home | introduction | examples | features | manual | site map | contact

brat installation

This page describes how to install a brat server and its third-party dependencies.

(If you already have a running server, you may wish to read about annotation configuration next.)

Please note: you do not need to install brat to try it out; brat runs in your browser and a live demo system usable with any supported browser is linked form the brat home page.

If you wish to set up your own local brat installation for your annotation project, please follow the instructions below.

Table of contents

Quick start

The following brief technical instructions present a quick minimal way to set up the brat server. If you are not sure if your setup fills the technical requirements, please skip this section and read the detailed instructions below.

The brat server is a Python (version 2.5+) program that runs by default as a CGI application, and the installation script assumes a UNIX-like environment. If you are setting up brat server in a compatible environment with an existing web server that supports CGI, the quick start instructions for using CGI should work. If you don't have a web server installed and want to try brat out locally, you may want try out the quick start instructions for the standalone server. (For security reasons, we strongly recommend serving brat via a full web server such as Apache in production environments.)

First, download a brat package from the brat home page.

Then, unpack the package (e.g. in the public_html subdirectory of your home directory)

tar xzf brat-v1.3_Crunchy_Frog.tar.gz

Next, enter the created brat directory

cd brat-v1.3_Crunchy_Frog

The next step depends on whether you intend to run brat in a CGI environment or as a standalone server.

Quick start installation: using CGI

Run the installation script

./install.sh

and follow the instructions. You will be prompted for the following information:

Finally, if you are not running the installation as superuser, the script will prompt you for the sudo password. (This is necessary to assign write permissions to the work and data directories to the web server.)

If this works, the brat installation is complete. Now that brat is installed you will have to install and configure a CGI-capable web server if one is not already available on your system. You then only need to make sure that your web server permits CGI for the brat directory (see the Apache 2.x section below for instructions).

Quick start installation: standalone server

First, please note the following:

Run the installation script in "unprivileged" mode

./install.sh -u

and follow the instructions (as above). Then, start the standalone server

python standalone.py

You should then be able to access the brat server from the address printed out by standalone.py. For more detail, see the standalone server instructions below.

Quick start: next steps

After initial installation, you may want to set up some data of your own in your brat installation. For this, look in the Placing Data section below.

If you want to add additional users, you can edit the config.py file, which contains further instructions.

If you feel that your installation requires a more fine-tuned configuration, please see the section Manual Installation below.

If you encountered any issues with these quick start instructions, read on for detailed instructions. You may also wish to check the troubleshooting page.

Detailed Instructions

Preliminaries

The brat server is implemented in Python, and requires version 2.5 (or higher in the 2.x series) of python to run. If you do not currently have python installed, we recommend installing the most recent version of python in the 2.x series from python.org.

In its standard setup, brat is a served through a web server. If you are installing brat server on a machine that is not currently running a web server, you should consider starting by installing one. brat is currently developed against Apache 2.x, and we recommend using Apache.

For serving brat via a web server, These instructions primarily assume Apache, but we do have also LigHTTPD configuration files in the repository if you prefer this alternative, and any CGI-capable web server can be configured to run the brat server.

Alternatively, you may wish to consider running brat using the experimental standalone server included in version 1.3 and above. In this case, no web server is required.

Finally, these installation instructions and parts of the brat server assume a UNIX-like environment. These instructions should work without changes on most linux distributions. If you are running OS X (mac), please note a couple of OS X-specific points in the instructions below. If you are running windows, the easiest way to run a brat server is in a virtual machine running a UNIX-like operating system such as Ubuntu.

Manual Installation

First, download a brat package from the brat home page.

Then, unpack the package

tar xzf brat-VERSION.tar.gz

(where VERSION is the version string, like "v1.1_Albatross"). Next, enter the created brat directory

cd brat-VERSION

Setting up working directories

The brat server needs to read and write data to several directories. Let's create them:

mkdir -p data work

We now need to set the permissions of the directories so that they can be read and written by the web server. We recommend doing this by assigning the directories to the apache group and setting group permissions. To do this, you will need to know the apache group. The following script is provided for convenience for determining this

./apache-group.sh

(if this doesn't work, see the "Finding Your Apache 2 Group" section of this document).

Then simply change the group of the directories (replace ${YOUR_APACHE_GROUP} with the output you got above) and set the correct permissions:

sudo chgrp -R ${YOUR_APACHE_GROUP} data work
chmod -R g+rwx data work

If you can't succeed with the above or you are not concerned with security (say that it is a single-user system), you can run the command below instead. This will make the directories write-able and read-able by every user on your system. (This is not necessary if the command above succeeded.)

chmod 777 data work

Setting up dependencies

brat uses JSON for client-server communication. As JSON support is not included in all supported versions of python, you may need to set up a JSON library.

First, you can determine the version of python you are running with

python --version

if your python version is 2.5 or lower, you will need to install an external JSON library. As of v1.2, brat supports two JSON libraries, UltraJSON and simplejson. You only need to set up one of these, and UltraJSON is recommended as it is considerably faster.

To install UltraJSON, run the following in the brat root directory:

( cd server/lib/ && unzip ujson-1.18.zip && cd ujson-1.18 &&
  python setup.py build && mv build/lib.*/ ../ujson )

If this fails, or you prefer simplejson, install simplejson as

( cd server/lib && tar xfz simplejson-2.1.5.tar.gz )

No other library setup is required for standard brat operation.

(if you require Japanese tokenization support, please set up MeCab, included in the external/ directory.)

brat server configuration

The overall configuration of the brat server is controlled by config.py in the brat root directory. This configuration file is automatically created when running install.sh, but can be set up manually as follows:

Put a configuration template in place:

cp config_template.py config.py

Edit the configuration to suit your environment using your favorite text editor, e.g.

emacs config.py

Detailed instructions for setup are contained in the configuration file itself.

brat standalone server

As of version 1.3, brat includes an experimental standalone server. This is an alternative for running the brat server that doesn't require a separate web server such as Apache, and can be substantially faster, reducing server response times by 90% in some cases.

However, the standalone server is not currently security-hardened or as comprehensively tested as the rest of brat. For these reason, we strongly recommend serving brat via a full web server such as Apache in production environments, projects involving sensitive data, and on systems accessible from the internet.

On a standard installation, the brat standalone server can typically be run simply as

APACHE_USER=`./apache-user.sh`
sudo -u $APACHE_USER python standalone.py

(this command will prompt for your password.)

In more detail: to run the standalone server, you need to execute the script standalone.py in the brat directory. This script should have write permissing to the working directories (normally work/ and data/). If you have set these directories up using install.sh or per the instructions in the section on working directories, they should be writable by the Apache user (if one exists). The above attempts to determine the username for Apache (./apache-user.sh), and then run standalone.py as this user (sudo -u).

If you do not have Apache installed, you may instead run the standalone server with

sudo -u USER python standalone.py

where USER is any user with write permissions to the brat work and data directories.

Placing Data

Your installation should now be ready, just place your data in the data directory and make sure it has the right permissions using chmod as you did above.

You can either place your data files directly in the data directory, or create arbitraty directory structure under data. We recommend creating a subdirectory into data for each (version of a) corpus you are working on.

Brat stores its data in a standoff format, using two files for each document: a .txt file containing the document text (UTF-8 encoding), and a .ann file containing the annotations. To start a new annotation project, you can simply place the .txt files in the desired location and create an empty .ann file corresponding to each .txt. After copying the .txt files in the data directory, the .ann files can be created as follows:

find data -name '*.txt' | sed -e 's|\.txt|.ann|g' | xargs touch

(Remember also to set write permissions for these files.)

You can now access your brat installation by retrieving the brat directory or index.xhtml file from your web server using a supported web browser.

Third-party components

This part largely focuses on Ubuntu, but the instructions should be readily applicable to other *NIX-based systems.

Apache 2.x

brat supports FastCGI which can speed up your installation by roughly x10 since you won't have to invoke the Python interpreter for every request. If you want to use FastCGI as opposed to CGI keep an eye out for configuration comments regarding it.

NOTE: FastCGI support in brat is still somewhat experimental and there are known issues relating in particular to failures to reload aspects of the system on change to configuration files. We recommend to only use FastCGI for fixed annotation configurations.

For FastCGI you need the flup Python library:

( cd server/lib/ && tar xfz flup-1.0.2.tar.gz )

Install Apache 2.x if you don't have it already:

sudo apt-get install apache2

From here, the setup process is different for OS X and for other UNIX-like systems such as linux. Please check the section relevant to your system below.

Apache setup on UNIX-like systems other than OS X

Next, let's edit httpd.conf:

sudo vim /etc/apache2/httpd.conf

If you are installing brat into the public_html subdirectory of your home directory, add the following lines:

<Directory /home/*/public_html>
    AllowOverride Options Indexes FileInfo Limit
    AddType application/xhtml+xml .xhtml
    AddType font/ttf .ttf
    # For CGI support
    AddHandler cgi-script .cgi
    # Comment out the line above and uncomment the line below for FastCGI
    #AddHandler fastcgi-script fcgi
</Directory>

# For FastCGI, Single user installs should be fine with anything over 8
#FastCgiConfig -maxProcesses 16

If you are not installing into public_html in your home directory, adjust the above (in particular the line <Directory /home/*/public_html>) accordingly. If you installed into public_html in your home directory, make sure that you have the userdir module enabled:

sudo a2enmod userdir

For FastCGI you also want to install its module and then add it and the rewrite module that we use to redirect the CGI requests to FastCGI:

sudo apt-get install libapache2-mod-fastcgi
sudo a2enmod fastcgi
sudo a2enmod rewrite

The final FastCGI step is detailed in .htaccess in the brat installation directory, which involves uncommenting and configuring the rewrite module.

Finally tell Apache 2.x to load your new configuration.

sudo /etc/init.d/apache2 reload

Apache setup on OS X

Next, let's edit the apache httpd.conf:

sudo vim /etc/apache2/users/USER.conf

where USER is your username.

If you are installing brat into the Sites directory in your home directory, make sure the following is included:

<Directory "/Users/USER/Sites/">
   Options Indexes MultiViews FollowSymLinks
   Order allow,deny
   Allow from all

   AllowOverride Options Indexes FileInfo Limit
   AddHandler cgi-script .cgi
   AddType application/xhtml+xml .xhtml
   AddType font/ttf .ttf
</Directory>

where USER is again your username.

Finally, restart apache

sudo apachectl restart

(For FastCGI, we're currenly trying to determine a configuration that works on OS X and will provide instructions once we have one.)

Finding Your Apache 2 Group

Ideally you should set all permissions as needed for the Apache 2 group, but finding it can be painful.

Find out what the Apache group name is, it is usually apache or www-data; it can be found in apache2.conf or httpd.conf under /etc/apache2/ or /etc/httpd/. Let's assume it's www-data. Then:

sudo chgrp -R www-data data work
chmod -R g+rwx data work

Actually, due to the joy of Linux segmentation you can find the group elsewhere as well. Here is a small heuristic that works on at least two Linux distributions:

locate --regex '(apache2|httpd)\.conf$' | xargs cat | grep 'Group'

If what you get from this looks funky (like a variable), say with a leading $, try this:

locate envvars | grep -E '(apache2|httpd)' | grep '/etc/' \
    | xargs cat | grep -i 'group'

If this doesn't work either dive into /etc/group and hope that you can find something that at least looks like apache or www-data:

cat /etc/group | cut -d : -f 1 | sort

Back-ups

There is a script that you can use with good old cron. Do as follows.

Add a line like the one below to the crontab for your Apache user:

0 5 * * * ${PATH_TO_BRAT_INSTALLATION}/tools/backup.py

You will now have a back-up made into your work directory every morning at five o'clock.

Next: setting up annotation project configuration