0.0.99

* add openai STT options

* wip

* allow apikey passed as param for openai stt

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.50",
+  "version": "0.0.99",
   "description": "Jambonz Verb Specification Utilities",
   "main": "index.js",
   "scripts": {
@@ -20,7 +20,7 @@
   "devDependencies": {
     "eslint": "^7.32.0",
     "eslint-plugin-promise": "^4.3.1",
-    "tape": "^5.6.1"
+    "tape": "^5.7.5"
   },
   "dependencies": {
     "debug": "^4.3.4",

specs.json

@@ -1,4 +1,11 @@
 {
+  "answer": {
+    "properties": {
+      "id": "string"
+    },
+    "required": [
+    ]
+  },
   "sip:decline": {
     "properties": {
       "id": "string",
@@ -27,6 +34,7 @@
       "id": "string",
       "referTo": "string",
       "referredBy": "string",
+      "referredByDisplayName": "string",
       "headers": "object",
       "actionHook": "object|string",
       "eventHook": "object|string"
@@ -41,18 +49,35 @@
       "synthesizer": "#synthesizer",
       "recognizer": "#recognizer",
       "bargeIn": "#bargeIn",
+      "ttsStream": "#ttsStream",
       "record": "#recordOptions",
       "listen": "#listenOptions",
+      "stream": "#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",
+      "vad":"#vad",
+      "referHook": "object|string",
+      "earlyMedia": "boolean"
     },
     "required": []
   },
+  "fillerNoise": {
+    "properties": {
+      "enable": "boolean",
+      "url": "string",
+      "startDelaySecs": "number"
+    },
+    "required": [
+      "enable"
+    ]
+  },
   "listenOptions": {
     "properties": {
       "enable": "boolean",
@@ -68,17 +93,28 @@
       "passDtmf": "boolean",
       "playBeep": "boolean",
       "disableBidirectionalAudio": "boolean",
+      "bidirectionalAudio": "#bidirectionalAudio",
       "timeout": "number"
     },
     "required": [
       "enable"
     ]
   },
+  "ttsStream": {
+    "properties": {
+      "enable": "boolean",
+      "synthesizer": "#synthesizer"
+    },
+    "required": [
+      "enable"
+    ]
+  },
   "bargeIn": {
     "properties": {
       "enable": "boolean",
       "sticky": "boolean",
       "actionHook": "object|string",
+      "partialResultHook": "object|string",
       "input": "array",
       "finishOnKey": "string",
       "numDigits": "number",
@@ -102,6 +138,24 @@
       "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",
@@ -159,13 +213,14 @@
     "properties": {
       "id": "string",
       "text": "string|array",
+      "stream": "boolean",
       "loop": "number|string",
       "synthesizer": "#synthesizer",
       "earlyMedia": "boolean",
-      "disableTtsCache": "boolean"
+      "disableTtsCache": "boolean",
+      "closeStreamOnEmpty": "boolean"
     },
     "required": [
-      "text"
     ]
   },
   "gather": {
@@ -188,6 +243,7 @@
       "recognizer": "#recognizer",
       "play": "#play",
       "say": "#say",
+      "fillerNoise": "#fillerNoise",
       "actionHookDelayAction": "#actionHookDelayAction"
     },
     "required": [
@@ -198,6 +254,8 @@
       "id": "string",
       "name": "string",
       "beep": "boolean",
+      "memberTag": "string",
+      "speakOnlyTo": "string",
       "startConferenceOnEnter": "boolean",
       "endConferenceOnExit": "boolean",
       "endConferenceDuration": "number",
@@ -229,13 +287,18 @@
       "dtmfHook": "object|string",
       "headers": "object",
       "anchorMedia": "boolean",
+      "exitMediaPath": "boolean",
+      "boostAudioSignal": "number|string",
       "listen": "#listen",
+      "stream": "#listen",
       "target": ["#target"],
       "timeLimit": "number",
       "timeout": "number",
       "proxy": "string",
       "transcribe": "#transcribe",
-      "amd": "#amd"
+      "amd": "#amd",
+      "dub": ["#dub"],
+      "tag": "object"
     },
     "required": [
       "target"
@@ -246,15 +309,17 @@
       "id": "string",
       "credentials": "object|string",
       "project": "string",
+      "agent": "string",
       "environment": "string",
-      "region": {
+      "region": "string",
+      "model": {
         "type": "string",
-        "enum": ["europe-west1", "europe-west2", "australia-southeast1", "asia-northeast1"]
+        "enum": ["es", "cx"]
       },
       "lang": "string",
       "actionHook": "object|string",
       "eventHook": "object|string",
-      "events": "[string]",
+      "events": "array",
       "welcomeEvent": "string",
       "welcomeEventParams": "object",
       "noInputTimeout": "number",
@@ -262,7 +327,8 @@
       "passDtmfAsTextInput": "boolean",
       "thinkingMusic": "string",
       "tts": "#synthesizer",
-      "bargein": "boolean"
+      "bargein": "boolean",
+      "queryInput": "#queryInput"
     },
     "required": [
       "project",
@@ -270,6 +336,16 @@
       "lang"
     ]
   },
+  "queryInput": {
+    "properties": {
+      "text": "string",
+      "intent": "string",
+      "event": "string",
+      "dtmf": "string"
+    },
+    "required": [
+    ]
+  },
   "dtmf": {
     "properties": {
       "id": "string",
@@ -320,6 +396,35 @@
       "passDtmf": "boolean",
       "playBeep": "boolean",
       "disableBidirectionalAudio": "boolean",
+      "bidirectionalAudio": "#bidirectionalAudio",
+      "sampleRate": "number",
+      "timeout": "number",
+      "transcribe": "#transcribe",
+      "url": "string",
+      "wsAuth": "#auth",
+      "earlyMedia": "boolean",
+      "channel": "number"
+    },
+    "required": [
+      "url"
+    ]
+  },
+  "stream": {
+    "properties": {
+      "id": "string",
+      "actionHook": "object|string",
+      "auth": "#auth",
+      "finishOnKey": "string",
+      "maxLength": "number",
+      "metadata": "object",
+      "mixType": {
+        "type": "string",
+        "enum": ["mono", "stereo", "mixed"]
+      },
+      "passDtmf": "boolean",
+      "playBeep": "boolean",
+      "disableBidirectionalAudio": "boolean",
+      "bidirectionalAudio": "#bidirectionalAudio",
       "sampleRate": "number",
       "timeout": "number",
       "transcribe": "#transcribe",
@@ -331,6 +436,26 @@
       "url"
     ]
   },
+  "llm": {
+    "properties": {
+      "id": "string",
+      "vendor": "string",
+      "model": "string",
+      "auth": "object",
+      "connectOptions": "object",
+      "actionHook": "object|string",
+      "eventHook": "object|string",
+      "toolHook": "object|string",
+      "events": "array",
+      "llmOptions": "object"
+    },
+    "required": [
+      "vendor",
+      "model",
+      "auth",
+      "llmOptions"
+    ]
+  },
   "message": {
     "properties": {
       "id": "string",
@@ -423,7 +548,9 @@
       "timeout": "number",
       "amd": "#amd",
       "dual_streams": "boolean",
-      "sipRequestWithinDialogHook": "string"
+      "sipRequestWithinDialogHook": "string",
+      "referHook": "object|string",
+      "timeLimit": "number"
     },
     "required": [
       "call_hook",
@@ -444,8 +571,10 @@
     "properties": {
       "id": "string",
       "transcriptionHook": "string",
+      "translationHook": "string",
       "recognizer": "#recognizer",
-      "earlyMedia": "boolean"
+      "earlyMedia": "boolean",
+      "channel": "number"
     },
     "required": [
       "transcriptionHook"
@@ -508,7 +637,7 @@
       "fallbackVoice": "string|object",
       "engine": {
         "type": "string",
-        "enum": ["standard", "neural"]
+        "enum": ["standard", "neural", "generative", "long-form"]
       },
       "gender": {
         "type": "string",
@@ -591,6 +720,7 @@
       "asrDtmfTerminationDigit": "string",
       "asrTimeout": "number",
       "fastRecognitionTimeout": "number",
+      "minConfidence": "number",
       "nuanceOptions": "#nuanceOptions",
       "deepgramOptions": "#deepgramOptions",
       "ibmOptions": "#ibmOptions",
@@ -600,7 +730,11 @@
       "awsOptions": "#awsOptions",
       "azureOptions": "#azureOptions",
       "assemblyAiOptions": "#assemblyAiOptions",
-      "customOptions": "#customOptions"
+      "googleOptions": "#googleOptions",
+      "customOptions": "#customOptions",
+      "verbioOptions": "#verbioOptions",
+      "speechmaticsOptions": "#speechmaticsOptions",
+      "openaiOptions": "#openaiOptions"
     },
     "required": [
       "vendor"
@@ -610,6 +744,7 @@
     "properties": {
       "authToken": "string",
       "uri": "string",
+      "sampleRate": "number",
       "options": "object"
     },
     "required": [
@@ -640,11 +775,43 @@
   },
   "azureOptions": {
     "properties": {
-      "speechSegmentationSilenceTimeoutMs": "number"
+      "speechSegmentationSilenceTimeoutMs": "number",
+      "postProcessing" : "string",
+      "languageIdMode": {
+        "type": "string",
+        "enum": [
+          "AtStart",
+          "Continuous"
+        ]
+      },
+      "speechRecognitionMode": {
+        "type": "string",
+        "enum": [
+          "CONVERSATION",
+          "DICTATION",
+          "INTERACTIVE"
+        ]
+      }
     },
     "required": [
     ]
   },
+  "googleOptions" : {
+    "properties": {
+      "serviceVersion": {
+        "type": "string",
+        "enum": [
+          "v1",
+          "v2"
+        ]
+      },
+      "recognizerId": "string",
+      "speechStartTimeoutMs": "number",
+      "speechEndTimeoutMs": "number",
+      "enableVoiceActivityEvents": "boolean",
+      "transcriptNormalization": "array"
+    }
+  },
   "cobaltOptions": {
     "properties": {
       "serverUri": "string",
@@ -689,6 +856,8 @@
   },
   "deepgramOptions": {
     "properties": {
+      "deepgramSttUri": "string",
+      "deepgramSttUseTls": "boolean",
       "apiKey": "string",
       "tier": "string",
       "model": "string",
@@ -696,6 +865,7 @@
       "version": "string",
       "punctuate": "boolean",
       "smartFormatting": "boolean",
+      "noDelay": "boolean",
       "profanityFilter": "boolean",
       "redact": {
         "type": "string",
@@ -715,11 +885,13 @@
       "search": "array",
       "replace": "array",
       "keywords": "array",
+      "keyterms": "array",
       "endpointing": "boolean | number",
       "utteranceEndMs": "number",
       "shortUtterance": "boolean",
       "vadTurnoff": "number",
-      "tag": "string"
+      "tag": "string",
+      "fillerWords" : "boolean"
     }
   },
   "sonioxOptions": {
@@ -735,6 +907,147 @@
     "required": [
     ]
   },
+  "verbioOptions": {
+    "properties": {
+      "enable_formatting": "boolean",
+      "enable_diarization": "boolean",
+      "topic": "number",
+      "inline_grammar": "string",
+      "grammar_uri": "string",
+      "label": "string",
+      "recognition_timeout": "number",
+      "speech_complete_timeout": "number",
+      "speech_incomplete_timeout": "number"
+    },
+    "required": [
+    ]
+  },
+  "openaiOptions": {
+    "properties": {
+      "apiKey": "string",
+      "model": "string",
+      "prompt": "string",
+      "language": "string",
+      "input_audio_noise_reduction": {
+        "type": "string",
+        "enum": [
+          "near_field",
+          "far_field"
+        ]
+      },
+      "turn_detection": "#turnDetection"
+    },
+    "required": [
+      "model"
+    ]
+  },
+  "turnDetection": {
+    "properties": {
+      "type": {
+        "type": "string",
+        "enum": [
+          "none",
+          "server_vad"
+        ]
+      },
+      "threshold": "number",
+      "prefix_padding_ms": "number",
+      "silence_duration_ms": "number"
+    },
+    "required": [
+      "type"
+    ]
+  },
+  "speechmaticsOptions": {
+    "properties": {
+      "transcription_config": "#sm_transcriptionConfig",
+      "translation_config": "#sm_translationConfig",
+      "audio_events_config_config": "#sm_audioEventsConfig"
+    },
+    "required": [
+    ]
+  },
+  "sm_transcriptionConfig": {
+    "properties": {
+      "language": "string",
+      "domain": "string",
+      "additional_vocab": "array",
+      "diarization": "string",
+      "speaker_diarization_config": "#sm_speakerDiarizationConfig",
+      "enable_partials": "boolean",
+      "max_delay": "number",
+      "max_delay_mode": {
+        "type": "string",
+        "enum": [
+          "fixed",
+          "flexible"
+        ]
+      },
+      "output_locale": "string",
+      "punctuation_overrides": "#sm_puctuationOverrides",
+      "operating_point": "string",
+      "enable_entities": "boolean",
+      "audio_filtering_config": "#sm_audioFilteringConfig",
+      "transcript_filtering_config": "#sm_transcriptFilteringConfig"
+    },
+    "required": [
+    ]
+  },
+  "sm_speakerDiarizationConfig": {
+    "properties": {
+      "speaker_sensitivity": "number",
+      "max_speakers": "number"
+    },
+    "required": [
+    ]
+  },
+  "sm_puctuationOverrides": {
+    "properties": {
+      "permitted_marks": "array",
+      "sensitivity": "number"
+    },
+    "required": [
+    ]
+  },
+  "sm_audioFilteringConfig": {
+    "properties": {
+      "volume_threshold": "number"
+    },
+    "required": [
+      "volume_threshold"
+    ]
+  },
+  "sm_transcriptFilteringConfig": {
+    "properties": {
+      "remove_disfluencies": "boolean"
+    },
+    "required": [
+      "remove_disfluencies"
+    ]
+  },
+  "sm_translationConfig": {
+    "properties": {
+      "target_languages": "array",
+      "enable_partials": "boolean"
+    },
+    "required": [
+      "target_languages"
+    ]
+  },
+  "sm_audioEventsConfig": {
+    "properties": {
+      "types": {
+        "type": "array",
+        "enum": [
+          "applause",
+          "music",
+          "laughter"
+        ]
+      }
+    },
+    "required": [
+    ]
+  },
   "sonioxStorage": {
     "properties": {
       "id": "string",
@@ -873,6 +1186,8 @@
     "properties": {
       "enable": "boolean",
       "voiceMs": "number",
+      "silenceMs": "number",
+      "strategy": "string",
       "mode": "number"
     },
     "required": [
@@ -883,6 +1198,7 @@
     "properties": {
       "actionHook": "object|string",
       "thresholdWordCount": "number",
+      "digitCount": "number",
       "timers": "#amdTimers",
       "recognizer": "#recognizer"
     },
@@ -904,7 +1220,15 @@
       "noResponseTimeout": "number",
       "noResponseGiveUpTimeout": "number",
       "retries": "number",
-      "actions": "array"
+      "actions": "array",
+      "giveUpActions": "array"
+    }
+  },
+  "bidirectionalAudio" : {
+    "properties": {
+      "enabled": "boolean",
+      "streaming": "boolean",
+      "sampleRate": "number"
     }
   }
-}
+}

test/jambonz-verb-test.js

@@ -81,6 +81,27 @@ test("validate correct verbs", async (t) => {
         }
       }
     },
+    {
+      "verb": "config",
+      "referHook": "https://referhook.com"
+    },
+    {
+      "verb": "config",
+      "referHook": {
+        "url": "https://referhook.com"
+      }
+    },
+    {
+      "verb": "config",
+      "recognizer": {
+        "vendor": "google",
+        "language": "de-DE",
+        "label": "label1",
+        "azureOptions": {
+          "languageIdMode": "Continuous"
+        }
+      }
+    },
     {
       "verb": "config",
       "record": {
@@ -118,6 +139,7 @@ test("validate correct verbs", async (t) => {
       "callerName": "Tom",
       "answerOnBridge": true,
       "dtmfCapture": ["*2", "*3"],
+      "timeLimit": 10,
       "dtmfHook": {
         "url": "/dtmf",
         "method": "GET"
@@ -183,7 +205,8 @@ test("validate correct verbs", async (t) => {
         "hintsBoost": 10,
         "fastRecognitionTimeout": 2000,
         "deepgramOptions": {
-          "endpointing": 500
+          "endpointing": 500,
+          "noDelay": true,
         }
       },
       "say": {
@@ -217,6 +240,31 @@ test("validate correct verbs", async (t) => {
         }
       }
     },
+    {
+      "verb": "gather",
+      "actionHook": "http://example.com/collect",
+      "input": ["digits", "speech"],
+      "bargein": true,
+      "dtmfBargein": true,
+      "finishOnKey": "#",
+      "numDigits": 5,
+      "timeout": 8,
+      "recognizer": {
+        "vendor": "verbio",
+        "language": "en-US",
+        "verbioOptions": {
+          "enable_formatting": true,
+          "enable_diarization": true,
+          "topic": 0,
+          "inline_grammar": "this is inline grammar",
+          "grammar_uri": "https://grammar_uri.com",
+          "label": "label",
+          "recognition_timeout": 500,
+          "speech_complete_timeout": 500,
+          "speech_incomplete_timeout": 500,
+        }
+      },
+    },
     {
       "verb": "gather",
       "actionHook": "http://example.com/collect",
@@ -303,6 +351,39 @@ 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": "config",
+      "vad": {
+        "enable": true,
+        "voiceMs": 250,
+        "silenceMs": 150,
+        "strategy": "one-shot",
+        "mode": 2
+      }
+    },
     {
       "verb": "message",
       "to": "15083084809",
@@ -361,6 +442,12 @@ test("validate correct verbs", async (t) => {
       "referTo": "+15083084809",
       "actionHook": "/action"
     },
+    {
+      "verb": "sip:refer",
+      "referTo": "+15083084809",
+      "referredByDisplayName": "Alice",
+      "actionHook": "/action"
+    },
     {
       "verb": "tag",
       "data": {
@@ -378,6 +465,39 @@ test("validate correct verbs", async (t) => {
         "interim": true
       }
     },
+    {
+      "verb": "transcribe",
+      "transcriptionHook": "http://example.com/transcribe",
+      "recognizer": {
+        "vendor": "nvidia",
+        "language" : "en-US",
+        "customOptions": {
+          "sampleRate": 16000
+        }
+      }
+    },
+    {
+      "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",
@@ -400,7 +520,8 @@ test("validate correct verbs", async (t) => {
         "actionHook": "/answeringMachineDetection",
     
       },
-      "dual_streams": true
+      "dual_streams": true,
+      "timeLimit" : 10
     }
   ];
   try {