0.0.70

* support google stt version 2

* fix review comments

* fix review comment

README.md

@@ -1,4 +1,4 @@
-# verb-specificiations
+# verb-specifications
 Jambonz Verb Specification Utilities
 
 #### Running the test suite

jambonz-app-json-validation.js

@@ -58,7 +58,8 @@ function validateVerb(name, data, logger) {
       const dSpec = specData.properties[dKey];
       debug(`Task:validate validating property ${dKey} with value ${JSON.stringify(dVal)}`);
 
-      if (typeof dSpec === 'string' && dSpec === 'array') {
+      if (typeof dVal === 'undefined') continue;
+      else if (typeof dSpec === 'string' && dSpec === 'array') {
         if (!Array.isArray(dVal)) throw new Error(`${name}: property ${dKey} is not an array`);
       }
       else if (typeof dSpec === 'string' && dSpec.includes('|')) {

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.30",
+  "version": "0.0.70",
   "lockfileVersion": 2,
   "requires": true,
   "packages": {
     "": {
       "name": "@jambonz/verb-specifications",
-      "version": "0.0.30",
+      "version": "0.0.70",
       "license": "MIT",
       "dependencies": {
         "debug": "^4.3.4",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.30",
+  "version": "0.0.70",
   "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",
@@ -43,13 +50,28 @@
       "bargeIn": "#bargeIn",
       "record": "#recordOptions",
       "listen": "#listenOptions",
+      "transcribe": "#transcribeOptions",
       "amd": "#amd",
+      "fillerNoise": "#fillerNoise",
       "notifyEvents": "boolean",
       "reset": "string|array",
-      "onHoldMusic": "string"
+      "onHoldMusic": "string",
+      "actionHookDelayAction": "#actionHookDelayAction",
+      "sipRequestWithinDialogHook": "object|string",
+      "boostAudioSignal": "number|string"
     },
     "required": []
   },
+  "fillerNoise": {
+    "properties": {
+      "enable": "boolean",
+      "url": "string",
+      "startDelaySecs": "number"
+    },
+    "required": [
+      "enable"
+    ]
+  },
   "listenOptions": {
     "properties": {
       "enable": "boolean",
@@ -60,7 +82,13 @@
         "type": "string",
         "enum": ["mono", "stereo", "mixed"]
       },
-      "metadata": "object"
+      "metadata": "object",
+      "maxLength": "number",
+      "passDtmf": "boolean",
+      "playBeep": "boolean",
+      "disableBidirectionalAudio": "boolean",
+      "bidirectionalAudio": "#bidirectionalAudio",
+      "timeout": "number"
     },
     "required": [
       "enable"
@@ -71,6 +99,7 @@
       "enable": "boolean",
       "sticky": "boolean",
       "actionHook": "object|string",
+      "partialResultHook": "object|string",
       "input": "array",
       "finishOnKey": "string",
       "numDigits": "number",
@@ -84,6 +113,34 @@
       "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",
@@ -169,7 +226,9 @@
       "timeout": "number",
       "recognizer": "#recognizer",
       "play": "#play",
-      "say": "#say"
+      "say": "#say",
+      "fillerNoise": "#fillerNoise",
+      "actionHookDelayAction": "#actionHookDelayAction"
     },
     "required": [
     ]
@@ -179,6 +238,8 @@
       "id": "string",
       "name": "string",
       "beep": "boolean",
+      "memberTag": "string",
+      "speakOnlyTo": "string",
       "startConferenceOnEnter": "boolean",
       "endConferenceOnExit": "boolean",
       "endConferenceDuration": "number",
@@ -210,13 +271,16 @@
       "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"
@@ -301,6 +365,7 @@
       "passDtmf": "boolean",
       "playBeep": "boolean",
       "disableBidirectionalAudio": "boolean",
+      "bidirectionalAudio": "#bidirectionalAudio",
       "sampleRate": "number",
       "timeout": "number",
       "transcribe": "#transcribe",
@@ -402,7 +467,9 @@
       "to": "#target",
       "headers": "object",
       "timeout": "number",
-      "amd": "#amd"
+      "amd": "#amd",
+      "dual_streams": "boolean",
+      "sipRequestWithinDialogHook": "string"
     },
     "required": [
       "call_hook",
@@ -424,7 +491,8 @@
       "id": "string",
       "transcriptionHook": "string",
       "recognizer": "#recognizer",
-      "earlyMedia": "boolean"
+      "earlyMedia": "boolean",
+      "channel": "number"
     },
     "required": [
       "transcriptionHook"
@@ -450,7 +518,8 @@
       "vmail": "boolean",
       "tenant": "string",
       "trunk": "string",
-      "overrideTo": "string"
+      "overrideTo": "string",
+      "proxy": "string"
     },
     "required": [
       "type"
@@ -479,11 +548,11 @@
       "vendor": "string",
       "label": "string",
       "language": "string",
-      "voice": "string",
+      "voice": "string|object",
       "fallbackVendor": "string",
       "fallbackLabel": "string",
       "fallbackLanguage": "string",
-      "fallbackVoice": "string",
+      "fallbackVoice": "string|object",
       "engine": {
         "type": "string",
         "enum": ["standard", "neural"]
@@ -568,11 +637,17 @@
       "azureSttEndpointId": "string",
       "asrDtmfTerminationDigit": "string",
       "asrTimeout": "number",
+      "fastRecognitionTimeout": "number",
       "nuanceOptions": "#nuanceOptions",
       "deepgramOptions": "#deepgramOptions",
       "ibmOptions": "#ibmOptions",
       "nvidiaOptions": "#nvidiaOptions",
       "sonioxOptions": "#sonioxOptions",
+      "cobaltOptions": "#cobaltOptions",
+      "awsOptions": "#awsOptions",
+      "azureOptions": "#azureOptions",
+      "assemblyAiOptions": "#assemblyAiOptions",
+      "googleOptions": "#googleOptions",
       "customOptions": "#customOptions"
     },
     "required": [
@@ -588,6 +663,71 @@
     "required": [
     ]
   },
+  "awsOptions": {
+    "properties": {
+      "accessKey": "string",
+      "secretKey": "string",
+      "securityToken": "string",
+      "region": "string",
+      "vocabularyName": "string",
+      "vocabularyFilterName": "string",
+      "vocabularyFilterMethod": {
+        "type": "string",
+        "enum": [
+          "remove",
+          "mask",
+          "tag"
+        ]
+      },
+      "languageModelName": "string",
+      "piiEntityTypes": "array",
+      "piiIdentifyEntities": "boolean"
+    },
+    "required": [
+    ]
+  },
+  "azureOptions": {
+    "properties": {
+      "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",
+      "enableConfusionNetwork": "boolean",
+      "metadata": "string",
+      "compiledContextData": "string",
+      "wordTimeOffsets": "boolean",
+      "contextToken": "string"
+    },
+    "required": [
+    ]
+  },
   "nvidiaOptions": {
     "properties": {
       "rivaUri": "string",
@@ -620,12 +760,15 @@
   },
   "deepgramOptions": {
     "properties": {
+      "deepgramSttUri": "string",
+      "deepgramSttUseTls": "boolean",
       "apiKey": "string",
       "tier": "string",
       "model": "string",
       "customModel": "string",
       "version": "string",
       "punctuate": "boolean",
+      "smartFormatting": "boolean",
       "profanityFilter": "boolean",
       "redact": {
         "type": "string",
@@ -646,6 +789,8 @@
       "replace": "array",
       "keywords": "array",
       "endpointing": "boolean | number",
+      "utteranceEndMs": "number",
+      "shortUtterance": "boolean",
       "vadTurnoff": "number",
       "tag": "string"
     }
@@ -720,6 +865,11 @@
     "required": [
     ]
   },
+  "assemblyAiOptions": {
+    "properties": {
+      "apiKey": "string"
+    }
+  },
   "resource": {
     "properties": {
       "externalReference": "#resourceReference",
@@ -820,5 +970,21 @@
       "toneTimeoutMs": "number",
       "greetingCompletionTimeoutMs": "number"
     }
+  },
+  "actionHookDelayAction" : {
+    "properties": {
+      "enabled": "boolean",
+      "noResponseTimeout": "number",
+      "noResponseGiveUpTimeout": "number",
+      "retries": "number",
+      "actions": "array"
+    }
+  },
+  "bidirectionalAudio" : {
+    "properties": {
+      "enabled": "boolean",
+      "streaming": "boolean",
+      "sampleRate": "number"
+    }
   }
-}
+}

test/jambonz-verb-test.js

@@ -37,7 +37,60 @@ test("validate correct verbs", async (t) => {
         "input" : ["speech"],
         "actionHook": "/userInput"
       },
-      "onHoldMusic": "http://server.com/hold"
+      "transcribe": {
+        "enable": true,
+        "transcriptionHook": "http://server.com/hook",
+        "recognizer": {
+          "vendor": "google",
+          "language": "de-DE",
+        }
+      },
+      "onHoldMusic": "http://server.com/hold",
+      "actionHookDelayAction": {
+        "enabled": true,
+        "noResponseTimeout": 5,
+        "noResponseGiveUpTimeout": 10,
+        "retries": 3,
+        "actions": [
+          {
+            "verb": "say",
+            "text": "To speak to Sales press 1 or say Sales.  To speak to customer support press 2 or say Support",
+            "synthesizer": {
+              "vendor": "google",
+              "language": "en-US"
+            }
+          },
+          {
+            "verb": "play",
+            "url": "https://example.com/example.mp3",
+            "timeoutSecs": 10,
+            "seekOffset": 8000,
+            "actionHook": "/play/action"
+          }
+        ]
+      }
+    },
+    {
+      "verb": "config",
+      "recognizer": {
+        "vendor": "google",
+        "language": "de-DE",
+        "label": "label1",
+        "assemblyAiOptions": {
+          "apiKey": "apikey"
+        }
+      }
+    },
+    {
+      "verb": "config",
+      "recognizer": {
+        "vendor": "google",
+        "language": "de-DE",
+        "label": "label1",
+        "azureOptions": {
+          "languageIdMode": "Continuous"
+        }
+      }
     },
     {
       "verb": "config",
@@ -50,6 +103,10 @@ test("validate correct verbs", async (t) => {
       }
       }
     },
+    {
+      "verb": "config",
+      "sipRequestWithinDialogHook": "https://jambonz.or/sipIndialogActionHook"
+    },
     {
       "verb": "config",
       "record": {
@@ -135,6 +192,7 @@ test("validate correct verbs", async (t) => {
         "language": "en-US",
         "hints": ["sales", "support"],
         "hintsBoost": 10,
+        "fastRecognitionTimeout": 2000,
         "deepgramOptions": {
           "endpointing": 500
         }
@@ -150,6 +208,24 @@ test("validate correct verbs", async (t) => {
           "fallbackLabel": "label1",
           "fallbackVoice": "voice"
         }
+      },
+      "say": {
+        "text": "To speak to Sales press 1 or say Sales.  To speak to customer support press 2 or say Support",
+        "synthesizer": {
+          "vendor": "google",
+          "language": "en-US",
+          "voice": {
+            "reportedUsage":"REALTIME",
+            "model":"path/to/model",
+           },
+          "fallbackVendor": "google",
+          "fallbackLanguage": "en-US",
+          "fallbackLabel": "label1",
+          "fallbackVoice": {
+            "reportedUsage":"REALTIME",
+            "model":"path/to/model",
+           }
+        }
       }
     },
     {
@@ -173,6 +249,28 @@ test("validate correct verbs", async (t) => {
           "endpointing": true
         }
       },
+      "actionHookDelayAction": {
+        "noResponseTimeout": 5,
+        "noResponseGiveUpTimeout": 10,
+        "retries": 3,
+        "actions": [
+          {
+            "verb": "say",
+            "text": "To speak to Sales press 1 or say Sales.  To speak to customer support press 2 or say Support",
+            "synthesizer": {
+              "vendor": "google",
+              "language": "en-US"
+            }
+          },
+          {
+            "verb": "play",
+            "url": "https://example.com/example.mp3",
+            "timeoutSecs": 10,
+            "seekOffset": 8000,
+            "actionHook": "/play/action"
+          }
+        ]
+      }
     },
     {
       "verb": "hangup",
@@ -216,6 +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",
@@ -291,6 +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",
@@ -312,7 +455,8 @@ test("validate correct verbs", async (t) => {
       "amd": {
         "actionHook": "/answeringMachineDetection",
     
-      }
+      },
+      "dual_streams": true
     }
   ];
   try {