SaneBB is a small and fast BBCode parser in Java, which I made as an alternative to the horribly overengineered and inner-platform effect victim KefirBB library.

It's a single Java file, the documentation is included at the top of the file.

You can find it here on GitHub.


Serf is a small HTTP server made for local testing. It can serve static files and you can customize the headers with a simple config file.

Here is an example configuration file:

# Set SocketType to Dual (IPv4 + IPv6), IPv4, IPv6 or Unix
SocketType = Dual
Port = 80
# UnixPath is ignored unless SocketType is set to Unix
UnixPath = /run/serf.sock

# Set directory to serve, relative to the location of the INI file
RootDirectory = .
# Enable AutoIndex to list content of directories without index.html
AutoIndex = On
# Maximum cache time in seconds
MaxAge = 0
# Generate E-tag based on file modification time and size
ETag = On

# List headers you want to add to all server responses
Cross-Origin-Embedder-Policy = require-corp
Cross-Origin-Opener-Policy = same-origin

You can find out more in the code repository.


SeatSH is a Windows service that enables you to launch graphical applications from an SSH connection on Windows 10. As an example, try to start notepad from an SSH connection: it won't work, because this session does not have access to the active console (desktop) session.

SeatSH runs as a service, waits for a client to connect, and launches the command line specified by the client on the active console session. Once the service is installed, launching a graphical application from the command line is as simple as seatsh notepad.

It will transparently forward standard streams (stdin, stdout, stderr) and report the application exit code as its own.

It is mainly useful for automated test systems, do not install it on production machines because it is a security risk. SeatSH only makes sure that the user connected to the SSH matches the one on the active console session, nothing more.

You can find out more in the code repository.


The DRD project strives to provide a fully-featured and up-to-date alternative french PMSI MCO classifier.

Using the data files (.tab) provided with GenRSA, it can classify MCO stays from 2013 to 2022 several hundred times faster than the official classifier (on a 4 core computer from 2015, it can classify around 4 million stays per second).

It can be used in three ways:

The classifier was carefully assembled from the documentation and through retro-engineering of the official classifier (GenRSA), without source code access.

You can find out more in the code repository.