watchparty/README.md
Nicolas b4c81d820c
All checks were successful
Publier l'image Docker / test (push) Successful in 20s
Publier l'image Docker / publish (push) Successful in 16s
Cast a party's video to a Chromecast
Implements a detailed spec (Google Cast / Chromecast integration) asking
for a member to cast the party's video to a Chromecast's Default Media
Receiver, with play/pause/seek shown as a remote on the sender's own
screen. The spec's description of this app's architecture (one shared
Emby stream per party, the server proxying 100% of Emby access) was
verified directly against the real code before building anything, since
it named specific existing services/routes as the integration points.

A Chromecast is a new kind of client for the existing party-stream/burn
relay: it can't carry the sender's httpOnly session cookie at all, so it
needed its own narrow way in without weakening "the server holds the
only real Emby token" the rest of the app is built around.

- New CastTokenService (same shape as AuthService: in-memory Map, opaque
  crypto.randomUUID() keys, TTL-checked) issues a token scoped to one
  {partyId, userId}, standing in for that member's own EmbySession.
  New POST/DELETE /api/parties/:partyId/cast-token to mint/revoke one.
- The 4 routes a Chromecast actually needs (party-stream/master.m3u8,
  party-stream/raw, burn/master.m3u8, burn/:segment) gained one small
  fallback (sessionForStream in index.ts): try the session cookie first,
  same as ever, then a castToken query param - either way the result is
  the same EmbySession type, so PartyStreamService/TranscodeService/
  EmbyClient needed zero changes. The image/subtitle routes stay
  cookie-only, outside the spec's own 4-route list.
- PartyService.leave() now revokes whatever cast token the leaving
  member held, unconditionally - a token dies the instant its owner
  actually leaves the party regardless of TTL or whether the client got
  a chance to call the explicit stop-casting route.
- New client/src/cast.ts is the only module touching the CAF Sender SDK
  (loaded as a pinned CDN script, same pattern as hls.js/socket.io) -
  exposes a small PlaybackDriver interface (currentTime/paused/play/
  pause) with two implementations, one over the local <video>, one over
  CAF's RemotePlayer/RemotePlayerController. This was the only new
  abstraction the feature needed, not a new sync mechanism: Player.ts's
  existing sendLocalState/applyRemoteState already read and write "the
  player" in one place each, so pointing them at whichever driver is
  active carried the host-only gate, drift-tolerance reconciliation, and
  3s heartbeat over unchanged. Seek was already "the host's next
  broadcast" with no separate event, so seeking on the TV is party-wide
  for free; a non-host caster already only ever received state, so
  redirecting that to the cast driver reproduces "still dependent on the
  host" automatically.
- Cast button in the player's action bar, shown to any member (not just
  the host) once cast.framework reports a receiver on the network.
  Starting a cast tears down the local hls.js instance; ending one
  (user-stopped, or the receiver dropping off on its own) resyncs local
  playback via the same loadVideo path an audio/quality change already
  uses. A real reload for any other reason (item change, burn-in/
  subtitle/stream-options change) simply ends an active cast first,
  rather than keeping a live session in sync with every kind of party
  transition - not worth the complexity for v1.

Scoped to video parties only - the spec's every concrete reference is
video-specific, and MusicBar.ts has no comparable control surface today.

Tests: CastTokenService (issue/resolve/reissue-invalidates-the-old-one/
revoke/wrong-party-rejection), PartyService.leave()'s new call. Full
typecheck/test/build clean on both projects (115/115 server tests).
Confirmed against the real running server that the 4 streaming routes
still reject a request with no auth exactly as before, and now also
reject an invalid castToken - both a plain 401, no regression. Confirmed
in a real browser that the Cast SDK loads and initializes correctly. Not
verified: no real Chromecast or Emby login credentials in this
environment to take a cast session, or the button's own live behavior,
any further - see docs/architecture.md's "Casting to a Chromecast"
section.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
2026-09-15 10:13:12 +02:00

315 lines
18 KiB
Markdown

# Watch Party for Emby
A standalone web app for watching the same title, in sync, with other users
of the same Emby server. It reads your Emby catalog and streams your Emby
media, but it's its own app with its own player — not an Emby plugin.
See [docs/architecture.md](docs/architecture.md) for the full design,
including why this replaced an earlier Emby-plugin attempt (still in
[src/WatchParty/](src/WatchParty/) for reference — see below).
## How it works
1. Sign in with your Emby account (just username/password — the server
address is fixed server-side, see Deploying below).
2. A sidebar lists your Emby libraries (movies, TV, music, music videos,
playlists if you have any — anything else Emby might have, like
photos or a boxset collection, isn't shown); the home page shows every ongoing
watch party live (host, how long it's been running, member count) plus a
"recently added" row per library, Emby-style. No join codes — if a party
is running, it's already on the list, updating in real time.
3. Pick a movie/episode tile and hit "Lancer une watch party" (or drill into
a Series' seasons/episodes first) — you become the host — or hit
"Rejoindre" on any party already listed on the home page. Already
connected to that same party from another device? You're asked to
confirm before it gets disconnected — reconnecting from the same
device (a reload, a second tab) never asks.
4. The host's play/pause/seek drives everyone else's player automatically;
starting from an episode queues the rest of the series so the host can
just hit "Épisode suivant" instead of restarting a party each time. If
the host leaves, the party hands off to another member automatically
(their player unlocks live, no reload needed) — only closing once
everyone's gone. Everyone also shares the exact same stream from
Emby — one transcode per party, not one per person — so the host's
language/quality pick applies for the whole party too.
5. Everyone present shows up as a small avatar overlay on the video (their
initials, on the player itself). If you're not the host, a small
toolbar lets you ask the host to pause or resume — everyone sees the
request, including the host, without needing chat.
6. Subtitles are a host-only, party-wide pick too, like the
language/quality pick above — the host picks one (or "Aucun") and
everyone's player follows. Any text-based track just renders
natively in each viewer's own browser, in sync (no reload, no extra
server load — only *which* one is synced). A Blu-ray/DVD-style image
subtitle (PGS/VobSub/DVB) is different — picking one has this app's
own ffmpeg burn it into the video, shared live by everyone in the
party, which does restart the stream for everyone since the video
itself changes.
7. Music works differently, on purpose: pick a track (browsing Artist →
Album → Track) to start a listening party. There's no host for
playback — anyone can add more tracks to the shared queue at any time,
anyone can pause/resume, and skipping the current track needs a
majority vote from whoever's currently listening. The player itself is
a persistent bar docked at the bottom of the screen, not a page — the
sidebar and the rest of the app stay fully usable while it plays, so
adding more tracks is just browsing to an album and clicking a track
(its button reads "Ajouter à la file" instead of "Écouter ensemble"
once a party's already going). Starting or joining a video watch party
always leaves the listen party first — the two don't run at once.
8. A playlist works the same idea as a series: open it to see what's
inside, then "Lancer cette playlist" starts one party pre-loaded with
everything in it — a video queue if it's movies/episodes, or the
participative music queue if it's tracks.
9. While you're in a party, your own progress is quietly pushed back to
your own Emby account (resume position, last-played state) — so
opening Emby's own app afterward reflects what you actually watched
or listened to, not just this app's own memory of it.
Everything — catalog, images, video, audio, sync — flows through this
app's own backend, which proxies whatever it needs from Emby server-side.
The browser never talks to Emby directly and never holds an Emby access
token; it only ever has a session cookie for this app.
## Status
Login, and the rendering/navigation logic for every page (sidebar, home,
library, series, artist/album, both players), have been exercised end to
end (login against the real Emby server with intentionally-wrong
credentials; the pages against mocked API responses; the music queue's
timer/vote/pause logic against the real server modules directly - no Emby
involved there at all). See
[docs/architecture.md](docs/architecture.md#assumptions-to-verify) for
exactly what that does and doesn't confirm — the real Emby response shapes
this now depends on (library `CollectionType`, seasons/episodes, artists/
albums/tracks, `PlaybackInfo`'s media streams) and actual HLS playback
(video or audio) are still unverified.
The TypeScript rewrite itself is separately verified: both projects
type-check in strict mode, the server has a real automated test suite
(`server/test/*.test.ts`, Node's built-in test runner, see Running
locally below) covering the party lifecycle, the music queue's
timer/vote/pause behavior, the socket protocol, and the catalog/streaming
overrides, and the compiled+bundled client was driven through every page
and both players end to end (mocked network/Socket.IO) in a real browser.
CI (Forgejo Actions) runs that same typecheck+test+build sequence on
every push and gates the Docker image publish on it passing — see
[docs/architecture.md](docs/architecture.md#continuous-integration).
**Not verified**: an actual `docker build` of the new multi-stage
Dockerfile - this dev environment has no Docker - so treat the first real
build as the thing to double-check, even though it mirrors exactly the
`npm run build` commands already confirmed working locally for both
projects. Same caveat for the self-hosted ffmpeg burn-in pipeline (see
docs/architecture.md): no ffmpeg binary is reachable here, so while its
pure logic (stream-index conversion, session lifecycle) is unit-tested
with `child_process.spawn` mocked, the actual command against a real
PGS file has never run in this environment. Treat the first real
attempt at a Blu-ray-style subtitle as the thing to double-check here.
**Resolved**: three Emby-API assumptions that used to be listed here as
unconfirmed (the ffmpeg direct-stream URL, the Playlists endpoint, the
Emby-playstate sync endpoint) were all checked directly against Emby's
own real OpenAPI spec. Two matched exactly. The third - playstate sync
- took three rounds: two request-shape bugs against Emby's
`PlayingItems/Progress` check-in endpoint (a missing `MediaSourceId`,
then a missing `PlaySessionId`), then a further report ("still don't
see this in my Emby profile") that questioned the whole approach - a
bare `Progress` report with no matching "Start" call may just never
reach the resume-position/watched-history data real clients read,
regardless of request shape. Switched entirely to a plain
`POST /Users/{userId}/Items/{itemId}/UserData` instead - confirmed
directly against the real server that this is exactly the data model
"Continue Watching" and the watched checkmark read, updated with no
playback session required at all - and added marking an item Played
(the other half of the same report) when the party moves past it.
Not yet confirmed end to end against a real server (no safe way to
fire a real, state-mutating test call without one) - see
docs/architecture.md's "Sync watch-party progress back to Emby"
section for the full story. The one-shared-stream-per-party work is a
different kind of unverified: its own Emby API calls were already live,
confirmed code, just newly shared instead of per-viewer - what's untested is behavior
under real multi-viewer load (whether a dead Emby session actually
surfaces as an HTTP failure the way the cache-invalidation logic
assumes, how a real transcode's seek handling feels once several
viewers depend on the exact same one).
Casting to a Chromecast (see docs/architecture.md) is the newest,
least-verified piece: the security-critical parts were checked against
the real running server directly (the party-stream/burn routes reject a
request with no auth exactly as before, and now also reject an invalid
`castToken`, both a plain 401) and the Cast SDK itself was confirmed to
load and initialize in a real browser, but there's neither a real
Chromecast nor Emby login credentials in this environment to take it
any further - the actual on-TV playback and the cast button's own
live behavior are unconfirmed.
## Project layout
Both `server/` and `client/` are TypeScript now (see
[docs/architecture.md](docs/architecture.md#typescript-and-a-service-oriented-architecture)
for the why and the migration itself) - each its own project with its own
`package.json`/`tsconfig.json`/build step; `server/src` and `client/src`
are the only source of truth, `server/dist`/`client/dist` are generated
(gitignored) and never hand-edited.
```
server/ Node.js + Express + Socket.IO backend: proxies ALL Emby
access (auth, catalog, images, HLS streaming) and
coordinates party/room state + realtime sync. The
browser only ever talks to this. TypeScript, compiled by
tsc to server/dist.
src/index.ts Express app + route wiring only - thin,
delegates to the services below
src/config.ts env-derived config (EMBY_SERVER_URL, PORT)
src/types.ts shared interfaces (Party, ItemDto, the
Socket.IO event maps, ...)
src/Logger.ts party/stream lifecycle logging
src/emby/EmbyClient.ts low-level Emby HTTP + response proxying (incl. HLS)
src/services/AuthService.ts server-side login + session store
src/services/CatalogService.ts all Emby browsing/streaming + DTO shaping
src/services/PartyService.ts in-memory party lifecycle (video queue, host handoff)
src/services/MusicQueueService.ts participative music queue - see docs/architecture.md
src/realtime/SocketGateway.ts Socket.IO protocol (see comments in the file)
client/ Frontend: TypeScript, esbuild-bundled to client/dist
(one app.js + one app.css - see client/build.mjs).
Talks only to server/ - never to Emby directly. Each
view is its own folder colocating its .ts/.html/.css.
src/app.ts entry point - registers routes, boots the app
src/router.ts tiny hash router (unchanged in spirit - see docs)
src/types.ts shared interfaces (client-side copies of the
server's DTO shapes - see docs/architecture.md)
src/api/embyApi.ts calls our own /api/catalog/* + auth routes
src/api/party.ts our own backend's party REST + Socket.IO client
src/tiles.ts + tiles.css poster-tile rendering shared by home/library/artist
src/styles/ base.css (global reset/vars/primitives) +
small shared component styles (presence,
detail header, track list)
src/shell/Shell.* persistent sidebar + #mainContent the router swaps
src/musicBar/MusicBar.* persistent bottom-bar listen party - NOT a route,
mounted once by Shell.ts so it survives navigation
(see docs/architecture.md's "Music bar" section)
src/views/login/Login.* sign-in form
src/views/home/Home.* live parties carousel + per-library "recently added"
src/views/library/Library.* one library's tile grid (or artist grid, for music)
src/views/series/Series.* Series detail + season tabs + episode list
src/views/artist/Artist.* one artist's albums
src/views/album/Album.* one album's tracks - each track starts or queues
onto the active listen party (see musicBar/ above)
src/views/player/Player.* the /party/:id route's dispatcher (video takeover,
or hands off to musicBar/ + bounces back for music)
src/WatchParty/ The earlier Emby-plugin attempt (C#, unrelated to the
TypeScript above). Builds fine and installs, but Emby
has no supported extension point for a player button,
and even menu integration required reverse-engineering
undocumented internals - see docs/architecture.md for
the full story. Kept for reference; not the active
direction.
docs/architecture.md full design doc for the standalone app
```
## Running locally
Both projects need `npm install` once; the server needs the client
already built at least once (it serves `client/dist` as static files).
```bash
cd client && npm install && npm run build && cd ..
cd server && npm install && npm run build
EMBY_SERVER_URL=https://your-emby-server npm start
```
Then open `http://localhost:3000`.
For active development, `client`'s `npm run watch` rebuilds on save, and
`server`'s `npm run dev` runs the server straight from TypeScript via
`tsx` (no separate compile step needed while iterating) - run both in
separate terminals.
Server tests (`server/test/*.test.ts`, Node's built-in test runner, no
extra dependencies or compile step):
```bash
cd server && npm test
```
`server`'s `npm run typecheck` also covers `server/test/`, not just `src/`
(see `tsconfig.test.json`).
## Deploying
Build the Docker image from the **repo root** (it needs both `server/` and
`client/`) - the Dockerfile compiles both TypeScript projects itself in a
build stage, so nothing needs building beforehand:
```bash
docker build -f server/Dockerfile -t watchparty .
docker run -d --name watchparty -p 3000:3000 -e EMBY_SERVER_URL=https://theater.media.kyuno.fr watchparty
```
Then put it behind your existing reverse proxy like everything else
(`theater.media.kyuno.fr`, `git.dev.kyuno.fr`, ...), e.g.
`watchparty.kyuno.fr` → `http://<container>:3000`.
## Roadmap
- [x] Design + scaffold the standalone app
- [x] Proxy all Emby access server-side; Emby-themed UI (sidebar, carousels)
- [x] TV show navigation (Series → Seasons → Episodes) + episode auto-queue
- [x] Player language/subtitle/quality pickers (all three are now
host-only, party-wide choices — see docs/architecture.md's "One
shared Emby stream per party" and "Text subtitles join burn-in as
a party-wide, host-only pick")
- [x] Socket.IO instead of raw WebSockets; fixed two real memory leaks
(dead-connection cleanup server-side, view teardown on navigation
client-side) — see docs/architecture.md
- [x] Live "watch parties en cours" list (host, elapsed time, member count),
live host handoff, no join codes
- [x] Presence overlay (avatars) + a request toolbar (pause/resume) in the
player — see docs/architecture.md
- [x] Device detection: reconnecting from the same device is silent, a
genuinely different device asks to confirm before disconnecting the
other one — see docs/architecture.md
- [x] Server-side logging of the party lifecycle (create/join/leave/destroy,
host handoff, play/pause, pause requests) — see docs/architecture.md
- [x] Music: dedicated player, Artist/Album/Track browsing, participative
queue (free add, vote-to-skip, no host) — see docs/architecture.md
- [x] Full TypeScript rewrite (client + server), component-style client
views (.ts/.html/.css each), server-side singleton services per
feature — see docs/architecture.md
- [x] Automated server test suite (`node:test`) + CI job gating the Docker
publish on typecheck+test+build passing — see docs/architecture.md
- [x] Music player as a persistent bottom bar instead of a page — browse
and queue tracks while listening; starting a video party always
leaves the listen party first — see docs/architecture.md
- [x] Self-hosted ffmpeg burn-in for PGS/image subtitles (Emby's own
transcode can't render these as text) — one shared stream per
party, host-driven — see docs/architecture.md
- [x] Browse Emby playlists and launch one as a pre-queued party (video
queue or the participative music queue) — see docs/architecture.md
- [x] Sync watch-party progress back to Emby's own PlaystateService, per
member's own account, including marking an item watched once the
party moves past it — see docs/architecture.md
- [x] One shared Emby stream per party instead of one per viewer — see
docs/architecture.md
- [x] Text subtitle choice joins burn-in as a host-only, party-wide pick
(was still each viewer's own local choice until now); fixed the
audio picker briefly showing an empty option on first load — see
docs/architecture.md
- [x] Fixed catalog browsing feeling slow: image thumbnails were never
cacheable (Emby's own Cache-Control/ETag were silently dropped) —
see docs/architecture.md
- [x] Cast a party's video to a Chromecast (Default Media Receiver),
with a play/pause/seek remote on the sender's own screen — see
docs/architecture.md
- [ ] Try a full party end-to-end with two real accounts (still the main gap)
- [ ] Try casting to a real Chromecast (only the SDK's own bootstrap is
confirmed here — no real device/session)
- [ ] Tune transcoding target / support direct play when possible
No chat, no OIDC — deliberately out of scope (Emby-only auth was the
explicit choice; see docs/architecture.md).
## License
MIT — see [LICENSE](LICENSE).