## Description:
- The lobby WebSocket broadcast (`/lobbies`) was re-sending the full
`PublicGames` snapshot — including each lobby's `gameConfig` — to every
connected client every 500ms. Almost nothing in that payload changes
tick-to-tick; only `numClients` moves.
- `WorkerLobbyService` now tracks the sorted set of `gameID`s it last
sent as a full snapshot. On each incoming broadcast it sends a `full`
only when that set changes; otherwise it sends a `counts` delta carrying
just `{gameID → numClients}`.
- This relies on the master-side coupling at
[MasterLobbyService.ts:140-159](src/server/MasterLobbyService.ts#L140-L159):
when master finds a lobby without `startsAt`, it both sets `startsAt`
AND schedules a fresh lobby on the same tick, so the gameID change
brings the `startsAt` (and `gameConfig`) along with it.
- New WS connections are primed with the worker's cached last `full` so
late joiners don't have to wait for the next structural change.
- `LobbySocket` parses the new discriminated union (`PublicLobbyMessage
= full | counts`), keeps the last full snapshot in memory, and merges
counts into it before invoking the existing callback. `GameModeSelector`
is unchanged.
- Master → worker IPC is unchanged — still sends the full snapshot every
500ms. The optimization only applies to the worker → WS-client boundary,
which is the fan-out point.
## Please complete the following:
- [x] I have added screenshots for all UI updates
- [x] I process any text displayed to the user through translateText()
and I've added it to the en.json file
- [x] I have added relevant tests to the test directory
## Please put your Discord username so you can be contacted if a bug or
regression is found:
evan
## Description:
This is a refactor to simplify config handling.
Replaces the per-environment DevConfig/PreprodConfig/ProdConfig class
hierarchy with two static classes: ClientEnv (browser main thread, reads
from window.BOOTSTRAP_CONFIG) and ServerEnv (Node server, reads from
process.env). The four config classes are deleted, the abstract
DefaultServerConfig is gone, and DefaultConfig is renamed to Config.
The values that flow server → client (gameEnv, numWorkers,
turnstileSiteKey, jwtAudience, instanceId) used to be baked into the
hardcoded per-env classes. They're now real env vars on the server,
embedded into a single window.BOOTSTRAP_CONFIG object in index.html at
request time (alongside the existing gitCommit/assetManifest/cdnBase
globals, which moved into the same object), and read back by ClientEnv
on the client. The dev defaults previously hidden inside DevServerConfig
are now explicit in start:server-dev (NUM_WORKERS=2,
TURNSTILE_SITE_KEY=1x..., JWT_AUDIENCE=localhost, etc.) and in
vite.config.ts's html plugin inject.data. Production deploys plumb
NUM_WORKERS and TURNSTILE_SITE_KEY through deploy.yml (GitHub vars) into
the remote env file; JWT_AUDIENCE is derived from DOMAIN in deploy.sh.
The dynamic /api/instance endpoint is gone — INSTANCE_ID rides along in
BOOTSTRAP_CONFIG now.
ServerEnv is the only thing server code touches; ClientEnv is
browser-only. The two classes have intentional overlap (env, numWorkers,
jwtIssuer, gameCreationRate, workerIndex, etc.) since they derive
identical logic from different sources — there's a TODO in each to
consolidate via a shared helper later. The game-logic Config no longer
stores a ServerConfig/ClientEnv reference and its serverConfig() getter
is gone; the one caller (MultiTabModal) now reads ClientEnv.env()
directly. Worker init no longer carries server-config values since
nothing in the worker actually reads them.
## Please complete the following:
- [x] I have added screenshots for all UI updates
- [x] I process any text displayed to the user through translateText()
and I've added it to the en.json file
- [x] I have added relevant tests to the test directory
- [x] I confirm I have thoroughly tested these changes and take full
responsibility for any bugs introduced
## Please put your Discord username so you can be contacted if a bug or
regression is found:
evan
## Description:
This refactors client configuration loading to make the environment
split explicit.
The app currently has two different env concerns:
- the browser main thread needs the live runtime env to select API /
Turnstile / JWT settings
- the worker and game-logic path need a build-time env to select game
config behavior
Before this change, both responsibilities were hidden behind the same
loader, which made the intent unclear and caused confusion around the
worker fallback behavior.
This PR separates those paths explicitly:
- main-thread browser code now uses `getRuntimeClientServerConfig()`
- game creation and worker/game-logic code now uses
`getGameLogicConfig()`
- the build-time game-logic env is represented explicitly as
`GameLogicEnv`
## What Changed
- Added `GameLogicEnv` to model the build-time game config choice
explicitly.
- Added `getRuntimeClientServerConfig()` for live runtime browser config
from `window.BOOTSTRAP_CONFIG`.
- Added `getBuildTimeGameLogicEnv()` and
`getServerConfigForGameLogicEnv()` for build-time worker/game-logic
config.
- Renamed game config loading from `getConfig()` to
`getGameLogicConfig()` to reflect what it actually does.
- Updated browser call sites to use the runtime client config loader.
- Updated worker/game creation paths to use the game-logic config
loader.
- Updated config loader tests to cover both paths.
## Behavior
This keeps the current intended behavior, but makes it explicit:
- Runtime client env:
- comes from `window.BOOTSTRAP_CONFIG`
- controls live browser integration settings such as API origin,
Turnstile, and JWT audience/issuer
- Build-time game-logic env:
- comes from bundled `process.env.GAME_ENV`
- maps:
- `dev` -> dev game config
- `staging` -> default/prod game config
- `prod` -> default/prod game config
That means preprod/staging deployments can continue using prod game
logic while still using staging API/auth settings on the main thread.
## Why
The previous setup worked, but the naming and loader boundaries were
misleading:
- the same function was used for both runtime browser config and
worker/game config
- the worker fallback looked like an implementation detail instead of an
intentional architectural split
This change makes that intent visible in code without changing the
desired deployment behavior.
## Please complete the following:
- [ ] I have added screenshots for all UI updates
- [ ] I process any text displayed to the user through translateText()
and I've added it to the en.json file
- [ ] I have added relevant tests to the test directory
- [ ] I confirm I have thoroughly tested these changes and take full
responsibility for any bugs introduced
## Please put your Discord username so you can be contacted if a bug or
regression is found:
DISCORD_USERNAME
## Description:
Calling `lobbySocket.stop()` closes the WebSocket, but the async `close`
event fires afterward and triggers `handleClose()` →
`scheduleReconnect()`, reopening the connection. In singleplayer, both
`onPrestart` and `onJoin` callbacks fire synchronously in a single
stack, so the second `stop()` call has nothing left to clean up - the
reconnect timeout gets set after both calls complete.
Adds a `stopped` flag that skips reconnection in `handleClose()` when
the socket was intentionally closed.
Surely this PR description is not AI generated.
Should reduce network load of the server because all the singleplayer
gamers no longer stay connected to the lobbies websocket forever. They
also no longer load all the map mainfest.jsons and thumbnails while they
are playing alone (rotation running in the background).
- [X] I have added screenshots for all UI updates
- [X] I process any text displayed to the user through translateText()
and I've added it to the en.json file
- [X] I have added relevant tests to the test directory
- [X] I confirm I have thoroughly tested these changes and take full
responsibility for any bugs introduced
## Please put your Discord username so you can be contacted if a bug or
regression is found:
FloPinguin
## Description:
Currently only the master process sends public lobby updates to clients.
This is not scalable since it could overload the master process.
In this PR, the master uses IPC to send public lobby info to all
workers. Then clients connect to a random worker to get public lobby
updates via websocket. This way clients never connect directly to the
master websocket.
The flow looks like this:
Every 100ms:
1. Master schedules a public game on a random worker if new games are
needed
2. Master broadcasts public lobby info to all workers (all public games
& num clients connected to each game)
3. Each worker responds to that update with the number of clients
connected to its own public games
4. Master then updates its public lobby state so it knows how many
clients are connected to each public game
## Please complete the following:
- [x] I have added screenshots for all UI updates
- [x] I process any text displayed to the user through translateText()
and I've added it to the en.json file
- [x] I have added relevant tests to the test directory
- [x] I confirm I have thoroughly tested these changes and take full
responsibility for any bugs introduced
## Please put your Discord username so you can be contacted if a bug or
regression is found:
evan
## Description:
Changes game lobbies into websockets instead of polling
## Please complete the following:
- [x] I have added screenshots for all UI updates
- [x] I process any text displayed to the user through translateText()
and I've added it to the en.json file
- [x] I have added relevant tests to the test directory
- [x] I confirm I have thoroughly tested these changes and take full
responsibility for any bugs introduced
## Please put your Discord username so you can be contacted if a bug or
regression is found:
w.o.n
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: iamlewis <lewismmmm@gmail.com>
Evan