Stock upstream README documented Moritz Jung's plugin, not this fork. Rewrite in setup-guide shape: TOC, plugin info table, watchlist frontmatter/sync-tier/watch-status-rule blocks, library type sources, command + settings tables, build/install steps, Issues with quoted error strings. Facts sourced from src: commands (main.ts), settings defaults (Settings.ts), active statuses + throttle + flip rule (SyncEngine.ts, build.ts), preserved fields (build.ts, parse.ts), TMDB v3/v4 branch (tmdb.ts), dev outDir (vite.config.mts). Upstream import/template/search-by-ID reference linked, GPL-3 credit kept.
16 KiB
Media DB Sync
An Obsidian plugin that fills in media metadata, re-syncs it on a schedule, and flips a series back to Unwatched when a new episode airs. Fork of Media DB by Moritz Jung.
The stock plugin writes a note once, at import time, and never touches it again. This fork adds the part that runs afterwards.
Table of Contents
- Media DB Sync
- Table of Contents
- What This Fork Adds
- Plugin Info
- Watchlist Notes
- Library Types
- Commands
- Settings
- Related Articles
- Pre-requisites
- Clone Repository
- Set Up Build Environment
- Install Into Obsidian
- Development
- Issues
- License and Credit
What This Fork Adds
| Capability | Stock Media DB | This fork |
|---|---|---|
| Import a title from an API | Yes | Yes, unchanged |
| Re-sync an existing note later | Manual, one note at a time | Scheduled, across the whole folder |
language, country, imdb_id, content_rating, trailer, homepage |
Left blank on TMDB imports | Filled |
producer |
Writes studio instead |
Individual producers |
| Episode tracking | No | last_episode, upcoming_episode, next_air_date |
| Watch-status automation | No | Watched flips to Unwatched on a new episode |
| Backfilling ids onto notes that predate the plugin | No | Watchlist: resolve missing TMDB ids |
| Manga, book, game, comic upkeep | Import only | Scheduled sync per type |
Plugin Info
| Property | Value |
|---|---|
| Plugin ID | media-db-sync |
| Display name | Media DB Sync |
| Minimum Obsidian version | 1.12.0 |
| Desktop only | No, runs on mobile |
| License | GPL-3.0 |
| Upstream | mProjectsCode/obsidian-media-db-plugin |
| Toolchain | Bun + Vite |
| Build output | dist/main.js, dist/manifest.json, dist/styles.css |
| Network calls | Obsidian requestUrl() only, never node fetch |
Watchlist Notes
A watchlist note is one markdown file per title, carrying type: watchlist_item in its frontmatter. Everything the sync reads and writes lives in that frontmatter; the body holds the rendered sections.
Frontmatter written on every sync
title engName media_type category
watch_status rating rating_stars year
runtime seasons episodes vod
genre status language country
director writer producer content_rating
tmdb_rating tmdb_id imdb_id release_date
last_air_date next_air_date last_episode upcoming_episode
poster trailer homepage imdb_page
notion_url synopsis cast
media_type is Movie or TV Series. category is Movie, Series, or Anime.
Body sections owned by the renderer: ## Synopsis, ## Cast, ## Links, ## My Notes.
Fields the sync never overwrites
watch_status kept, except for the episode rule below
rating kept
rating_stars kept
notion_url kept
## My Notes body text under this heading is carried across verbatim
A note with no watch_status yet starts at Unwatched.
Which notes get synced
Watchlist: sync now skips anything inactive, so a daily run costs a handful of requests instead of one per note. A note counts as active when either of these holds:
status is one of: Returning Series, In Production, Planned, Pilot
watch_status is Watching
Everything else is only touched by Watchlist: full sync. Requests are spaced 250 ms apart by default.
The watch-status rule
if media is a TV series
and watch_status == "Watched"
and TMDB's last_air_date is newer than the last_air_date stored in the note
then watch_status -> "Unwatched"
That is the whole rule. A movie you marked Watched stays Watched, because a movie has no last_air_date that can move.
Anime detection
genre contains "Animation" AND language == "Japanese" -> category: Anime
otherwise, if it is a movie -> category: Movie
otherwise -> category: Series
This is why the blank language field on stock TMDB imports matters. Without it, every anime lands in Series.
Library Types
Each type is a separate folder with its own toggle. All four run on the watchlist's sync interval.
| Type | Default folder | Data sources | API key needed |
|---|---|---|---|
| Manga | Mangas |
Jikan (MyAnimeList), AniList, MangaDex RSS for chapter counts | None |
| Book | Books |
Open Library | None |
| Game | Games |
Steam store, RAWG | RAWG |
| Comic | Comics |
Comic Vine | Comic Vine |
When a resolve pass turns up several plausible matches for one note, it opens a picker instead of guessing. The picker offers Never resolve (mark no_resolve); a note with no_resolve: true is skipped by every future resolve pass and left out of the "missing id" count.
Notes that resolve to nothing still get converted into the canonical shape locally, with no API calls, so they appear in Bases views rather than staying invisible.
Commands
Inherited from upstream, unchanged:
| Command |
|---|
| Create entry |
| Create entry (per media type) |
| Create entry (advanced search) |
| Create entry by id |
| Update open note (this will recreate the note) |
| Update metadata |
| Insert link |
Added here:
| Command | What it does |
|---|---|
Watchlist: sync now (airing/active only) |
Syncs active notes only |
Watchlist: full sync (all entries) |
Syncs every note in the folder |
Watchlist: dry-run full sync (log only, no writes) |
Walks the full sync and logs the diff without writing |
Watchlist: resolve missing TMDB ids |
Searches TMDB for notes that have no tmdb_id |
Library: sync <type> now |
One type |
Library: resolve <type> ids |
Backfills ids for one type |
Library: dry-run <type> full sync |
Log only |
Library: sync all |
Every enabled type |
Run the dry-run against a real folder before the first live sync. If the vault is under git, commit it first.
Settings
Under Watchlist sync:
| Setting | Default | Notes |
|---|---|---|
| Enable watchlist sync | Off | |
| Watchlist folder | Watchlist |
|
| Sync interval (hours) | 24 |
Accepts 1 to 168 |
Under Library sync, per type: an enable toggle and a folder path.
The scheduler is catch-up based rather than a wall-clock timer. On load it checks whether lastSync + interval is already in the past and runs if so, so a laptop that was asleep at the scheduled hour still syncs when it wakes.
API keys are held in the Obsidian keychain, not in data.json. TMDB takes either credential: a token starting with eyJ is sent as Authorization: Bearer, anything else is sent as ?api_key=.
Related Articles
- Upstream Media DB README, for the import features, templates, property mappings, and the per-API "search by ID" reference
- TMDB API docs
- BRAT, for installing a plugin straight from a repo
- Obsidian Bases, which is what the
watchlist_itemshape is written for
Pre-requisites
- Obsidian 1.12.0 or newer
- A TMDB API key or read access token, from your TMDB API settings
- Bun 1.x, if you are building from source
- Git
Clone Repository
git clone git@github.com:afiqzudinhadi/obsidian-media-db-sync.git
cd obsidian-media-db-sync
The upstream remote is kept, so changes from Moritz Jung's plugin can be pulled in:
git fetch upstream
Set Up Build Environment
1. Install Bun
curl -fsSL https://bun.sh/install | bash
This project uses Bun and Vite. It does not use npm or esbuild; older notes that say otherwise are stale.
2. Install Dependencies
bun install
3. Build
bun run build
Type-checks first, then writes main.js, manifest.json, and styles.css into dist/.
4. Run the Checks
bun run check
Formatter check, both type-check passes, ESLint at zero warnings, then the test suite.
Install Into Obsidian
Option A: BRAT
Install BRAT, then add afiqzudinhadi/obsidian-media-db-sync as a beta plugin. BRAT reads manifest-beta.json and updates from releases.
Option B: Copy the Build
mkdir -p /path/to/vault/.obsidian/plugins/media-db-sync
cp dist/main.js dist/manifest.json dist/styles.css /path/to/vault/.obsidian/plugins/media-db-sync/
Reload Obsidian, then enable Media DB Sync under Settings, Community plugins.
The plugin id is media-db-sync, not obsidian-media-db, so this installs alongside the stock plugin and neither one overwrites the other's folder or data.json.
Configure
- Paste the TMDB key into Settings, Media DB Sync.
- Turn on Enable watchlist sync and set the folder.
- Run
Watchlist: dry-run full syncand read the console output. - If the diff looks right, run
Watchlist: full sync.
Development
Live Build Into a Test Vault
bun run dev
In development mode the output goes to exampleVault/.obsidian/plugins/media-db-sync/ instead of dist/, and rebuilds on save.
Tests
bun test --preload ./tests/setup.ts
Use bun run test:log to see the sync engine's log lines. The suite covers the serializer, YAML quoting, the diff-on-write path, the candidate picker loop, and the watch-status rule.
Issues
TMDB API key not configured
TMDB API key not configured (Media DB Sync settings).
The key sits in the Obsidian keychain, not in data.json, so it does not travel with a synced vault. Enter it again on each device.
Watchlist sync already running
Watchlist sync already running
A sync is in progress. Both a manual command and the scheduler can start one, and the second caller is turned away rather than allowed to write over the first. Wait for it to finish. The library sync reports the same way with Library sync already running.
Notes skipped for having no id
The sync report counts notes it could not act on because they carry no tmdb_id. Anything imported before this plugin existed will be in that group. Run Watchlist: resolve missing TMDB ids, then sync again. Notes marked no_resolve: true are deliberate opt-outs and are excluded from the count.
TMDB works in Obsidian but not from a script
nodename nor servname provided, or not known
Some ISP DNS resolvers return nothing for TMDB's API host, which breaks curl and any Python or Node script on that connection. Obsidian's requestUrl() goes through Chromium's network stack and resolves it fine. That is why every network call here goes through requestUrl(), and why an external cron script is not a working substitute on such a connection.
License and Credit
GPL-3.0, inherited from upstream. The import pipeline, API adapters, modals, and property mapping are Moritz Jung's work: see mProjectsCode/obsidian-media-db-plugin. The sync engine, watchlist schema, library specs, and resolve flow are added in this fork.