7.9 KB


This guide explains how to create a setup for running and developing the bot. Whether you want to just run it, or also develop it and test your changes, this guide lists the steps.

Since one of the bot's goals is provide an introductory space for Haskell, it details all the steps of starting to work with Haskell and provides hints. Even if you're just starting, this guide is friendly and useful to follow.

Programming Interface

Here is a quick summary of what the bot API offers. It is being developed and more components will probably be added.

  • Bot commands with multiple prefixes and multiple names
  • Reaction to various events: joins, parts, private messages, etc.
  • Can run in multiple IRC channels (but on a single IRC server)
  • Bot commands can run any I/O (network, files, terminal, etc.)
  • Bot commands can run any IRC commands: Join, part, notice, etc.
  • Persistent state can be managed by the bot commands while the bot runs
  • The state includes a hierarchical central settings system
  • Extra event sources: Web listener, newsfeed watcher


If you already have the Haskell tools installed, you can skip this section.

The bot is written in Haskell, a functional programming language. A good place to start learning it is the Haskell Wikibook. The definition of Haskell is published in the form of reports, the latest being the Haskell 2010 Report. While learning, and actually in general, a very useful resource to keep open in a browser tab is the API reference of the base package, here.

To work with Haskell you'll need 3 primary tools:

  1. GHC - a compiler, interpreter, package manager and more
  2. Cabal - project manager for installing packages and packaging your own
  3. stack - a command-line interface for working with Haskell projects

You'll need to install them both. Preferrably GHC 7.10.3 and a recent release of stack. If you use a Debian based distro, you can install GHC easily from a PPA. Trisquel should import it for you, so all you need to do is apt-get install.

For Parabola, check the versions supplied by the distro's packages. Instructions for more distros can be found online.

Add ~/.cabal/bin:/opt/cabal/1.22/bin:/opt/ghc/7.10.3/bin to your PATH. For example, add this to the bottom of your .bashrc:

# add haskell programs to PATH
export PATH=~/.cabal/bin:/opt/cabal/1.22/bin:/opt/ghc/7.10.3/bin:$PATH

The Haskell community has:

  • IRC channels on Freenode: #haskell, #haskell-beginners
  • Mailing lists
  • A wiki
  • Hackage, where people upload package releases


You can either use a release version, or the latest development version.

The development version many require recent dependency versions which aren't available on Hackage yet. If building fails, it's probably because you need those recent versions. These commands will download the dependencies most likely to be needed. In the same way you can download more.

If you'll be using the release version, there is no need for this.

Install Darcs, a version control system. The bot itself is in a Git repository, but some of these dependencies are in Darcs repositories. You can either download their files using regular HTTP, e.g. with wget) or use Darcs. Since Darcs is a popular version control system for Haskell projects (and is itself written in Haskell), the latter option is demonstrated below (just change the Darcs specific lines if you chose the former option).

If your distro has a recent enough version (at least 2.8.4):

$ sudo apt-get install darcs

Otherwise install Darcs from Hackage/Stackage:

$ stack install darcs-2.10.3

Get the bot code:

$ git clone

You now have the latest development version. You can switch to the latest release version (use git tag --list to find out its number), e.g.:

$ git checkout 0.3


Before running the bot, you'll need to configure 3 places:

1. State data, stored in JSON files in state/

Create initial state data:

$ cp -r state-default state

Some features (like channel logs and quotes) place files in separate subdirectories.

$ mkdir state/chanlogs state/quotes

2. Configuration file config/settings.yaml

First, you'll need to set up postgresql. You'll need one user and one database, owned by that user. Setting this up will vary slightly by operating system.

Then, you should funbot's config file to match your database:

$ cp config/settings-default.yaml config/settings.yaml
$ vim config/settings.yaml

3. Configuration, in a Haskell source file, src/FunBot/Config.hs

Most of the settings moved to config/settings.yaml, but a few still remain in Config.hs. They will soon move too.

If you make a git commit, make sure you don't commit your personal changes.

$ vim src/FunBot/Config.hs

You'll need to rebuild the bot for changes in Haskell files to take effect.

$ stack build

You can also customize the event matching and behavior definitions in src/Main.hs.

Running and Exploring

Now you can do things like running it, debugging it, exploring it in the REPL (i.e. interpreter).

To run the bot:

$ stack exec funbot

To load the source into the GHCi, the REPL, and play/explore it:

$ stack repl

For the explorers, there is an IRC server package in Haskell, hulk. You could run the bot against a locally running instance of the server.


The Git repository contains a shell script you can run in a cron job. There is also a supervisord configuration file. See these files for details. If you want to run the bot as a separate system user (and not as your own user), the steps are:

  1. Create a Linux system user with its own home directory. For example, /var/lib/ircbot
  2. Edit its PATH (e.g. by creating minimal .profile and .bashrc) and install the setup as described above. But you don't need:
    • a Cabal sandbox. You can install to the new user's user package repository. Simply skip the sandbox creation steps.
    • Write access to the git repo. Clone the repositories using read-only access. It's safer not to give this new user write access to its own code, just in case anything bad happens.
  3. Switch to the system user in the terminal and run the bot. Edit the configuration source file (nickname, password, channels, etc.). Setup the cron job.
  4. When there are updates to the bot or its dependencies, darcs pull or git pull them, build/install and relaunch (e.g. pkill funbot && dist/build/funbot/funbot &)


There are basically 3 ways to contribute code:

  • If you prefer to have your code checked by someone else for any reason, you can create a merge request (i.e. work in a clone or branch).
  • Or create a patch with git format-patch
  • Otherwise, this project just gives commit access to any interested community member / contributor.

The details follow.

Merge Requests

The steps are:

  1. Create a branch with your changes
  2. Open an issue containing the URL of the branch

The branch can be in:

  • The upstream funbot repository at NotABug. Feature/wip branches can be added as needed.
  • A repository in NotABug (e.g. your personal clone of funbot) or some other free-software hosting platform instance online. Merge requests from proprietary ones, such as githu8, won't be accepted. Please use NotABug or GNU savannah or Rel4tion's server or your own server or some other free-as-in-freedom option.

Direct Commits

This means access to pushing to the master branch of the upstream repository. Ask fr33domlover.


Open an issue in NotABug with the patch file attached (or post a link to it, if you host it somewhere else).