0.0.111

* support enable stt latency calculation

* fixed review comment

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.86",
+  "version": "0.0.111",
   "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 +1,13 @@
 {
+  "alert" : {
+    "properties": {
+      "id": "string",
+      "message": "string"
+    },
+    "required": [
+      "message"
+    ]
+  },
   "answer": {
     "properties": {
       "id": "string"
@@ -34,6 +43,7 @@
       "id": "string",
       "referTo": "string",
       "referredBy": "string",
+      "referredByDisplayName": "string",
       "headers": "object",
       "actionHook": "object|string",
       "eventHook": "object|string"
@@ -48,19 +58,24 @@
       "synthesizer": "#synthesizer",
       "recognizer": "#recognizer",
       "bargeIn": "#bargeIn",
+      "ttsStream": "#ttsStream",
       "record": "#recordOptions",
       "listen": "#listenOptions",
+      "stream": "#listenOptions",
       "transcribe": "#transcribeOptions",
       "amd": "#amd",
       "fillerNoise": "#fillerNoise",
       "notifyEvents": "boolean",
+      "notifySttLatency": "boolean",
       "reset": "string|array",
       "onHoldMusic": "string",
       "actionHookDelayAction": "#actionHookDelayAction",
       "sipRequestWithinDialogHook": "object|string",
       "boostAudioSignal": "number|string",
       "vad":"#vad",
-      "referHook": "object|string"
+      "referHook": "object|string",
+      "earlyMedia": "boolean",
+      "autoStreamTts": "boolean"
     },
     "required": []
   },
@@ -96,6 +111,15 @@
       "enable"
     ]
   },
+  "ttsStream": {
+    "properties": {
+      "enable": "boolean",
+      "synthesizer": "#synthesizer"
+    },
+    "required": [
+      "enable"
+    ]
+  },
   "bargeIn": {
     "properties": {
       "enable": "boolean",
@@ -200,13 +224,15 @@
     "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": {
@@ -276,6 +302,7 @@
       "exitMediaPath": "boolean",
       "boostAudioSignal": "number|string",
       "listen": "#listen",
+      "stream": "#listen",
       "target": ["#target"],
       "timeLimit": "number",
       "timeout": "number",
@@ -283,7 +310,8 @@
       "transcribe": "#transcribe",
       "amd": "#amd",
       "dub": ["#dub"],
-      "tag": "object"
+      "tag": "object",
+      "forwardPAI": "boolean"
     },
     "required": [
       "target"
@@ -367,6 +395,34 @@
     ]
   },
   "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",
@@ -400,6 +456,7 @@
       "model": "string",
       "auth": "object",
       "connectOptions": "object",
+      "mcpServers": ["#mcpServer"],
       "actionHook": "object|string",
       "eventHook": "object|string",
       "toolHook": "object|string",
@@ -408,11 +465,20 @@
     },
     "required": [
       "vendor",
-      "model",
       "auth",
       "llmOptions"
     ]
   },
+  "mcpServer": {
+    "properties": {
+      "url": "string",
+      "auth": "object",
+      "roots": ["#root"]
+    },
+    "required": [
+      "url"
+    ]
+  },
   "message": {
     "properties": {
       "id": "string",
@@ -506,7 +572,8 @@
       "amd": "#amd",
       "dual_streams": "boolean",
       "sipRequestWithinDialogHook": "string",
-      "referHook": "object|string"
+      "referHook": "object|string",
+      "timeLimit": "number"
     },
     "required": [
       "call_hook",
@@ -689,7 +756,8 @@
       "googleOptions": "#googleOptions",
       "customOptions": "#customOptions",
       "verbioOptions": "#verbioOptions",
-      "speechmaticsOptions": "#speechmaticsOptions"
+      "speechmaticsOptions": "#speechmaticsOptions",
+      "openaiOptions": "#openaiOptions"
     },
     "required": [
       "vendor"
@@ -738,6 +806,14 @@
           "AtStart",
           "Continuous"
         ]
+      },
+      "speechRecognitionMode": {
+        "type": "string",
+        "enum": [
+          "CONVERSATION",
+          "DICTATION",
+          "INTERACTIVE"
+        ]
       }
     },
     "required": [
@@ -812,6 +888,7 @@
       "version": "string",
       "punctuate": "boolean",
       "smartFormatting": "boolean",
+      "noDelay": "boolean",
       "profanityFilter": "boolean",
       "redact": {
         "type": "string",
@@ -831,11 +908,17 @@
       "search": "array",
       "replace": "array",
       "keywords": "array",
+      "keyterms": "array",
       "endpointing": "boolean | number",
       "utteranceEndMs": "number",
       "shortUtterance": "boolean",
       "vadTurnoff": "number",
-      "tag": "string"
+      "tag": "string",
+      "fillerWords" : "boolean",
+      "preflightThreshold": "number",
+      "eotThreshold": "number",
+      "eotTimeoutMs": "number",
+      "mipOptOut": "boolean"
     }
   },
   "sonioxOptions": {
@@ -866,6 +949,60 @@
     "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",
@@ -878,6 +1015,7 @@
   "sm_transcriptionConfig": {
     "properties": {
       "language": "string",
+      "domain": "string",
       "additional_vocab": "array",
       "diarization": "string",
       "speaker_diarization_config": "#sm_speakerDiarizationConfig",
@@ -1014,7 +1152,18 @@
   },
   "assemblyAiOptions": {
     "properties": {
-      "apiKey": "string"
+      "apiKey": "string",
+      "serviceVersion": {
+        "type": "string",
+        "enum": [
+          "v2",
+          "v3"
+        ]
+      },
+      "formatTurns": "boolean",
+      "endOfTurnConfidenceThreshold": "number",
+      "minEndOfTurnSilenceWhenConfident": "number",
+      "maxTurnSilence": "number"
     }
   },
   "resource": {
@@ -1095,7 +1244,16 @@
       "voiceMs": "number",
       "silenceMs": "number",
       "strategy": "string",
-      "mode": "number"
+      "mode": "number",
+      "vendor": {
+        "type": "string",
+        "enum": [
+          "webrtc",
+          "silero"
+        ]
+      },
+      "threshold": "number",
+      "speechPadMs": "number"
     },
     "required": [
       "enable"
@@ -1105,6 +1263,7 @@
     "properties": {
       "actionHook": "object|string",
       "thresholdWordCount": "number",
+      "digitCount": "number",
       "timers": "#amdTimers",
       "recognizer": "#recognizer"
     },

test/jambonz-verb-test.js

@@ -72,12 +72,18 @@ test("validate correct verbs", async (t) => {
     },
     {
       "verb": "config",
+      "notifySttLatency": true,
       "recognizer": {
         "vendor": "google",
         "language": "de-DE",
         "label": "label1",
         "assemblyAiOptions": {
-          "apiKey": "apikey"
+          "apiKey": "apikey",
+          "serviceVersion": "v3",
+          "formatTurns": true,
+          "endOfTurnConfidenceThreshold": 0.5,
+          "minEndOfTurnSilenceWhenConfident": 500,
+          "maxTurnSilence": 2000
         }
       }
     },
@@ -139,6 +145,7 @@ test("validate correct verbs", async (t) => {
       "callerName": "Tom",
       "answerOnBridge": true,
       "dtmfCapture": ["*2", "*3"],
+      "timeLimit": 10,
       "dtmfHook": {
         "url": "/dtmf",
         "method": "GET"
@@ -204,7 +211,12 @@ test("validate correct verbs", async (t) => {
         "hintsBoost": 10,
         "fastRecognitionTimeout": 2000,
         "deepgramOptions": {
-          "endpointing": 500
+          "endpointing": 500,
+          "noDelay": true,
+          preflightThreshold: 1000,
+          eotThreshold: 500,
+          eotTimeoutMs: 5000,
+          mipOptOut: true
         }
       },
       "say": {
@@ -219,6 +231,10 @@ 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": {
@@ -372,6 +388,10 @@ test("validate correct verbs", async (t) => {
        }
       }
     },
+    {
+      "verb": "config",
+      "autoStreamTts": true
+    },
     {
       "verb": "config",
       "vad": {
@@ -379,7 +399,10 @@ test("validate correct verbs", async (t) => {
         "voiceMs": 250,
         "silenceMs": 150,
         "strategy": "one-shot",
-        "mode": 2
+        "mode": 2,
+        "vendor": "webrtc",
+        "threshold": 0.5,
+        "speechPadMs": 1000
       }
     },
     {
@@ -440,6 +463,12 @@ test("validate correct verbs", async (t) => {
       "referTo": "+15083084809",
       "actionHook": "/action"
     },
+    {
+      "verb": "sip:refer",
+      "referTo": "+15083084809",
+      "referredByDisplayName": "Alice",
+      "actionHook": "/action"
+    },
     {
       "verb": "tag",
       "data": {
@@ -512,7 +541,28 @@ 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 {