It’s a fairly significant upgrade from 0.8.
Some of the changes are:
Auto-install support for Ubuntu 8.04 & 8.10 and Fedora 9 and 10
Parameter-passsing from Asterisk dialplan
Easy alternate TTS support
Fixed VXML 2.1 tag DOM functions in Ubuntu
Fixed mangling of certain characters in TTS and grammars
Fixed recording hangups sometimes not being handled properly
Fixed improper grammar type detection
and lots of bug fixes.
You can get it in the Downloads section.
Voiceglue documentation has also been improved with this release, and is now in the new voiceglue wiki so that the community can help maintain and extend it.
Over at Ampersand, we had reason to write a simple VXML script that exercises some of the main areas of VXML. We thought we’d release it under GPL as part of the voiceglue project, in case anyone else in the community finds it of use. The exerciser is a series of VXML scripts that exercise various functions relating to input, output, flow, dialogs, events, etc. More details here.
The script is available at http://www.voiceglue.org/vxml_tests/ex.vxml ; and feel free to point a voice browser at it and run it directly off our site if you like.
It’s been awhile since we posted a status update on the project, so here goes:
One big focus over the last 5 months has been to clean up and fix various small issues in the 0. series of releases as well as supporting voiceglue users on the forum. We’re now at 0.6, which has been relatively stable. We try to answer all questions on the mailing list within a day or two.
We’ve also been working on integrating and testing speech recognition functionality. We’ve been making use of Asterisk’s res_speech.so interface, which is the Digium-blessed approach to interfacing with speech engines. Right now we are only aware of support for Lumenvox under this interface which we are using as our test environment (and we should note that they have been helpful and supportive of our effort). As a future project goal we will probably build a Sphinx interface if no-one else is doing so. The res_speech interface was exposed to the AGI via a source-level patch in Asterisk 1.4, but this patch is now rolled into the 1.6 Beta; so we are now working with the 1.6 Beta. Our coding is complete for this effort, and we are in debug.
We’ve been given a development key for Cepstral, and plan to integrate it as a commercial alternative to Flite.
We’ve been keeping a list of feature requests, to plan into upcoming releases.
Our goal is to get speech recognition working at which point we will call it a 1.0 version. We will then take our feature request list, poll our user community to get your input, and schedule dot releases.
One last note, I (Steve) will be at Spring VON in San Jose if any voiceglue users would like to get together. Drop me a note.
DIgium Asterisk World turned out to be a good conference. Attendance was higher than I had anticpated, perhaps 150 people at the keynote and more than that aggregated amongst the different tracks and on the floor. Lot’s of interested parties, including enterprise buyers, vars, and a good showing by the business-oriented parts of the community, which was the goal of the event.
Our panel was well attended, and we had a good discussion about what “sophisticated services” mean from both a carrier and an application perspective. Moderator Bryan Johns did a good job and made some thoughtful comments. Co-panelist Chris Gatch talked about the benefits of using a next gen carrier and how asterisk as a pbx can take full advantage of such services.
Our slides are attached here, short and sweet. We overviewed different areas of sophisticated services possible on asterisk, such as conferencing, outbound campaigns, ACD/CTI, etc., then we focused on IVR and gave a short 4-slide overview of VXML, openVXI, and Voiceglue (including the voiceglue architecture diagram below — you can click thru to flickr for a bigger version). Several people asked some insightful questions, and in the post-panel discussions we had some requests for future project directions such as call control.
On last item of note is that we had a good talk with Craig Campbell of Cepstral, who donated a SDK so that we can integrate his TTS into Asterisk — which we plan to do as soon as we finish with our ASR integration. We do have a bit a of a roadmap that we’ve been putting together, and we’ll get a post out on that in the near future.
We’ve been asked to talk about using Asterisk to build scalable open-source IVRs at Pulver’s Fall VON and Digium Asterisk World. The conference is held here in Boston (our stomping grounds), and our panel is at 2pm on Tuesday, Oct 30.
I (Steve) will be on the panel along with a next gen service provider, and a very experienced Asterisk integrator. The title of the panel is Expanding to Sophisticated Services.
This is the inaugural year of the Asterisk World conference; which is the business-oriented counterpart to Astricon.
You can sign up for the full asterisk world conference, which gets you into all the keynotes and panels, or you can come to the expo only. As of today, you can still get a free expo pass if you go to the main VON site.
If you’re interested in Voiceglue, come talk to us at VON. Doug and I will be there all day Tuesday, and I’ll be back for Wednesday and part of Thursday as well.
Version 0.5 of Voiceglue is now available on the download page.
This version fixes a bug with validation of identifiers in VXML files.
If you’re getting errors like these:
Datatype error: Type:InvalidDatatypeValueException, Message:Value 'caller' does not match regular expression facet '([a-zA-Z]|[a-zA-Z$][a-zA-Z0-9_$]*[a-zA-Z0-9_])’
Failed to parse the URI’s content. Make sure that this document consists of valid VXML.
then you should definitely upgrade.
This version turns off Xerces XML validation, which reports valid
VXML documents (confirmed valid by XSV) as errors.
