| SIP Express Media Server - README |
SEMS is a free, high performance, extensible media server
for SIP (RFC3261) based VoIP services.
It is intended to complement proxy/registrar servers
in VoIP networks for all applications where server-
side processing of audio is required, for example away
or pre-call announcements, voicemail, or network side
SEMS can be used to implement simple high performance
components like announcement servers as building
blocks of more complex applications, or, using its powerful
framework for application development including back-to-back
user agent (B2BUA) and state machine scripting functionality,
complex VoIP services can be realized completely in SEMS.
SEMS only supports patent free codecs. That means that codecs
like g729 won't be supported. On the other hand, SEMS supports
all important patent free codecs out of the box (g711u, g711a,
GSM06.10, speex, G726, L16 and iLBC). Integrating other codecs
in SEMS is very simple (patented or not).
SEMS shows very good performance on current standard
PC architecture based server systems. It has sucessfully
been run with 1200 G.711 conference channels on a quad-core
Intel(R) Xeon at 2GHz (700 GSM, 280 iLBC channels), and up to
5000 channels on a dual quad Xeon at 2.9GHz. On the
other hand it also runs on very small devices - for example
small embedded systems like routers running OpenWRT, for which
of course the achievable channel count is not that high.
SEMS is free (speech+beer) software. It is licensed under dual
license terms, the GPL (v2+) and proprietary license.
See doc/COPYING for details.
The following applications are shipped with SEMS :
* announcement plays an announcement
* voicemail records voice messages and sends them
as email, saves them to a voicebox, or
* conference enables many people to talk together
at the same time
* webconference conference application that can be
controlled from an external program,
e.g. a website
* echo test module to echo the caller's voice
* voicebox users can dial in to the voicebox to check
* annrecorder users can record their personal greeting
* mailbox auto-attendant that saves voicemails into
an IMAP server. Users can dial in to check
their messages (simpler version)
* ann_b2b pre-call-announcement, plays announcement
before connecting the callee in B2BUA mode
* announce_transfer pre-call-announcement, plays announcement
and then transfers the caller to the callee
* early_announce (pre-call) announcement using early media (183),
optionally continues the call in B2BUA mode
* auth_b2b B2BUA app that changes identity and authenticates
second call leg
* conf_auth collect a PIN number, verify it against an
XMLRPC authentication server and connects in
* pin_collect collect a PIN, optionally verify it, and transfer
the call into a conference
* callback reject the call, call back caller later and have
her enter a number to call in b2bua with media relay
* reg_agent SIP REGISTER to register SEMS' contact to an aor
* sw_prepaid_sip prepaid application
* call_timer B2BUA, ends the call after a certain time
* precoded_announce plays preencoded announcements
Developing and customizing Applications and services:
SEMS comes with a set of example applications intended to help
development of custom services, including a calling card
application, a traffic generator, a component to control the
media server via XMLRPC, and announcements played from DB.
DSM state machine scripting is a powerful yet simple method
to rapidly implement custom applications. With this method,
the service logic is written as an easy to understand
textual definition of a state machine, which is interpreted
and executed for every call. The (domain specific) language
for defining state machines can be extended by implementing
modules. A set of useful modules are shipped with SEMS,
including MySQL database access module, Python module,
conference support, Amazon AWS and more.
SEMS' core implements basic call and audio processing,
and loads plug-ins which extend the system. Audio
plug-ins enable new codecs and file formats,
application plug-ins implement the services' logic.
Other modules called component modules provide
functionality for other modules to use.
You can easily extend SEMS by creating your own plug-ins.
Applications can be written using the SEMS framework API
in C++, or in Python using an embedded python interpreter
of the ivr or py_sems modules, or the DSM.
All requirements are optional.
1. Python version >= 2.3 for the ivr (embedded python interpreter)
and py_sems, optional
2. flite speech synthesizer for TTS in the ivr, optional
3. lame >= 3.95 for mp3 file output, optional. mpg123 for mp3 playback,
4. xmlrpcpp for XMLRPC control, optional
5. spandsp library for DTMF detection and PLC, optional
(SEMS has its own implementations for both)
6. libZRTP SDK (http://zfoneproject.com) for ZRTP, optional
7. libev for jsonrpc, optional
How to get started with SEMS:
To try out SEMS, the easiest is to get a release from
http://ftp.iptel.org/pub/sems/, unpack and install it
using the usual make && make install. After installation, the
configuration file /usr/local/etc/sems/sems.conf needs to be
On debian, you can add the iptel.org SEMS repository to
deb http://ftp.iptel.org/pub/sems/debian lenny free
deb-src http://ftp.iptel.org/pub/sems/debian lenny free
and install SEMS from binary packages with:
$ apt-get install sems sems-python-modules
If you want to build SEMS from source on debian, see below.
On Fedora, simply do
$ sudo yum install sems
and sems package will be installed.
You can also follow one of the tutorials
linked from the SEMS homepage
(e.g. http://www.iptel.org/howto_sems_voicemail).The Application
Modules Documentation page then gives an overview of the application
modules that come with SEMS
If you are interested in writing your own applications, the
application development tutorial is a good start
together with the design overview
(http://www.iptel.org/files/semsng-designoverview.pdf) and the example
1. Download SEMS
Download the source tarball from ftp://ftp.iptel.org/pub/sems/,
or get the newest version through svn:
$ svn checkout svn://svn.berlios.de/sems/trunk
$ svn checkout http://svn.berlios.de/svnroot/repos/sems/trunk
2. Compile SEMS
$ make all
$ make install
Notice: you don't need to execute 'make install' if you want to
run SEMS from the source tree, but in this case some paths
in the config files will need to be adapted.
If you want to install to a different prefix than /usr/local,
For detailed instructions for some modules, have a look at
doc/COMPILING as well.
3. Look at the default configuration file whether it fits your need:
If you installed SEMS with 'make install', the configuration
is at '/usr/local/etc/sems/sems.conf'.
Else make your own using sems.conf.sample.
4. Start SEMS:
If you installed SEMS with 'make install', sems can be
found at '/usr/local/sbin/sems'.
If you need help starting 'sems', try 'sems -h'.
5. Using SEMS
SEMS needs to be told from the many possible applications that are
loaded which one to run. You can simply set the application
in sems.conf, e.g. application=conference. You can also define an
application mapping, i.e. numbers (R-URIs) that will be mapped to
applications, see the explanation of 'application' parameter in
Alternatively, the SIP-stack-SER can be configured in ser-sems.cfg
to send all calls directly to one application, in that case the line
in ser-sems.cfg could be changed to e.g.
A sample ser.cfg file in this style can be found at
Creating packages on debian (ubuntu, ...):
install debian package build tools:
$ apt-get install debhelper devscripts
$ apt-get install g++ make libspandsp-dev flite-dev libspeex-dev \
libssl-dev python-dev python-sip4-dev openssl
get the source:
$ wget ftp.iptel.org/pub/sems/sems-x.y.z.tar.gz ; tar xzvf sems-x.y.z.tar.gz
$ cd sems-x.y.z ; ln -s pkg/debian .
set version in changelog if not correct
$ dch -v x.y.z "SEMS x.y.z release"
$ dpkg-buildpackage -rfakeroot -us -uc
install package in .. using dpkg.
Installed files using 'make install':
/usr/local/sbin/sems : SEMS executable
/usr/local/lib/sems/plug-in/* : plug-ins
/usr/local/lib/sems/audio/* : default path for audio files
/usr/local/lib/sems/ivr/* : precompiled IVR scripts
/usr/local/etc/sems/sems.conf : configuration file
/usr/local/etc/sems/etc/* : modules configuration files
/usr/local/share/doc/sems/README : this README.
source_path/scripts/sems[.redhat] : example start-up scripts.
source_path/sems.conf.example : example configuration file.
In the doc/ directory there is a set of files describing the
applications shipped with SEMS, alongside some more documentation.
Generate the doxygen documentation with 'make doc' in doc/doxygen_doc,
that contains all these files as well.
All this and more documentation is available online linked from
the SEMS homepage:
Support, mailing lists, bugs and contact:
Please have a look at the documentation and other information on
the SEMS homepage (www.iptel.org/sems).
Best-effort support is given through the mailing lists for SEMS,
firstname.lastname@example.org and email@example.com, which are the first address
to ask for help, report bugs and improvements. You need to be
subscribed to be able to post to the lists: http://lists.iptel.org.
The mailing list archives at http://lists.iptel.org/pipermail/sems/
and http://lists.iptel.org/pipermail/semsdev/ can be a great help as
well (especially with google site search on lists.iptel.org, e.g.
The bug tracker for SEMS is at http://tracker.iptel.org/browse/SEMS
Please submit all bugs, crashes and feature requests you encounter.
Raphael Coeffic (firstname.lastname@example.org), the father of SEMS,
Stefan Sayer (email@example.com), current lead developer,
and all contributors:
Alfred E Heggestad
Andriy I Pylypenko
Greger Viken Teigre
Rui Jin Zheng
Tom van der Geer
(if you want to be added or removed from this list, please mail to firstname.lastname@example.org)
Special thanks to IPTEGO GmbH and iptelorg GmbH for sponsoring development of SEMS.
All kinds of contributions and bug fixes are very welcome, for
example new application or codec modules, documentation pages, howtos
etc. Please email one of the lists or the authors.
Please also have a look at the contributions license policy
SEMS - the media-S in the SLAMP.