mangabaka-seeder
A tool for seeding edits to the MangaBaka manga/comics database, in the spirit of yambs for MusicBrainz.
Status: v0.1 works end to end locally (Tower of God: current and archived covers found, checked and rehosted); not deployed yet. Design: the spec.
Today's goal is one thing done well: given a MangaBaka series link and a WEBTOON or Tapas series link, find every cover that series has used, current and historical, and request them added to the MangaBaka series. The current poster becomes the main (default) cover for the edition's language.
Two ways to use it
- Web page (primary). Open the page, paste the two links, review the covers. A small server behind the page reads Webtoons and the Wayback Machine for you, because browsers can't. It runs on your own machine with Bun.
- Userscript. Install
mbseed.user.jsin Tampermonkey or Violentmonkey. On a MangaBaka series page it adds a Seed covers panel; paste the Webtoons link. It reads Webtoons directly from your browser, and with litterbox or catbox it needs no server at all.
A command-line version (Bun) may follow.
How it works
-
Read both links. The MangaBaka link gives the series id; the Webtoons link gives the title,
title_no, language, and official or canvas. If the MangaBaka series has no link to thattitle_no, you're warned. -
Collect covers.
- Current: the portrait poster, the square thumbnail, and the banner and hero art.
- Historical (tick "Old covers from the Wayback Machine"; off by default, it takes minutes): cover images in archived copies of the series page on the Wayback Machine, up to one snapshot per month.
-
Check each image. It downloads each image to get its size and hash, classifies it by shape (portrait →
volume, square →other, wide →banner), and drops duplicates and images MangaBaka already has. -
Build the submission. Existing covers are kept and new ones added. The current poster is the default (the square for canvas series, which have no poster). Every new cover gets a note recording where and when it was seen.
-
Rehost the images when you click Rehost selected, to the host you pick:
Host Links last the server's own storage (default, if configured) 30 days litterbox 1 hour – 72 hours catbox.moe permanently none ( direct): submit the Webtoons URL itselfas long as Webtoons keeps it A link must still work when a MangaBaka moderator looks at the request. If it has expired by then, run the tool again.
-
Hand it to you.
- Without API access (most users): press Download selected (ZIP) for
the cover files plus a
covers.txtwith each one's type, rating and note, and upload them in MangaBaka's cover form for the series (the page links to it). No rehosting needed. Or rehost them and paste the hosted URLs instead. - With a personal access token: the tool shows MangaBaka's preview diff and, after you confirm, submits the request for review.
- Without API access (most users): press Download selected (ZIP) for
the cover files plus a
The page keeps a history of the series you searched, in your browser: the covers, your ticks, notes and ratings, and what you did with them (downloaded, rehosted, filed). Pick one from History or the start screen to reopen it instantly; Search again refreshes it and keeps your edits. It holds the 30 latest series (up to 300 MB) and stays in this browser only.
Nothing is uploaded or sent to MangaBaka without your action, and submissions
always go to review. You choose the content rating (default safe); titles
Webtoons marks as mature show a warning. Your token is sent only to
MangaBaka's API, never to our server.
MangaBaka's submissions API is only open to contributors, or to users with at least 25 approved submissions.
Supported sources
| Source | Links | Current covers | Historical covers |
|---|---|---|---|
| WEBTOON (official & canvas) | https://www.webtoons.com/<lang>/…/list?title_no=N, m.webtoons.com |
poster, square, banner, foreground | Wayback Machine |
| Tapas (comics & novels) | https://tapas.io/series/<name> (or …/info, the series number, or an /episode/N link) |
portrait cover (full size), banner | Wayback Machine |
Tapas is English only. Mature Tapas series are shown only to signed-in adults: the page can't read their current covers (it says so and still searches the archive), but the userscript can, when you are signed in to Tapas in that browser.
Each Webtoons language edition has its own title_no. Run the tool once per
edition; the covers are tagged with that edition's language.
Notes and limitations
- Some covers may only exist in the archive. If an old cover is no longer on Webtoons' server, its archive.org copy is used instead.
- Not everything is a cover. Some historical images are promotional or landing-page art. Review the list before submitting.
- Requests are rate-limited to stay polite to Webtoons, the Wayback Machine and MangaBaka.
Running it
Needs Bun 1.3 or newer.
bun install --frozen-lockfile
bun run build # the page and the userscript into dist/web/
MBSEED_STORAGE_DIR=/srv/mbseed/covers MBSEED_PUBLIC_URL=https://mbseed.example.org bun run start
Open the address it prints. Without MBSEED_STORAGE_DIR the server runs without
its own storage and the page offers litterbox, catbox and no rehosting. All
settings, and how to put it behind a reverse proxy, are in
docs/ARCHITECTURE.md.
With Docker
Every push to main builds an image and publishes it to
gitea.lucaswinther.info/lucasw89/mangabaka-seeder (latest, the branch name,
sha-<commit>, and the version for v* tags). Copy
docker-compose.example.yml to docker-compose.yml,
set MBSEED_PUBLIC_URL, and run docker compose up -d. To build it yourself:
docker build -t mangabaka-seeder .
The CI workflow (.gitea/workflows/ci.yml) publishes with a repository or user
secret REGISTRY_TOKEN: a Gitea access token with package write permission
(Gitea's job token cannot publish packages).
The userscript
Install /mbseed.user.js from the running server (or dist/web/mbseed.user.js)
in Tampermonkey or Violentmonkey. A Seed covers button appears on MangaBaka
series pages. To rehost on your server from the userscript, enter its address
under Settings in the panel; build with MBSEED_PUBLIC_URL set so the script
is allowed to reach it.
Development
Working on this repo, by hand or with an AI agent: start at
AGENTS.md. Design: docs/specs/; current
architecture: docs/ARCHITECTURE.md.