A self-hosted page for your ranked favorites: movies, TV shows, albums, songs, video games, theater, and anything else you care to rank in a custom list. One owner, as many lists as you like, each with its own size (Top 5, Top 100, … up to 65,535), with proper posters and cover art pulled from the best free metadata sources.

Your page is titled "Your name's Canon" by default. Set your name under Settings → Profile, or give the site any title you like there.

Visitors just see your lists. There's no sign-up, no feed and no login button. You manage everything from /login.

Ranked lists per category: Movies, TV Shows, Albums, Songs, Video Games and Theater. Each list sets its own size, title, description and style.
Search-to-add from TMDB, TheTVDB, IGDB, Deezer/MusicBrainz and Wikipedia, with add-manually as a fallback for anything they miss.
Custom lists for anything else, like sports teams, drinks or restaurants. You name the list and add entries by hand (with your own photo or logo), or pull a name and picture from Wikipedia.
Theater by production: record which staging you mean (Broadway, West End, North American Tour, a revival…) plus venue and cast, and use that production's own poster.
Four styles per list: Podium (top 3 on a medal podium, the rest behind "Show more"), Spotlight (#1 as a big feature), Grid and List. You pick each list's style in List settings, and visitors can switch styles on a list's own page.
Drag to rank, or use the arrow buttons or type a new number. Paginated editing keeps huge lists usable.
Refresh from source: re-pull an entry, or a whole list, when its metadata changes upstream. Your notes, ranking and theater productions are kept, and artwork you chose yourself is only replaced if you say so. Custom lists and hand-added entries have no source, so they aren't refreshed.
Art is cached on your server, so pages never hotlink and keep working if a source changes. You can also paste an image URL or upload your own.
Preview clips for songs and albums: a play button plays 30 seconds from Deezer (an album plays its most popular track). Clips stream through your server, so visitors never connect to Deezer. You can turn this off in Settings → Profile.
Export any list as a titles-only .csv or .txt.
Descriptions and your reasons: each entry keeps the source's description (TMDB overview, IGDB summary, Wikipedia intro), and you can add a "Why it's #N" note explaining its spot. Both show in the detail view.
12 theme presets plus a custom CSS box. Every color is a CSS variable.
Single binary + SQLite, no external database. Around 15 MB, and it runs anywhere Docker does.

Quick start (Docker Compose)

git clone https://git.iamchrisama.com/cdnutter/canon && cd canon
cp .env.example .env        # optional: admin account + API keys
docker compose up -d
docker compose logs canon # shows the one-time setup code if you didn't set an admin in .env

Open http://localhost:4242/setup, enter the setup code and create your account. Then go to Settings → Integrations to add API keys.

To skip the setup page, set CANON_ADMIN_USERNAME and CANON_ADMIN_PASSWORD in .env before the first start. They're only used when no account exists yet.

Opening it from other devices

Canon listens on all network interfaces (0.0.0.0) by default, so phones and other computers on your network can open http://<this computer's IP>:4242. When run outside Docker, the startup log prints the exact addresses. If another device can't connect, the host's firewall is probably blocking the port (for example sudo ufw allow 4242/tcp). To keep Canon private to one machine, set CANON_HOST=127.0.0.1, or publish 127.0.0.1:4242:4242 in docker-compose.yml.

With Arcane (Git sync)

The compose file works as an Arcane project synced from this repository:

Projects → New → Git sync, direction Pull, compose file path docker-compose.yml.
Turn on Redeploy after sync to rebuild whenever you push. pull_policy: build makes every deploy build the image from the synced source; it never looks for canon:latest on a registry.
Put your settings (admin account, API keys) in the project's environment in Arcane. Arcane writes them to .env, which Canon reads.

The database lives in the canon_canon-data named volume, outside the synced folder, so a sync can't overwrite or delete it. To use a host folder instead, set CANON_DATA_PATH to an absolute path (it must be writable by UID 1000). To move data from an existing ./data folder into the volume:

docker compose stop canon
docker run --rm -v "$PWD/data:/from" -v canon_canon-data:/to alpine sh -c 'cp -a /from/. /to/ && chown -R 1000:1000 /to'

The compose file joins the external chromanet network; create it first (docker network create chromanet) or remove the networks entries.

Without Compose

docker build -t canon .
docker run -d --name canon -p 4242:4242 -v canon-data:/data canon

Or from source (Go 1.26+): go run ./cmd/canon stores data in ./data.

Data sources and API keys

Category	Source	Key?	Get one
Movies	TMDB	Yes (free)	themoviedb.org/settings/api. The v3 key or the v4 Read Access Token both work.
TV Shows	TheTVDB (default) or TMDB	Yes	thetvdb.com/api-information. Switch TV to TMDB in Settings to reuse your TMDB key.
Video Games	IGDB	Yes (free, via Twitch)	Register an app at dev.twitch.tv/console and use its Client ID + Secret (guide).
Albums & Songs	Deezer and MusicBrainz + Cover Art Archive, searched together (or either alone)	No	Nothing to set up.
Theater	Wikipedia	No	Nothing to set up.
Custom	Manual entry, or Wikipedia	No	Nothing to set up.

Why these for music and theater? Neither needs a key.

Music: Canon searches Deezer and MusicBrainz together and merges the results. Deezer's public API has no key, generous limits, 1000×1000 covers and 30-second preview clips, but its search misses some albums it carries (Death Magnetic only turns up for an exact artist + album query) and has few game soundtracks. MusicBrainz is fully open, finds those, and tags soundtracks, but allows only one request per second. Entries added from MusicBrainz or by hand are matched to Deezer when saved, so they can play previews too. Spotify, Apple Music and Last.fm all need keys or OAuth and have stricter terms. For game music, VGMdb has the deepest catalogue but no public API, so add anything missing by hand.
Theater: there's no open theater database with an API (IBDB and Playbill don't offer one). Wikipedia covers nearly every musical, play and opera, usually with poster or logo art. Canon filters search to pages that use the musical, play or opera infobox. Wikipedia describes the work, so you pick the production yourself and can swap in that production's poster.

Keys entered in the UI are stored in the SQLite database on your server. Keys set as environment variables take precedence and show as locked in the UI.

Configuration

Variable	Default
`CANON_HOST`	`0.0.0.0`	Interface to listen on. `0.0.0.0` makes it reachable from other devices on your network; `127.0.0.1` keeps it on this machine
`CANON_PORT`	`4242`	Listen port
`CANON_DATA_DIR`	`/data` (Docker), `./data`	Database + cached images
`PUID` / `PGID`	`1000` (image), `99`/`100` (Compose)	Docker only: the user and group that own everything Canon writes. The Compose file defaults to Unraid's `nobody:users`
`UMASK`	`022`	Docker only: permissions for new files (`002` makes them group-writable)
`CANON_ADMIN_USERNAME` / `_PASSWORD`		Create the owner account on first start
`CANON_TMDB_API_KEY`
`CANON_TVDB_API_KEY` / `CANON_TVDB_PIN`		The PIN is only for user-supported keys
`CANON_IGDB_CLIENT_ID` / `_SECRET`		Twitch app credentials
`CANON_TRUST_PROXY`	`false`	Trust `X-Forwarded-*` headers (set behind a reverse proxy)
`CANON_SECURE_COOKIES`	`false`	Force `Secure` cookies (set if served over HTTPS)
`CANON_LOG_LEVEL`	`info`	`debug` logs every request

Reverse proxy

Canon serves plain HTTP. Put it behind Caddy, Traefik or nginx for HTTPS, and set CANON_TRUST_PROXY=true so logins are rate-limited per real client and cookies are marked secure. Caddy example:

favorites.example.com {
    reverse_proxy canon:4242
}

Theming

Pick a preset under Settings → Appearance: Midnight, Obsidian, Nord, Dracula, Mocha, Rosé Pine, Gruvbox, Forest, Synthwave, Crimson (red on OLED black), Daylight, Paper or Latte. Then add your own CSS if you like. It loads after the theme on every page, so overriding variables is enough for a whole new look:

:root, [data-theme] {
  --bg: #0d0b12;
  --surface: #16131d;
  --accent: #ff5a7a;
  --font-display: "Georgia", serif;
  --radius-art: 14px;
}

See web/static/css/themes.css for every variable.

List sizes

A list can hold 1 to 65,535 entries. The editor recommends 100 or fewer and warns past 1,000. Long lists work (public pages and the editor paginate at 100), but they're hard for visitors to take in.

Private sites

To share with friends and family without being fully public, set a password under Settings → Privacy → Site password. Visitors then see a password prompt over a blurred placeholder; none of your lists, images or profile are sent until they enter it. Once they do, that browser stays unlocked for 90 days. Changing the password asks everyone for the new one, and you never need it while signed in. A protected site also tells search engines not to index it.

Ranking

Entries sit at places from #1 to the list's size, and a place can be left empty: a game can be #28 with nothing at #29 and the next at #30. In the list editor:

Type a place in an entry's number box and press Enter (↑/↓ jump between boxes, Alt+↑/↓ moves the entry one place). Or drag it.
Moving onto an empty place just puts it there. Moving onto a taken place shifts the entries in between one step, stopping at the first empty place, so entries you placed further down stay put.
An entry's ⋯ menu can insert an empty place above it, move it to #1 or the end, or remove it (you choose whether the entries below move up or the place stays empty).
Empty places show as their own rows: Add here makes the next entry you add go there, and Remove closes them up. Close gaps renumbers everything 1, 2, 3….

Edits show instantly and save in the background.

Importing a list

Import in the list editor adds entries from a file. Line numbers are places: line 1 becomes #1, line 38 becomes #38, and a blank line leaves its place empty (entries already at those places move down to make room).

Text (.txt, or paste): one title per line. If every line is numbered (38. Super Mario World), the numbers are the places instead. A trailing (1999) is read as the year. For songs and albums, Artist - Title works either way round.
CSV / TSV: a header row picks the columns: Title (or Name), Year, Artist/Director/Developer, Note (or Why), and Rank to give places directly. Without a header the first column is the title. The first row after the header is #1.

Each entry is searched for and matched. Canon only adds a match on its own when it's certain: the title matches exactly and nothing else fits as well (a year or artist in the file helps). Everything else is left for you to check, with a recommended pick, the other results, Add as written to make a new entry from your text, Skip, and a box to search again. Custom lists search all of Wikipedia, so there a match is only certain when a year or subtitle confirms it.

Exporting a list

Every list can be downloaded as its titles only, one per line, where line N is place #N: an empty place is a blank line. There are no ranks, header row or other fields, and importing the file puts everything back where it was.

Export → Titles as CSV (/l/<list>/export?format=csv): a single column. It starts with a UTF-8 BOM so Excel shows accented titles correctly.
Export → Titles as text (/l/<list>/export?format=txt): plain lines.

The Export menu appears in the list editor and, when you're signed in, on the list's public page. The links follow the list's visibility: anyone can download a public list, and only you can download a hidden one.

Upgrading from TopDeck

Canon was called TopDeck, and upgrading keeps all your data:

The database is renamed from topdeck.db to canon.db automatically on first start.
Environment variables changed prefix from TOPDECK_ to CANON_. Rename them in your .env.
You'll be signed out once (the session cookie was renamed).
Docker Compose now always uses the project name canon. If you ran the old compose file, your data is in the topdeck_topdeck-data volume. Copy it into the new one before starting:
```
docker volume create canon_canon-data
docker run --rm -v topdeck_topdeck-data:/from -v canon_canon-data:/to alpine cp -a /from/. /to/
```

Backups

Canon backs itself up every day at 3:00 and keeps the last 7. Change this under Settings → Backups: pick a preset (hourly, every 6 hours, daily, weekly, monthly), write your own cron expression, or turn scheduled backups off. Back up now makes a manual backup; manual backups are never deleted automatically. Every backup can be downloaded or deleted from the same page.

Each backup is a .tar.gz holding canon.db (a consistent snapshot, taken while Canon keeps running) and the images/ folder.

Where: CANON_BACKUP_DIR, by default data/backups. Docker Compose mounts the canon_canon-backups volume at /backups; set CANON_BACKUP_PATH to a host folder (a NAS share, say) to keep them off the data volume.
Time zone: schedules use the server's time zone, UTC in Docker unless you set TZ (e.g. TZ=America/Los_Angeles) in .env.
Cron syntax: minute hour day month weekday, with *, lists (1,15), ranges (mon-fri), steps (*/6) and @daily-style shorthands. For example, 30 2 * * mon-fri is 2:30 on weekdays.

To restore, stop Canon, extract a backup into the data folder, and start it again:

docker compose stop canon
docker run --rm -v canon_canon-data:/data -v canon_canon-backups:/backups alpine \
  sh -c 'rm -f /data/canon.db-wal /data/canon.db-shm && tar xzf /backups/<file>.tar.gz -C /data && chown -R 1000:1000 /data'
docker compose start canon

Development

go run ./cmd/canon        # http://localhost:4242
go test ./...
go test -tags live ./internal/providers   # hits the real music APIs

Templates and static assets in web/ are embedded into the binary, so rebuild after editing them. The front end is plain ES modules with no build step.

cmd/canon         entry point, first-run bootstrap
internal/store      SQLite schema, lists, items, ranking, settings, sessions
internal/providers  TMDB, TheTVDB, IGDB, Deezer, MusicBrainz, Wikipedia
internal/imgcache   downloads and stores artwork
internal/server     HTTP routes, auth, JSON API, page rendering
web/                templates, CSS themes, JS, icons

README.md Unescape Escape

⚠️ THIS PROJECT WAS ENTIRELY VIBECODED AND IS SLOP. DO NOT PROCEED IF YOU DO NOT ALIGN WITH THESE VALUES. DO NOT COMPLAIN ABOUT IT IN PULL RELEASES, ISSUES, ETC. UNLESS THERE ARE SEVERE SECURITY RISKS. THEY WILL BE REMOVED. PROCEED AT YOUR OWN RISK! ⚠️

Canon