extensions_pngnpbx_hh.conf

; extensions_pngnpbx_hh.conf
;
; Dial plan demo for Asterisk for nearly continuous text-to-speech
; as well as basic Speech...() examples to help test your setup.
;
; Feeds call audio from one channel into a new conference,
; and also attaches additional channels to the conference,
; in order to record and/or transcribe the audio in chunks,
; while letting the OG channel carry-on in the dial plan.
;
; When installed with working Speech...() engines, such
; as res_speech_vosk, the words spoken by the OG caller
; are available in the OG channel's SHARED() variables.
;
; Copyright 2024 Penguin PBX Solutions
;
; Licensed under the GPLv3
;
; Sponsored by Helping Hands India NGO
;
; Tested with Asterisk version 20.6.0, and the Vosk module for Asterisk,
; and a very small patch for partial speech results handling.


; DEMO 0
;
; After the beep, try saying a word and/or pressing a button on your phone.
; This is one basic way to do Automatic Speech Recognition (ASR) in Asterisk.
; Good starting point if changing from Background() to SpeechBackground().
; Nothing special here. Helpful for quickly debugging setup eg. does Vosk work?
; Because if this does not work, then nothing else will be fun in this file ;(
;
[pngnpbx-hh-demo-0]
exten = s,1,Answer()
 same = n,Wait(1)
;same = n,Set(DENOISE(rx)=on) ; might help detection
;same = n,Set(AGC(rx)=8000) ; might help detection
 same = n,SpeechCreate()
 same = n,SpeechBackground(beep,5)
 same = n,Set(safe_speech_text_0=${FILTER(a-zA-Z0-9 ,${SPEECH_TEXT(0)})})
 same = n,Set(safe_speech_score_0=${FILTER(a-zA-Z0-9 ,${SPEECH_SCORE(0)})})
 same = n,Set(safe_speech_text_1=${FILTER(a-zA-Z0-9 ,${SPEECH_TEXT(1)})})
 same = n,Set(safe_speech_score_1=${FILTER(a-zA-Z0-9 ,${SPEECH_SCORE(1)})})
 same = n,SpeechDestroy()
 same = n,Playback(ascending-2tone)
 same = n,Set(speech_digits=)
 same = n,Set(dtmf_digits=${FILTER(0-9,${safe_speech_text_0})})
 ; Vosk returns digits when DTMF touch tones are used
 same = n,GotoIf($[${LEN(${dtmf_digits})}]?${dtmf_digits},1)
 ; but this current file provides helper context to convert speech to digits
 same = n,Gosub(pngnpbx-hh-grammar-digits-en-us,s,1(${safe_speech_text_0}))
 same = n,Set(speech_digits=${FILTER(0-9,${GOSUB_RETVAL})})
 same = n,GotoIf($[${LEN(${speech_digits})}]?${speech_digits},1)

 same = n(words),Playback(you-entered)
 same = n,SayAlpha(${safe_speech_text_0})
 same = n,Playback(vm-goodbye)
 same = n,Hangup()

exten = _X,1,NoOp(one digit)
 same = n,Goto(d${EXTEN},1)
exten = _XX,1,NoOp(two digits)
 same = n,Goto(d${EXTEN},1)
exten = _XXX,1,NoOp(three digits)
 same = n,Goto(d${EXTEN},1)
exten = _XXX.,1,NoOp(four or more digits)
 same = n,Goto(d${EXTEN},1)

exten = _d.,1,NoOp()
 same = n,Playback(you-entered&digits)
 same = n,GotoIf($[${LEN(${speech_digits})}]?digits)
 same = n,SayAlpha(DTMF)
 same = n(digits),SayDigits(${EXTEN:1})
 same = n,Playback(vm-goodbye)
 same = n,Hangup()


; DEMO 1
;
; After the beep, say a word, or press a button;
; Hear tones, then say another word, or press a button.
; If nothing works, then make sure you have Vosk running.
; If it works a bit but not quite right, then try talking s-l-o-w-e-r.
;
[pngnpbx-hh-demo-1]
exten = s,1,Answer()
 same = n,Gosub(pngnpbx-hh-init,s,1) ; Minimal Required Part
 same = n,NoOp(modify the sample glue below per your requirements)
 same = n,Set(palace=${GOSUB_RETVAL})
 same = n,Background(silence/1)
 same = n,Background(phonetic/e_p)
 same = n,Background(beep)

 same = n,Verbose("SAY A WORD")
 same = n,Set(talking1=${SHARED(hh_word)})
 same = n,While($[!${LEN(${talking1})}])
 same = n,Background(silence/1)
 same = n,Set(talking1=${SHARED(hh_word)})
 same = n,EndWhile()
 same = n,Background(ascending-2tone)

 same = n,Verbose("SAY ANOTHER WORD")
 same = n,Set(talking2=${SHARED(hh_word)})
 same = n,While($["${talking1}" == "${talking2}"])
 same = n,Background(silence/1)
 same = n,Set(talking2=${SHARED(hh_word)})
 same = n,EndWhile()
 same = n,Background(ascending-2tone)

 same = n,NoOp(Or you could dial something instead...)
 ;same = n,Dial(PJSIP/7004) ; or go to an IVR, etc.
 same = n,NoOp(You can stop the ASR Conference Palace early...)
 same = n,Gosub(pngnpbx-hh-hup,${palace},1)
 same = n,Verbose("YOU SAID '${talking1}' THEN '${talking2}'")
 same = n,NoOp(...or it will stop when the OG caller hangs up.)
 same = n,Playback(you-entered)
 same = n,SayAlpha(${talking1})
 same = n,Playback(and)
 same = n,SayAlpha(${talking2})
 same = n,Playback(vm-goodbye)
 same = n,Hangup()


; DEMO 2
;
; Plays music until you say "penguin",
; then Allison says something silly.
;
[pngnpbx-hh-demo-2]
exten = s,1,Answer()
 same = n,Gosub(pngnpbx-hh-init,s,1(talk))
 same = n(moh),MusicOnHold(default)
 same = n,Hangup()
 same = n(talk),NoOp(what did you say?)
 same = n,Set(talking=${SHARED(hh_word)})
 same = n,GotoIf($["${talking}"!="penguin"]?moh)
 same = n,Playback(why-no-answer-mystery) ; LOL
 same = n,Goto(moh)


; INIT
;
; Main place you should Gosub() to from elsewhere in your dialplan.
; Cascades signals to downstream contexts found further below,
; including hangup handlers and several new helper calls (via async originate.)
;
; Ordered Parameters:
;   ARG1 - return priority name (or number) that the originating channel
;          is redirected to when talking is detected - this lets you
;          speech-enable any application the caller may be in eg. MusicOnHold()
;
; Returns:
;   palace - The name of the ConfBridge() - useful to shut down ASR
;            before the OG call is hung up eg. you are done with transcription.
;
[pngnpbx-hh-init]
exten = s,1,NoOp(Initializing ASR)
 same = n,Set(LOCAL(ogpritalk)=${ARG1})
 same = n,Set(LOCAL(ogctx)=${STACK_PEEK(1,c)})
 same = n,Set(LOCAL(ogext)=${STACK_PEEK(1,e)})
 same = n,Set(LOCAL(caller)=${CHANNEL(name)})
 same = n,Set(LOCAL(opts1)=caller="${caller}")
 same = n,Set(LOCAL(opts2)=ogext=${ogext}^ogctx=${ogctx}^ogpritalk=${ogpritalk})
 same = n,Set(LOCAL(palace)=p${MD5(${RAND}${UNIQUEID}):24})
 same = n,Set(LOCAL(body)=Local/${palace}@pngnpbx-hh-bridge)
 same = n,Set(LOCAL(ear)=pngnpbx-hh-listen)
 same = n,Set(LOCAL(hand)=pngnpbx-hh-record)
 same = n,Set(LOCAL(brain)=pngnpbx-hh-analyze)
 same = n,Set(CHANNEL(hangup_handler_push)=pngnpbx-hh-hup,${palace},1)
 ; Asynchronously spool three calls to do the work.
 ; 1. ear - listens - uses ChanSpy() to feed OG caller audio into ConfBridge()
 same = n,Originate(${body},exten,${ear},${palace},1,2,av(${opts1}))
 same = n,NoOp(ear originate_status=${ORIGINATE_STATUS})
 ; 2. hand - records - saves ConfBridge() audio using Record()
 ; LOOK OUT 8-) a lot of recordings are made, so you might want to comment out!
 same = n,Originate(${body},exten,${hand},${palace},1,2,a)
 same = n,NoOp(hand originate_status=${ORIGINATE_STATUS})
 ; 3. brain - analyzes - lets Speech...() convert speech-to-text into SHARED()
 same = n,Originate(${body},exten,${brain},${palace},1,2,av(${opts1}^${opts2}))
 same = n,NoOp(brain originate_status=${ORIGINATE_STATUS})
 same = n,DumpChan(5)
 same = n,Return(${palace})


; HUP
;
; Hangup handler for shutting down the ASR Conference Palace.
;
[pngnpbx-hh-hup]
exten = _p.,1,DumpChan(5)
 same = n,Set(LOCAL(palace)=${FILTER(pabcdef0123456789,${EXTEN})})
 same = n,Set(LOCAL(palace_count)=${CONFBRIDGE_INFO(parties,${palace})})
 same = n,GotoIf($[!${LEN(${palace_count})}]?out)
 same = n,GotoIf($[!${palace_count}]?out)
 same = n,ConfKick(${palace})
 same = n(out),Return()


; BRIDGE
;
; Sets reasonable defaults and enters the Conference Palace.
; Called from several places.
;
[pngnpbx-hh-bridge]
exten = _p.,1,DumpChan(5)
 same = n,Set(palace=${FILTER(pabcdef0123456789,${EXTEN})})
 same = n,Answer()
 same = n,Set(CONFBRIDGE(user,announce_join_leave)=no)
 same = n,Set(CONFBRIDGE(user,announce_only_user)=no)
 same = n,Set(CONFBRIDGE(user,announce_user_count)=no)
 same = n,Set(CONFBRIDGE(user,announce_user_count_all)=no)
 same = n,Set(CONFBRIDGE(user,startmuted)=no)
 same = n,Set(CONFBRIDGE(user,quiet)=yes)
 same = n,ConfBridge(${palace})
 same = n,NoOp(confbridge_result=${CONFBRIDGE_RESULT})
 same = n,Hangup()


; LISTEN
;
; The "ear". Listens to the caller. Sends their audio to the Conference Palace.
;
[pngnpbx-hh-listen]
exten = _p.,1,DumpChan(5)
 same = n,Set(palace=${FILTER(pabcdef0123456789,${EXTEN})})
 same = n,Answer()
 same = n,ChanSpy(${caller:1:-1},EoqsSu)
 same = n,NoOp(spy_channel=${SPY_CHANNEL})
 same = n,Hangup()


; RECORD
;
; The "hand". Repeatedly record Conference Palace audio into many wav files.
;
[pngnpbx-hh-record]
exten = _p.,1,DumpChan(5)
 same = n,Set(palace=${FILTER(pabcdef0123456789,${EXTEN})})
 same = n,Set(rectimeout=2)
 ; Yes the recording will end early when the OG hangs up,
 ; but recording in four second chunks means not much is lost,
 ; and these recordings are really only useful for debugging.
 same = n,Set(reclength=4)
 same = n,Set(recbase=/var/spool/asterisk/recording/pngnpbx-hh-${palace})
 same = n,Set(recformat=wav)
 same = n,Answer()
 same = n,While(1)
 same = n,Record(${recbase}-${EPOCH}.${recformat},${rectimeout},${reclength},q)
 same = n,NoOp(recorded_file=${RECORDED_FILE})
 same = n,NoOp(record_status=${RECORD_STATUS})
 same = n,EndWhile()
 same = n,Hangup()


; ANALYZE
;
; The "brain". Repeatedly does ASR on Conference Palace audio.
; Sets variables with speech detection result words and scores.
; Optionally redirects the original caller channel when they are done talking.
;
[pngnpbx-hh-analyze]
exten = _p.,1,DumpChan(5)
 same = n,Set(palace=${FILTER(pabcdef0123456789,${EXTEN})})
 same = n,Set(speechengine=)
 same = n,Set(speechtimeout=3)
 same = n,Answer()
 same = n,While(1)
 same = n,SpeechCreate(${speechengine})
 ; The SpeechBackground() app plays this file, which makes the speech
 ; detection less than continuous; so probably should modify Speech...() apps
 ; to handle a timeout without a file, or some thing like that (TODO...)
 ;same = n,SpeechStart() ; do not use this, needs upstream improvements
 same = n,SpeechBackground(ascending-2tone,${speechtimeout})
 same = n,Set(safe_speech_text_0=${FILTER(a-zA-Z0-9 ,${SPEECH_TEXT(0)})})
 same = n,Set(safe_speech_score_0=${FILTER(a-zA-Z0-9 ,${SPEECH_SCORE(0)})})
 ; Current res_speech_vosk only returns primary result with score of 100
 ; but maybe your speech engine is different...
 same = n,Set(safe_speech_text_1=${FILTER(a-zA-Z0-9 ,${SPEECH_TEXT(1)})})
 same = n,Set(safe_speech_score_1=${FILTER(a-zA-Z0-9 ,${SPEECH_SCORE(1)})})
 same = n,SpeechDestroy()
 same = n,GotoIf($[!${LEN(${safe_speech_text_0})}]?endloop)
 ; Can not do AstDB stuff here unless live_dangerously is set in asterisk.conf
 same = n,Set(SHARED(hh_last,${caller:1:-1})=${EPOCH})
 same = n,Set(SHARED(hh_word,${caller:1:-1})=${safe_speech_text_0})
 same = n,Set(SHARED(hh_score,${caller:1:-1})=${safe_speech_score_0})
 same = n,GotoIf($[!${LEN(${ogpritalk})}]?endloop)
 same = n,ChannelRedirect(${caller},${ogctx},${ogext},${ogpritalk})
 same = n(endloop),EndWhile()
 same = n,Hangup()


; GRAMMAR DIGITS EN US
;
; Current res_speech_vosk does not support grammars.
; Use this in a Gosub() to help out with detection of spoken numbers 0-20.
;
; Ordered Parameters:
;   ARG1 - word string to inspect if it is a spoken number
;
; Returns:
;   out - either the number as digit(s), or blank if NaN
;
[pngnpbx-hh-grammar-digits-en-us]
exten = s,1,DumpChan(5)
 same = n,Set(LOCAL(in)=${ARG1})
 same = n,Set(LOCAL(out)=)
 same = n,GotoIf($["${in}"==""]?i,1)
 same = n,GotoIf($["${in}"=="zero"]?0,1)
 same = n,GotoIf($["${in}"=="one"]?1,1)
 same = n,GotoIf($["${in}"=="two"]?2,1)
 same = n,GotoIf($["${in}"=="to"]?2,1)
 same = n,GotoIf($["${in}"=="too"]?2,1)
 same = n,GotoIf($["${in}"=="three"]?3,1)
 same = n,GotoIf($["${in}"=="tree"]?3,1)
 same = n,GotoIf($["${in}"=="thee"]?3,1)
 same = n,GotoIf($["${in}"=="the"]?3,1)
 same = n,GotoIf($["${in}"=="four"]?4,1)
 same = n,GotoIf($["${in}"=="for"]?4,1)
 same = n,GotoIf($["${in}"=="or"]?4,1)
 same = n,GotoIf($["${in}"=="five"]?5,1)
 same = n,GotoIf($["${in}"=="six"]?6,1)
 same = n,GotoIf($["${in}"=="seven"]?7,1)
 same = n,GotoIf($["${in}"=="eight"]?8,1)
 same = n,GotoIf($["${in}"=="ate"]?8,1)
 same = n,GotoIf($["${in}"=="nine"]?9,1)
 same = n,GotoIf($["${in}"=="ten"]?10,1)
 same = n,GotoIf($["${in}"=="eleven"]?11,1)
 same = n,GotoIf($["${in}"=="twelve"]?12,1)
 same = n,GotoIf($["${in}"=="thirteen"]?13,1)
 same = n,GotoIf($["${in}"=="fourteen"]?14,1)
 same = n,GotoIf($["${in}"=="fifteen"]?15,1)
 same = n,GotoIf($["${in}"=="sixteen"]?16,1)
 same = n,GotoIf($["${in}"=="seventeen"]?17,1)
 same = n,GotoIf($["${in}"=="eightteen"]?18,1)
 same = n,GotoIf($["${in}"=="nineteen"]?19,1)
 same = n,GotoIf($["${in}"=="twenty"]?20,1)
 same = n,Goto(i,1)

exten = i,1,Set(out=)
 same = n,Return(${out})
exten = _X,1,Set(out=${EXTEN})
 same = n,Return(${out})
exten = _XX,1,Set(out=${EXTEN})
 same = n,Return(${out})

; end of extensions_pngnpbx_hh.conf