Agamotto

A watch-order guide to every film and series in eleven franchises: Marvel, X-Men, DC, Star Wars, Alien & Predator, Godzilla & the MonsterVerse, Planet of the Apes, Dune, Middle-earth, the Wizarding World and Westeros. In English and Chinese.

Live: timeline.luminoid.dev (EN · 中文 from the language switch)

Features

• One section per continuity. The MCU, Sony's Spider-Man Universe, Toho's Shōwa era and the rest each get their own section on a vertical timeline.
• Story or release order. Either can run in both directions.
  • Story order follows the official timeline where one exists: Marvel's Disney+ order, StarWars.com, and the years in Tolkien's appendices.
  • Elsewhere it follows each continuity's own chronology.
• Details on every card. Runtime, seasons and episodes, plus IMDb and Douban scores with their vote counts and Rotten Tomatoes' Tomatometer, Certified Fresh and audience score, with links to each site and to JustWatch. Tap a poster to see it whole.
• Links to any franchise or section.
  • #xmen opens a franchise.
  • #xmen/fox-xmen, or just #fox-xmen, also scrolls to that section.
• My ratings. Import your IMDb ratings (their CSV export) and your Douban watched list (a bookmarklet) to see your scores next to the site's.
  • They stay in your browser and are never uploaded.
• A few hidden extras.

Layout

site/            the built site: index.html, posters/, icons, 404.html, _headers, robots.txt, sitemap.xml.
                 Deploy this folder as is.
pipeline/        the scripts, template and data that build it
  brand/         sources for the icons and the social card
  tests/         headless Chrome checks and a screenshot tool
  local/         the data scripts, Douban matches and caches; not in the repository (see Caches)

Building

You need:

• Python 3 with Pillow (pip install -r requirements.txt)
• curl
• Node 22 or later, for the built-in fetch and WebSocket
• Google Chrome, for the Douban scripts, the icons and the tests. The scripts look in the macOS location; set CHROME to the browser's path elsewhere.

Common tasks have make targets (make lists them):

make build       # rebuild site/ from the template and data (seconds, offline)
make serve       # build, then serve site/ at http://localhost:8000 (PORT=... to change)
make test        # easter eggs, links, rating imports and release checks in headless Chrome, offline
make test-live   # the same, plus the Douban bookmarklet on a public profile
make icons       # redraw the favicons, touch icon and social card from pipeline/brand/
make data        # full refresh: fetch, douban, enrich, build (slow, needs network)

The steps behind make data, run from pipeline/. Everything under pipeline/local/ (the fetch, Douban and enrich scripts, the Douban match files and the caches) is not in the repository: the scripts drive third-party sites and stay on the maintainer's machine, so a clone can edit titles and rebuild the site from the committed JSON, but not refresh the data.

Step	Command	Writes
Edit titles, story order and settings	seed.py; Chinese names in zh_titles.py
Resolve titles on Wikipedia and Wikidata, download posters (a 200×300 card crop and the whole poster for the zoom)	python3 local/fetch.py	data.json, ../site/posters/
Queue Douban lookups for new titles	python3 local/douban/add_jobs.py	local/douban/jobs.json, local/douban/split.json
Match films through their Douban pages	node local/douban/subj.mjs	local/douban/subj.json
Match series seasons, and films without a Douban ID	node local/douban/m.mjs	local/douban/m.json
Read every later-season page those found, to check its IMDb ID and votes	node local/douban/check.mjs	local/douban/check.json
Add ratings, runtimes and seasons, and cross-check sources	python3 local/enrich.py	data_enriched.json, report.txt, rt.json, wiki_infobox.json, local/imdb/subset.json
Build the site	python3 build.py	../site/index.html, 404.html, _headers, robots.txt, sitemap.xml

A Douban match counts only when the Douban page lists the title's IMDb ID; a later season counts when its page lists the series or one of its episodes. report.txt lists every place where the sources disagree.

Caches and refreshing ratings

Nothing here expires, so a rerun reuses everything it already has.

Not committed, all under pipeline/local/:

• fetch.py, enrich.py and douban/: the scripts (see above); douban/ also holds the job queue, the matched subject and season pages, and the headless Chrome profile. The Douban scripts skip any title or page that already has an entry.
• cache/ holds every HTTP response, named by a hash of its URL. Files starting with c_ are Rotten Tomatoes pages (and Douban responses from an older version of the scripts); the rest are Wikipedia, Wikidata, TMDB and poster downloads.
• imdb/ holds IMDb's title.basics, title.episode and title.ratings files, downloaded by hand from https://datasets.imdbws.com/, and subset.json, the slice of them the site uses (rebuilt when it lacks an ID).

Committed, and reused the same way:

• rt.json: Rotten Tomatoes scores; enrich.py fetches only titles missing from it.
• wiki_infobox.json: runtimes, seasons and episode counts parsed from Wikipedia infoboxes.

To refresh:

• IMDb: download the three files again, delete local/imdb/subset.json, then make enrich build.
• Rotten Tomatoes: REFRESH_RT=1 make enrich, then make build. It refetches every page.
• Douban: remove the titles you want refreshed from local/douban/subj.json and local/douban/m.json (and their season pages from local/douban/check.json), then make douban enrich build. Each title takes a few seconds, and Douban throttles long runs.
• Wikipedia counts: delete wiki_infobox.json and the files in local/cache/ whose names don't start with c_, then make enrich build. The next make fetch downloads Wikidata and posters again too.

Checks

make test runs the three check scripts in pipeline/tests/. Each prints one line per check and exits non-zero on a failure:

• eggs.mjs covers the hidden extras.
• ratings.mjs covers links, history, the Douban bookmarklet and both rating imports. With OFFLINE=1 it only parses the bookmarklet; without it, it runs the bookmarklet on a public Douban profile.
• site.mjs serves site/ with its _headers and covers the release files and meta tags, the Content-Security-Policy, keyboard focus and screen-reader announcements, recovery from bad stored settings, and the English and Chinese strings.

Screenshots of any view:

node pipeline/tests/shot.mjs out.png 1300 900 "file://$PWD/site/index.html#dune"

Data sources

• Wikipedia and Wikidata: titles, release dates, runtimes, episode counts and most posters.
• TMDB: posters where Wikipedia has none.
• IMDb: ratings, runtimes and episodes, from IMDb's non-commercial datasets, which IMDb licenses for personal, non-commercial use only. Information courtesy of IMDb (https://www.imdb.com). Used with permission.
• Douban: ratings and Chinese titles. Every match is checked against the IMDb ID shown on the Douban page.
• Rotten Tomatoes: Tomatometer and audience scores, per season where they exist.

This is an unofficial fan index, not affiliated with any studio or rights holder. Posters belong to their owners.

Deploying

site/ is a static site with no build step at serve time, made for Cloudflare Pages (output directory site, no build command). Links use URL fragments, so the host needs no rewrite rules.

• _headers carries the security headers and a Content-Security-Policy that allows the page's one inline script by its hash. make build rewrites both files, so deploy index.html and _headers together.
• Poster URLs carry a version, so _headers lets browsers keep posters for a year.
• On Cloudflare, set the zone's Caching → Browser Cache TTL to "Respect Existing Headers". Otherwise its four-hour default overrides the page's revalidation and visitors can see a stale page.

License

CC BY-NC-SA 4.0 for the code, the design, the watch-order compilation and the original text. Ratings, titles and posters belong to their sources and owners; LICENSE lists them.