[asterisk-dev] Audio to/from Asterisk

Dan Jenkins dan at nimblea.pe
Wed Jul 24 10:50:53 CDT 2019


So even in its most basic form - lets take a Node process thats made a
websocket connection to Asterisk for ARI purposes... within the event loop
for that ARI session, I cant easily handle the audio going out to asterisk
within  the same "thread" (I use that term loosely) because when the
connection comes into Node (say im listening as a audio providing server on
port X ready),,, those two things within the same node process are within
two separate events within the same process - as soon as you start sharing
state outside of that ARI "thread" thats dealing with that one session we
get into dangerous territory. I'd much rather have a websocket out to
Asterisk, have an event come in via websocket to say heres a new session
and we create an ARI "thread" within that process. When I need to make a
session with bidirectional audio in/out of asterisk to my process I should
be able to ask the ARI for where to connect to in order to  send/receive
that media, make that connection with a uuid and then instruct ARI to
bridge the original channel and the new one.

Simple ARI apps shouldnt need messaging between apps to handle instructions
and media. The same process and event loop should be able to handle both.

Theres a lot of personal preference in all of that I do grant you. But I do
believe that I shouldnt have to make specific node processes available to
external resources but asterisk already is (to a degree). I see both sides
of the coin - im basically saying i want to build it this way and youre
saying, but i dont want to  have asterisk  open to external
applications.... so who wins? In an ideal world you should be able to do
both because Asterisk needs to be a flexible media engine!

On Wed, Jul 24, 2019 at 6:33 PM Seán C. McCord <ulexus at gmail.com> wrote:

> AudioSocket was initially designed precisely to be able to slot into the
> role of MRCP, for a client who was more interested in designing their own
> system (and paying for its development) than paying the license fees for
> UniMRCP.  George's solution should also fit that need.  I have since used
> AudioSocket for a number of other operations, but that was its original
> impetus.
>
> The channel interface for AudioSocket really should solve the ARI use case
> (the app interface is simpler for AMI and AGI).
>
> As to the port-in rather than port-out mechanism... I'm sure there are
> definitely use cases for it, but in any sufficiently large system, you're
> going to have to do routing one way or the other:  to Asterisk or from
> Asterisk.  Ultimately, you are going to have multiple instances of all
> components, so the directionality of the socket connection doesn't seem to
> me to be a large factor.  That is to say, adding an inbound port to
> Asterisk to initiate the two-way socket doesn't seem to be particularly
> more useful than outgoing.  Am I missing something, Dan, about your
> scenarios?  I didn't really design AudioSocket with inbound connections in
> mind at all.
>
>
> On Wed, Jul 24, 2019 at 10:03 AM George Joseph <gjoseph at digium.com> wrote:
>
>>
>>
>> On Wed, Jul 24, 2019 at 7:11 AM Dan Cropp <dan at amtelco.com> wrote:
>>
>>> Out of curiosity, would this be an alternative to unimrcp’s asterisk
>>> support for MRCP (TTS/ASR)?
>>>
>>
>> Well it wasn't intended to implement MRCP but yes, it's intended to
>> provide the same very-high-level functionality controlled via ARI.
>>
>>
>>
>>>
>>>
>>>
>>>
>>> *From:* asterisk-dev <asterisk-dev-bounces at lists.digium.com> *On Behalf
>>> Of *Luca Pradovera
>>> *Sent:* Monday, July 22, 2019 3:12 AM
>>> *To:* Asterisk Developers Mailing List <asterisk-dev at lists.digium.com>
>>> *Subject:* Re: [asterisk-dev] Audio to/from Asterisk
>>>
>>>
>>>
>>> Hello,
>>>
>>> I remember this being talked about, and it's essentially tied to the
>>> mechanism that would allow streaming ASR/TTS services to be used.
>>>
>>> +1 on this feature!
>>>
>>>
>>>
>>> On Mon, Jul 22, 2019 at 10:01 AM Dan Jenkins <dan at nimblea.pe> wrote:
>>>
>>> Also coming back to this with some real-life case issues I'm currently
>>> facing and why I can't use audiosocket :(
>>>
>>>
>>>
>>> I need to be able to ask the ARI/AGI/AMI for an IP/port combo and for an
>>> external app to then connect into asterisk rather than asterisk connecting
>>> to a URI elsewhere. Lets say I already have a nodejs (or any other
>>> language) process taking care of controlling that call via ARI or even AGI
>>> (all the different ways) - I need that same process to handle the media I'm
>>> sending and receiving to/from asterisk and so if I already have that
>>> process up and then Asterisk calls out to a generic URI then that media
>>> will never make it back to the original nodejs process.
>>>
>>>
>>>
>>> I think its of upmost importance that I be able to ask asterisk for a
>>> host:port pair and then be able to connect to that port from my external
>>> application.
>>>
>>>
>>>
>>> What do people think? I thought we'd talked about this mechanism at
>>> devcon?
>>>
>>>
>>>
>>> Dan
>>>
>>>
>>>
>>> On Sat, Jul 20, 2019 at 2:39 PM Dan Jenkins <dan at nimblea.pe> wrote:
>>>
>>> Just  going to chime in and say I don't see a one way audio solution as
>>> enough and I'd worry that doing that would lead to "oh but only so many
>>> people need to chuck audio in". This has been discussed at at least 3
>>> devcons now.
>>>
>>>
>>>
>>> On Thu, Jul 18, 2019 at 2:09 PM Seán C. McCord <ulexus at gmail.com> wrote:
>>>
>>> I certainly don't mind if a better-designed system comes along and
>>> obviates my AudioSocket implementation.  I built it because I needed it.
>>> However, bidirectional audio flow is critical for me (speech synthesis,
>>> external interfacing, real-time processed audio, processed injections,
>>> etc).  While I would actually prefer a system which was a bit beefier than
>>> my own (simple protocol aside, there's a good deal of range between my
>>> protocol and MRCP), my meagre C skills (and patience) don't allow me to
>>> venture forth into those difficult waters.
>>>
>>>
>>>
>>> I do like the separate connection (unlike Wazo's) and the support of TLS
>>> (unlike mine)... and yours is certainly (even without looking) more
>>> performant.  Mine also probably needs a multi-threaded, dedicated-receiver
>>> approach like most of the other channels which handle socket-received
>>> media, rather than the simple non-blocking I/O with null frame insertion.
>>> No perfect solution yet.
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> On Thu, Jul 18, 2019 at 8:01 AM George Joseph <gjoseph at digium.com>
>>> wrote:
>>>
>>> Hey Guys,
>>>
>>>
>>>
>>> I was on vacation when this thread happened but I'm also working on this
>>> now.  The implementation uses the existing ARI channel and bridge recording
>>> endpoints ands add the ability to specify a URI in the form of
>>> (udp|tcp|tls)://hostname:port to stream the media.  This makes use of the
>>> existing chan_bridge_media driver and only requires a scheme handler
>>> similar to Seán's res_audiosocket.   This implementation is more targeted
>>> to real-time speech recognition/transcription/captioning and is therefore
>>> one way (outbound).  A future enhancement is planned that would send each
>>> channel in a bridge as a separate audio channel in a multi-channel
>>> container.
>>>
>>>
>>>
>>> I'm not suggesting that this should replace Seán's audiosocket stuff but
>>> I did want to let you know what was in the pipeline.
>>>
>>>
>>>
>>> george
>>>
>>>
>>>
>>> On Fri, Jul 5, 2019 at 7:38 AM Seán C. McCord <ulexus at gmail.com> wrote:
>>>
>>> Solutions such as Jack are non-network oriented and severely limited in
>>> scalability.  There are, of course, many other options, but the closest are
>>> chan_rtp and chan_nbs.  RTP is a good option except for the difficulty for
>>> non-telephony people to interact with it.  NBS is deprecated, undocumented,
>>> and unsupported by any locatable resources.
>>>
>>>
>>>
>>> While the original app interface from last year required dialplan, the
>>> channel interface does not.  It is a plain channel which can be used by ARI
>>> directly.
>>>
>>>
>>>
>>>
>>>
>>> On Fri, Jul 5, 2019, 15:28 Sylvain Boily <sylvain at wazo.io> wrote:
>>>
>>> Hello Seán,
>>>
>>> On 2019-07-05 4:45 a.m., Seán C. McCord wrote:
>>>
>>> A brief update:
>>>
>>>
>>>
>>> I have adapted my app_audiosocket from last year to become
>>> chan_audiosocket, a full bidirectional audio channel interface for Asterisk
>>> to any AudioSocket service (which itself is a dead-simple implementation).
>>> I'll be demoing it in my talk next week at CommCon, for anyone who might be
>>> interested.  I'm going to try to have it ready to push to gerrit for review
>>> this weekend.
>>>
>>>
>>> I'll be there.
>>>
>>>
>>>
>>> For now, you can see it in the 'channel' branch of
>>> github.com/CyCoreSystems/audiosocket.
>>>
>>>
>>> This is very different from what we did. You need dialplan to use it. In
>>> our case we don't need any dialplan to use it, it's more ARI approach.
>>>
>>> Sylvain
>>>
>>> --
>>> _____________________________________________________________________
>>> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>>>
>>> asterisk-dev mailing list
>>> To UNSUBSCRIBE or update options visit:
>>>    http://lists.digium.com/mailman/listinfo/asterisk-dev
>>>
>>>
>>>
>>>
>>> --
>>>
>>> *George Joseph*
>>>
>>> Digium - A Sangoma Company | Software Developer | Software Engineering
>>>
>>> 445 Jan Davis Drive NW - Huntsville, AL 35806 - US
>>>
>>> direct/fax: +1 256 428 6012
>>>
>>> Check us out at: https://digium.com · https://sangoma.com
>>>
>>>
>>>
>>>
>>>
>>>
>>> --
>>>
>>> Seán C. McCord
>>>
>>> ulexus at gmail.com
>>>
>>> CyCore Systems
>>>
>>> --
>>> _____________________________________________________________________
>>> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>>>
>>> asterisk-dev mailing list
>>> To UNSUBSCRIBE or update options visit:
>>>    http://lists.digium.com/mailman/listinfo/asterisk-dev
>>>
>>> --
>>> _____________________________________________________________________
>>> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>>>
>>> asterisk-dev mailing list
>>> To UNSUBSCRIBE or update options visit:
>>>    http://lists.digium.com/mailman/listinfo/asterisk-dev
>>>
>>> --
>>> _____________________________________________________________________
>>> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>>>
>>> asterisk-dev mailing list
>>> To UNSUBSCRIBE or update options visit:
>>>    http://lists.digium.com/mailman/listinfo/asterisk-dev
>>
>>
>>
>> --
>> *George Joseph*
>> Digium - A Sangoma Company | Software Developer | Software Engineering
>> 445 Jan Davis Drive NW - Huntsville, AL 35806 - US
>> direct/fax: +1 256 428 6012
>> Check us out at: https://digium.com · https://sangoma.com
>>
>> --
>> _____________________________________________________________________
>> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>>
>> asterisk-dev mailing list
>> To UNSUBSCRIBE or update options visit:
>>    http://lists.digium.com/mailman/listinfo/asterisk-dev
>
>
>
> --
> Seán C. McCord
> ulexus at gmail.com
> CyCore Systems
> --
> _____________________________________________________________________
> -- Bandwidth and Colocation Provided by http://www.api-digital.com --
>
> asterisk-dev mailing list
> To UNSUBSCRIBE or update options visit:
>    http://lists.digium.com/mailman/listinfo/asterisk-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20190724/54ef859c/attachment-0001.html>


More information about the asterisk-dev mailing list