🎹 Ludwig

Ludwig is a Lemmy-compatible Reddit-style link board and forum server, available as a zero-configuration single binary.

The current release of Ludwig is an alpha-quality MVP. Many features are missing. Major bugs almost certainly exist. Breaking changes to the database are possible, and will require exporting and re-importing the entire database to upgrade.

MVP Features

A Reddit-style link board with posts, comments, voting, link previews, and user-creatable sub-boards.
Minimal web interface is usable without JavaScript; users or admins can disable JS entirely.
User signups can be restricted with admin approval or invite codes.
The Lemmy API is supported: Lemmy apps can interact with the site, read, vote, and post.

Usage

Currently you have to build Ludwig manually; I recommend using the Docker container (docker/Dockerfile) to build and run Ludwig, since there are a lot of specific dependencies and the current build does not produce a fully static executable.

Run task build-release from within the Docker container, then copy the ludwig binary from build.release to somewhere else and run it.

Ludwig will create two .mdb files in its directory by default: ludwig.mdb contains the site database, and search.mdb is a full-text search index.

By default, Ludwig serves a web interface on port 2023. Just put this behind a reverse proxy, and it should be ready to go.

Options

Ludwig does not have a configuration file, but it does accept a few command-line options.

--setup: Run an interactive first-run setup wizard on the terminal, instead of in the web interface. Can only be used when the site is not yet set up.
--log-level: One of debug, info, warn, error, or critical. Default is info. Logs are currently written to the terminal; future versions will support logging to files.
-p N, --port=N: Sets the port for the web interface to N.
-s N, --map-size=N: The maximum LMDB database size, in MiB. Default is 4096 (4 GiB). Currently the server will crash if this database size is exceeded; future versions will handle low space more gracefully. Also sets the maximum size of the search database, if applicable.

Can be safely increased, but not decreased.
--db=FILE: The database filename; will be created if it does not exist. Default is ludwig.mdb.
--search=PROVIDER: Full-text search configuration. Default is lmdb:search.mdb. Potential providers are:
- none - Full-text search is disabled.
- lmdb:FILE - Use sentencepiece to build an LMDB search index of tokens. This approach is very limited: it only has token mappings for English, and it can only check whether posts contain all of the searched tokens, but not the tokens' order. FILE is the filename of the search database.
More search providers will be added in the future.

Note that, currently, switching search providers will only index future posts; if you change from none to lmdb or switch lmdb files, the new search index will not contain any posts from before the switch.
-r N, --rate-limit=N: Maximum HTTP requests allowed per 5 minutes from a single IP. POST and PUT requests count as 10 requests. Default is 3000.
-t N, --threads=N: Number of request handler threads. Default is the number of physical cores on the machine.
--export=FILE: Exports the database to a dump file, then exits.
--import=FILE: Imports a database from a dump file, then exits. The database (--db) must not exist yet; import will refuse to run if it would overwrite an existing database.

FAQ

Why clone Lemmy?

A few reasons:

I like the idea of a Fediverse forum and a Reddit clone, but Lemmy is annoyingly difficult to set up.
I wanted a few specific features, like single-board sites and old-school forum homepages.
Lemmy is AGPL. I prefer permissive licenses; many developers (myself included) are reluctant to even read AGPL-encumbered code for fear of legal liability if we then write something similar in future projects.

Why C++, when Lemmy is already written in Rust?

Honestly, it's because I wanted to try a complex project to finally learn C++. I've already written a lot of C, so I'm familiar enough with manual memory management and the related pitfalls. C++ was easier in some ways, harder in others.

I knew I wanted to use LMDB, and that requires a language with manual memory management. I hadn't used C++ or Rust before, and, since this is a clean-room reimplementation of Lemmy, I didn't want to use Rust and likely duplicate a lot of Lemmy's code anyway. uWebSockets is also cool, and can be incredibly fast.

Why "Ludwig"?

Because, unlike Lemmy, my project is named after the best Koopaling.

ar-nelson/ludwig