0.0.69

* support google stt version 2

* fix review comments

* fix review comment

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.49",
+  "version": "0.0.69",
   "lockfileVersion": 2,
   "requires": true,
   "packages": {
     "": {
       "name": "@jambonz/verb-specifications",
-      "version": "0.0.49",
+      "version": "0.0.69",
       "license": "MIT",
       "dependencies": {
         "debug": "^4.3.4",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.49",
+  "version": "0.0.69",
   "description": "Jambonz Verb Specification Utilities",
   "main": "index.js",
   "scripts": {

specs.json

@@ -1,4 +1,11 @@
 {
+  "answer": {
+    "properties": {
+      "id": "string"
+    },
+    "required": [
+    ]
+  },
   "sip:decline": {
     "properties": {
       "id": "string",
@@ -43,15 +50,28 @@
       "bargeIn": "#bargeIn",
       "record": "#recordOptions",
       "listen": "#listenOptions",
+      "transcribe": "#transcribeOptions",
       "amd": "#amd",
+      "fillerNoise": "#fillerNoise",
       "notifyEvents": "boolean",
       "reset": "string|array",
       "onHoldMusic": "string",
       "actionHookDelayAction": "#actionHookDelayAction",
-      "sipRequestWithinDialogHook": "object|string"
+      "sipRequestWithinDialogHook": "object|string",
+      "boostAudioSignal": "number|string"
     },
     "required": []
   },
+  "fillerNoise": {
+    "properties": {
+      "enable": "boolean",
+      "url": "string",
+      "startDelaySecs": "number"
+    },
+    "required": [
+      "enable"
+    ]
+  },
   "listenOptions": {
     "properties": {
       "enable": "boolean",
@@ -67,6 +87,7 @@
       "passDtmf": "boolean",
       "playBeep": "boolean",
       "disableBidirectionalAudio": "boolean",
+      "bidirectionalAudio": "#bidirectionalAudio",
       "timeout": "number"
     },
     "required": [
@@ -91,6 +112,34 @@
       "enable"
     ]
   },
+  "transcribeOptions": {
+    "properties": {
+      "enable": "boolean",
+      "transcriptionHook": "string",
+      "recognizer": "#recognizer"
+    },
+    "required": [
+      "enable"
+    ]
+  },
+  "dub": {
+    "properties": {
+      "id": "string",
+      "action": {
+        "type": "string",
+        "enum": ["addTrack", "removeTrack", "silenceTrack", "playOnTrack", "sayOnTrack"]
+      },
+      "track": "string",
+      "play": "string",
+      "say": "string|object",
+      "loop": "boolean",
+      "gain": "number|string"
+    },
+    "required": [
+      "action",
+      "track"
+    ]
+  },
   "dequeue": {
     "properties": {
       "id": "string",
@@ -177,6 +226,7 @@
       "recognizer": "#recognizer",
       "play": "#play",
       "say": "#say",
+      "fillerNoise": "#fillerNoise",
       "actionHookDelayAction": "#actionHookDelayAction"
     },
     "required": [
@@ -187,6 +237,8 @@
       "id": "string",
       "name": "string",
       "beep": "boolean",
+      "memberTag": "string",
+      "speakOnlyTo": "string",
       "startConferenceOnEnter": "boolean",
       "endConferenceOnExit": "boolean",
       "endConferenceDuration": "number",
@@ -218,13 +270,16 @@
       "dtmfHook": "object|string",
       "headers": "object",
       "anchorMedia": "boolean",
+      "boostAudioSignal": "number|string",
       "listen": "#listen",
       "target": ["#target"],
       "timeLimit": "number",
       "timeout": "number",
       "proxy": "string",
       "transcribe": "#transcribe",
-      "amd": "#amd"
+      "amd": "#amd",
+      "dub": ["#dub"],
+      "tag": "object"
     },
     "required": [
       "target"
@@ -309,6 +364,7 @@
       "passDtmf": "boolean",
       "playBeep": "boolean",
       "disableBidirectionalAudio": "boolean",
+      "bidirectionalAudio": "#bidirectionalAudio",
       "sampleRate": "number",
       "timeout": "number",
       "transcribe": "#transcribe",
@@ -434,7 +490,8 @@
       "id": "string",
       "transcriptionHook": "string",
       "recognizer": "#recognizer",
-      "earlyMedia": "boolean"
+      "earlyMedia": "boolean",
+      "channel": "number"
     },
     "required": [
       "transcriptionHook"
@@ -589,6 +646,7 @@
       "awsOptions": "#awsOptions",
       "azureOptions": "#azureOptions",
       "assemblyAiOptions": "#assemblyAiOptions",
+      "googleOptions": "#googleOptions",
       "customOptions": "#customOptions"
     },
     "required": [
@@ -629,11 +687,34 @@
   },
   "azureOptions": {
     "properties": {
-      "speechSegmentationSilenceTimeoutMs": "number"
+      "speechSegmentationSilenceTimeoutMs": "number",
+      "languageIdMode": {
+        "type": "string",
+        "enum": [
+          "AtStart",
+          "Continuous"
+        ]
+      }
     },
     "required": [
     ]
   },
+  "googleOptions" : {
+    "properties": {
+      "serviceVersion": {
+        "type": "string",
+        "enum": [
+          "v1",
+          "v2"
+        ]
+      },
+      "recognizerId": "string",
+      "speechStartTimeoutMs": "number",
+      "speechEndTimeoutMs": "number",
+      "enableVoiceActivityEvents": "boolean",
+      "transcriptNormalization": "array"
+    }
+  },
   "cobaltOptions": {
     "properties": {
       "serverUri": "string",
@@ -678,6 +759,8 @@
   },
   "deepgramOptions": {
     "properties": {
+      "deepgramSttUri": "string",
+      "deepgramSttUseTls": "boolean",
       "apiKey": "string",
       "tier": "string",
       "model": "string",
@@ -895,5 +978,12 @@
       "retries": "number",
       "actions": "array"
     }
+  },
+  "bidirectionalAudio" : {
+    "properties": {
+      "enabled": "boolean",
+      "streaming": "boolean",
+      "sampleRate": "number"
+    }
   }
-}
+}

test/jambonz-verb-test.js

@@ -37,6 +37,14 @@ test("validate correct verbs", async (t) => {
         "input" : ["speech"],
         "actionHook": "/userInput"
       },
+      "transcribe": {
+        "enable": true,
+        "transcriptionHook": "http://server.com/hook",
+        "recognizer": {
+          "vendor": "google",
+          "language": "de-DE",
+        }
+      },
       "onHoldMusic": "http://server.com/hold",
       "actionHookDelayAction": {
         "enabled": true,
@@ -73,6 +81,17 @@ test("validate correct verbs", async (t) => {
         }
       }
     },
+    {
+      "verb": "config",
+      "recognizer": {
+        "vendor": "google",
+        "language": "de-DE",
+        "label": "label1",
+        "azureOptions": {
+          "languageIdMode": "Continuous"
+        }
+      }
+    },
     {
       "verb": "config",
       "record": {
@@ -295,6 +314,29 @@ test("validate correct verbs", async (t) => {
       "url": "wss://myrecorder.example.com/calls",
       "mixType" : "stereo"
     },
+    {
+      "verb": "listen",
+      "url": "wss://myrecorder.example.com/calls",
+      "mixType" : "stereo",
+      "bidirectionalAudio": {
+        enabled: true,
+        streaming: true,
+        sampleRate: 8000
+     }
+    },
+    {
+      "verb": "config",
+      "listen": {
+        "enable": true,
+        "url": "wss://myrecorder.example.com/calls",
+        "mixType" : "stereo",
+        "bidirectionalAudio": {
+          enabled: true,
+          streaming: true,
+          sampleRate: 8000
+       }
+      }
+    },
     {
       "verb": "message",
       "to": "15083084809",
@@ -370,6 +412,28 @@ test("validate correct verbs", async (t) => {
         "interim": true
       }
     },
+    {
+      "verb": "transcribe",
+      "transcriptionHook": "http://example.com/transcribe",
+      "recognizer": {
+        "vendor": "nvidia",
+        "language" : "en-US",
+        "interim": true,
+        "googleOptions": {
+          "serviceVersion": "v2",
+          "speechStartTimeoutMs": 500,
+          "speechEndTimeoutMs": 1000,
+          "enableVoiceActivityEvents": true,
+          "transcriptNormalization" : [
+            {
+              "search": "dog",
+              "replace": "cat",
+              "case_sensitive": true
+            }
+          ]
+        }
+      }
+    },
     {
       "verb": "rest:dial",
       "account_sid": "1291964182631236912836123912",