0.0.107

* support assembly ai v3

* wip

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.69",
+  "version": "0.0.107",
   "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

@@ -34,6 +34,7 @@
       "id": "string",
       "referTo": "string",
       "referredBy": "string",
+      "referredByDisplayName": "string",
       "headers": "object",
       "actionHook": "object|string",
       "eventHook": "object|string"
@@ -48,8 +49,10 @@
       "synthesizer": "#synthesizer",
       "recognizer": "#recognizer",
       "bargeIn": "#bargeIn",
+      "ttsStream": "#ttsStream",
       "record": "#recordOptions",
       "listen": "#listenOptions",
+      "stream": "#listenOptions",
       "transcribe": "#transcribeOptions",
       "amd": "#amd",
       "fillerNoise": "#fillerNoise",
@@ -58,7 +61,11 @@
       "onHoldMusic": "string",
       "actionHookDelayAction": "#actionHookDelayAction",
       "sipRequestWithinDialogHook": "object|string",
-      "boostAudioSignal": "number|string"
+      "boostAudioSignal": "number|string",
+      "vad":"#vad",
+      "referHook": "object|string",
+      "earlyMedia": "boolean",
+      "autoStreamTts": "boolean"
     },
     "required": []
   },
@@ -94,11 +101,21 @@
       "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",
@@ -197,13 +214,15 @@
     "properties": {
       "id": "string",
       "text": "string|array",
+      "instructions": "string",
+      "stream": "boolean",
       "loop": "number|string",
       "synthesizer": "#synthesizer",
       "earlyMedia": "boolean",
-      "disableTtsCache": "boolean"
+      "disableTtsCache": "boolean",
+      "closeStreamOnEmpty": "boolean"
     },
     "required": [
-      "text"
     ]
   },
   "gather": {
@@ -270,8 +289,10 @@
       "dtmfHook": "object|string",
       "headers": "object",
       "anchorMedia": "boolean",
+      "exitMediaPath": "boolean",
       "boostAudioSignal": "number|string",
       "listen": "#listen",
+      "stream": "#listen",
       "target": ["#target"],
       "timeLimit": "number",
       "timeout": "number",
@@ -290,15 +311,17 @@
       "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",
@@ -306,7 +329,8 @@
       "passDtmfAsTextInput": "boolean",
       "thinkingMusic": "string",
       "tts": "#synthesizer",
-      "bargein": "boolean"
+      "bargein": "boolean",
+      "queryInput": "#queryInput"
     },
     "required": [
       "project",
@@ -314,6 +338,16 @@
       "lang"
     ]
   },
+  "queryInput": {
+    "properties": {
+      "text": "string",
+      "intent": "string",
+      "event": "string",
+      "dtmf": "string"
+    },
+    "required": [
+    ]
+  },
   "dtmf": {
     "properties": {
       "id": "string",
@@ -350,6 +384,34 @@
     ]
   },
   "listen": {
+    "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",
+      "url": "string",
+      "wsAuth": "#auth",
+      "earlyMedia": "boolean",
+      "channel": "number"
+    },
+    "required": [
+      "url"
+    ]
+  },
+  "stream": {
     "properties": {
       "id": "string",
       "actionHook": "object|string",
@@ -376,6 +438,36 @@
       "url"
     ]
   },
+  "llm": {
+    "properties": {
+      "id": "string",
+      "vendor": "string",
+      "model": "string",
+      "auth": "object",
+      "connectOptions": "object",
+      "mcpServers": ["#mcpServer"],
+      "actionHook": "object|string",
+      "eventHook": "object|string",
+      "toolHook": "object|string",
+      "events": "array",
+      "llmOptions": "object"
+    },
+    "required": [
+      "vendor",
+      "auth",
+      "llmOptions"
+    ]
+  },
+  "mcpServer": {
+    "properties": {
+      "url": "string",
+      "auth": "object",
+      "roots": ["#root"]
+    },
+    "required": [
+      "url"
+    ]
+  },
   "message": {
     "properties": {
       "id": "string",
@@ -468,7 +560,9 @@
       "timeout": "number",
       "amd": "#amd",
       "dual_streams": "boolean",
-      "sipRequestWithinDialogHook": "string"
+      "sipRequestWithinDialogHook": "string",
+      "referHook": "object|string",
+      "timeLimit": "number"
     },
     "required": [
       "call_hook",
@@ -489,6 +583,7 @@
     "properties": {
       "id": "string",
       "transcriptionHook": "string",
+      "translationHook": "string",
       "recognizer": "#recognizer",
       "earlyMedia": "boolean",
       "channel": "number"
@@ -554,7 +649,7 @@
       "fallbackVoice": "string|object",
       "engine": {
         "type": "string",
-        "enum": ["standard", "neural"]
+        "enum": ["standard", "neural", "generative", "long-form"]
       },
       "gender": {
         "type": "string",
@@ -637,6 +732,7 @@
       "asrDtmfTerminationDigit": "string",
       "asrTimeout": "number",
       "fastRecognitionTimeout": "number",
+      "minConfidence": "number",
       "nuanceOptions": "#nuanceOptions",
       "deepgramOptions": "#deepgramOptions",
       "ibmOptions": "#ibmOptions",
@@ -647,7 +743,10 @@
       "azureOptions": "#azureOptions",
       "assemblyAiOptions": "#assemblyAiOptions",
       "googleOptions": "#googleOptions",
-      "customOptions": "#customOptions"
+      "customOptions": "#customOptions",
+      "verbioOptions": "#verbioOptions",
+      "speechmaticsOptions": "#speechmaticsOptions",
+      "openaiOptions": "#openaiOptions"
     },
     "required": [
       "vendor"
@@ -657,6 +756,7 @@
     "properties": {
       "authToken": "string",
       "uri": "string",
+      "sampleRate": "number",
       "options": "object"
     },
     "required": [
@@ -688,12 +788,21 @@
   "azureOptions": {
     "properties": {
       "speechSegmentationSilenceTimeoutMs": "number",
+      "postProcessing" : "string",
       "languageIdMode": {
         "type": "string",
         "enum": [
           "AtStart",
           "Continuous"
         ]
+      },
+      "speechRecognitionMode": {
+        "type": "string",
+        "enum": [
+          "CONVERSATION",
+          "DICTATION",
+          "INTERACTIVE"
+        ]
       }
     },
     "required": [
@@ -768,6 +877,7 @@
       "version": "string",
       "punctuate": "boolean",
       "smartFormatting": "boolean",
+      "noDelay": "boolean",
       "profanityFilter": "boolean",
       "redact": {
         "type": "string",
@@ -787,11 +897,13 @@
       "search": "array",
       "replace": "array",
       "keywords": "array",
+      "keyterms": "array",
       "endpointing": "boolean | number",
       "utteranceEndMs": "number",
       "shortUtterance": "boolean",
       "vadTurnoff": "number",
-      "tag": "string"
+      "tag": "string",
+      "fillerWords" : "boolean"
     }
   },
   "sonioxOptions": {
@@ -807,6 +919,165 @@
     "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",
+      "promptTemplates": "#promptTemplates",
+      "language": "string",
+      "input_audio_noise_reduction": {
+        "type": "string",
+        "enum": [
+          "near_field",
+          "far_field"
+        ]
+      },
+      "turn_detection": "#turnDetection"
+    },
+    "required": [
+    ]
+  },
+  "promptTemplates": {
+    "properties": {
+      "hintsTemplate": "string",
+      "conversationHistoryTemplate": "string"
+    },
+    "required": [
+    ]
+  },
+  "turnDetection": {
+    "properties": {
+      "type": {
+        "type": "string",
+        "enum": [
+          "none",
+          "server_vad",
+          "semantic_vad"
+        ]
+      },
+      "eagerness": {
+        "type": "string",
+        "enum": [
+          "low",
+          "medium",
+          "high",
+          "auto"
+        ]
+      },
+      "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",
@@ -866,7 +1137,18 @@
   },
   "assemblyAiOptions": {
     "properties": {
-      "apiKey": "string"
+      "apiKey": "string",
+      "serviceVersion": {
+        "type": "string",
+        "enum": [
+          "v2",
+          "v3"
+        ]
+      },
+      "formatTurns": "boolean",
+      "endOfTurnConfidenceThreshold": "number",
+      "minEndOfTurnSilenceWhenConfident": "number",
+      "maxTurnSilence": "number"
     }
   },
   "resource": {
@@ -945,6 +1227,8 @@
     "properties": {
       "enable": "boolean",
       "voiceMs": "number",
+      "silenceMs": "number",
+      "strategy": "string",
       "mode": "number"
     },
     "required": [
@@ -955,6 +1239,7 @@
     "properties": {
       "actionHook": "object|string",
       "thresholdWordCount": "number",
+      "digitCount": "number",
       "timers": "#amdTimers",
       "recognizer": "#recognizer"
     },
@@ -976,7 +1261,8 @@
       "noResponseTimeout": "number",
       "noResponseGiveUpTimeout": "number",
       "retries": "number",
-      "actions": "array"
+      "actions": "array",
+      "giveUpActions": "array"
     }
   },
   "bidirectionalAudio" : {

test/jambonz-verb-test.js

@@ -77,10 +77,25 @@ test("validate correct verbs", async (t) => {
         "language": "de-DE",
         "label": "label1",
         "assemblyAiOptions": {
-          "apiKey": "apikey"
+          "apiKey": "apikey",
+          "serviceVersion": "v3",
+          "formatTurns": true,
+          "endOfTurnConfidenceThreshold": 0.5,
+          "minEndOfTurnSilenceWhenConfident": 500,
+          "maxTurnSilence": 2000
         }
       }
     },
+    {
+      "verb": "config",
+      "referHook": "https://referhook.com"
+    },
+    {
+      "verb": "config",
+      "referHook": {
+        "url": "https://referhook.com"
+      }
+    },
     {
       "verb": "config",
       "recognizer": {
@@ -129,6 +144,7 @@ test("validate correct verbs", async (t) => {
       "callerName": "Tom",
       "answerOnBridge": true,
       "dtmfCapture": ["*2", "*3"],
+      "timeLimit": 10,
       "dtmfHook": {
         "url": "/dtmf",
         "method": "GET"
@@ -194,7 +210,8 @@ test("validate correct verbs", async (t) => {
         "hintsBoost": 10,
         "fastRecognitionTimeout": 2000,
         "deepgramOptions": {
-          "endpointing": 500
+          "endpointing": 500,
+          "noDelay": true,
         }
       },
       "say": {
@@ -209,6 +226,10 @@ test("validate correct verbs", async (t) => {
           "fallbackVoice": "voice"
         }
       },
+      "say": {
+        "text": "To speak to Sales press 1 or say Sales.  To speak to customer support press 2 or say Support",
+        "instructions": "Voice: High-energy, upbeat, and encouraging, projecting enthusiasm and motivation."
+      },
       "say": {
         "text": "To speak to Sales press 1 or say Sales.  To speak to customer support press 2 or say Support",
         "synthesizer": {
@@ -228,6 +249,31 @@ 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",
@@ -337,6 +383,20 @@ test("validate correct verbs", async (t) => {
        }
       }
     },
+    {
+      "verb": "config",
+      "autoStreamTts": true
+    },
+    {
+      "verb": "config",
+      "vad": {
+        "enable": true,
+        "voiceMs": 250,
+        "silenceMs": 150,
+        "strategy": "one-shot",
+        "mode": 2
+      }
+    },
     {
       "verb": "message",
       "to": "15083084809",
@@ -395,6 +455,12 @@ test("validate correct verbs", async (t) => {
       "referTo": "+15083084809",
       "actionHook": "/action"
     },
+    {
+      "verb": "sip:refer",
+      "referTo": "+15083084809",
+      "referredByDisplayName": "Alice",
+      "actionHook": "/action"
+    },
     {
       "verb": "tag",
       "data": {
@@ -412,6 +478,17 @@ 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",
@@ -456,7 +533,28 @@ test("validate correct verbs", async (t) => {
         "actionHook": "/answeringMachineDetection",
     
       },
-      "dual_streams": true
+      "dual_streams": true,
+      "timeLimit" : 10
+    },
+    {
+      "verb": "llm",
+      "vendor": 'ultravox',
+      "model": 'fixie-ai/ultravox',
+      "auth": {
+        "apiKey": "sk-1234567890abcdefg"
+      },
+      "llmOptions": {
+        "firstSpeaker": 'FIRST_SPEAKER_AGENT',
+      },
+      "mcpServers": [
+        {
+          "url": 'https://mcp.example.com',
+          "auth": {
+            "username": 'username',
+            "password": 'password'
+          }
+        }
+      ]
     }
   ];
   try {