuring, tags and genres (the library is small)	none yet
Detail page	player, kicker, title, description, love and respect, share, credits card, rider links, comments, JSON-LD Movie and VideoObject	backdrop header, poster, kicker, tagline, save, two reactions, share, synopsis, providers, trailer, credits, warnings, actions, related shelf, JSON-LD Movie with duration and BreadcrumbList	"This moved me" and "I took action" instead of love and respect
Related	none on the client	shared topics and genres, most shared first	none
Reactions	love and respect, no unique index	moved and acted, one row per member and type, counts public, authors private	none
Comments	flat list with a listing bug	not in this milestone	see below
Watchlist	none	private saved list per member, on web at /app/watchlist and on mobile	never a history
View counts	incremented on every metadata read including crawlers	none, by design	none
Admin	Next.js pages with a YouTube prefill and a direct upload	admin API for titles and shelves; every hand-set field is locked against ingest	admin pages are the next milestone
Import	batch JSON in the web repo, applied over SSH	the seed file in the API repo plus Wikidata discovery and TMDB enrichment, run by the data refresh	the seed carries the editorial layer
SEO	canonical, Open Graph video.movie, sitemap of one sport	canonical, Open Graph video.movie, sitemap and llms.txt of every published title and shelf	none
Mobile	a media tab: hero, shelves, explore grid, detail with a WebView player	reached from Home's Learn step: hero, shelves, explore grid with search and chips, detail with providers, trailer, save and reactions	none

The survey that produced this table lives with the project notes; the short version is that the Couch's strongest idea is its rights-aware catalogue record and its weakest is that nothing on it is private, and Vegan Grove keeps the first and fixes the second.

What ships

• A library home with a hero, editor shelves and automatic rows, then a browse grid with search, kind pills, sort and load more.
• An item page with everything a member needs to decide, watch, and act.
• Seven editor shelves seeded from media-collections.json: Start here, Investigations, Health and food, The planet, Oceans and wildlife, Ethics and philosophy, Activism.
• Twenty-four curated titles, Christspiracy among them, each with topic tags, a content warning where it applies, an official site, a free official stream where the maker publishes one, and one action link.
• A private watchlist and two reactions.
• An admin API for titles and shelves.

Privacy choices in the details

• Films are not hosted. watchLinks point out to the platform; no playback happens inside the app.
• Trailers are YouTube embeds that load only after a click, through youtube-nocookie.com, so simply visiting the page sends nothing to Google. The click is the consent.
• No "watched" tracking and no view counts. The watchlist is a saved list, not a history.
• Reaction and save counts are public on a title; who saved or reacted is answered only to that member, and those reads are never cached.
• Posters and backdrops are re-hosted on the app's own CDN. Nothing is hot-linked from TMDB or YouTube.

Data and visibility

media_items and media_collections are public. saved_media and media_reactions are private to the member and appear in the data inventory. See the data model.

API

Public: GET /api/media/home, GET /api/media?q=&kind=&tag=&year=&free=&maxRuntime=&sort=&cursor=, GET /api/media/collections, GET /api/media/collections/:slug, GET /api/media/:slug (adds viewer with a session), GET /api/media/:slug/related. Member: POST and DELETE /api/media/:id/save, GET /api/me/watchlist, POST /api/media/:id/reactions, DELETE /api/media/:id/reactions/:type. Admin: /api/admin/media and /api/admin/media/collections. Public reads carry Cache-Control: public, max-age=300, stale-while-revalidate=3600; anything with a viewer is private, no-store.

Screens

Web: /media, /media/[slug], /media/collections/[slug], /app/watchlist. Mobile: a stack under /media reached from Home's Learn step (home, explore, item, collection); not a tab.

Curation

Titles enter through the seed file, Wikidata, TMDB, or Open Library, or through the admin API, and land draft unless the source is trusted. Open Library adds kind: book rows (authors, year, one ISBN) with covers re-hosted in the media bucket. Admins rewrite the synopsis in the app's own words, check the watch links, set warnings and actions, place the title on a shelf, and publish. Items whose links all break are unpublished rather than left broken. The automation refreshes providers, posters, and books on a schedule and never publishes anything.

Next milestone

• Comments on titles, with a moderation queue (graphic-content threads attract abuse).
• "Suggest a film": a form that creates a draft, admin publishes.
• A screening kit per title (synopsis, discussion questions, runtime, licence status, how to obtain a public-performance licence, follow-up actions) and a "host a screening" form that creates an event of type screening.
• Admin pages on the web for what the admin API already allows.