add statusHook to redirect

* support houndify stt

* add testcase

* wip

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.97",
+  "version": "0.0.123",
   "lockfileVersion": 2,
   "requires": true,
   "packages": {
     "": {
       "name": "@jambonz/verb-specifications",
-      "version": "0.0.97",
+      "version": "0.0.123",
       "license": "MIT",
       "dependencies": {
         "debug": "^4.3.4",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.97",
+  "version": "0.0.123",
   "description": "Jambonz Verb Specification Utilities",
   "main": "index.js",
   "scripts": {

specs.json

@@ -1,4 +1,13 @@
 {
+  "alert" : {
+    "properties": {
+      "id": "string",
+      "message": "string"
+    },
+    "required": [
+      "message"
+    ]
+  },
   "answer": {
     "properties": {
       "id": "string"
@@ -57,6 +66,7 @@
       "amd": "#amd",
       "fillerNoise": "#fillerNoise",
       "notifyEvents": "boolean",
+      "notifySttLatency": "boolean",
       "reset": "string|array",
       "onHoldMusic": "string",
       "actionHookDelayAction": "#actionHookDelayAction",
@@ -64,7 +74,9 @@
       "boostAudioSignal": "number|string",
       "vad":"#vad",
       "referHook": "object|string",
-      "earlyMedia": "boolean"
+      "earlyMedia": "boolean",
+      "autoStreamTts": "boolean",
+      "disableTtsCache": "boolean"
     },
     "required": []
   },
@@ -213,6 +225,7 @@
     "properties": {
       "id": "string",
       "text": "string|array",
+      "instructions": "string",
       "stream": "boolean",
       "loop": "number|string",
       "synthesizer": "#synthesizer",
@@ -266,7 +279,8 @@
       "statusEvents": "array",
       "statusHook": "object|string",
       "enterHook": "object|string",
-      "record": "#record"
+      "record": "#record",
+      "distributeDtmf": "boolean"
     },
     "required": [
       "name"
@@ -298,7 +312,8 @@
       "transcribe": "#transcribe",
       "amd": "#amd",
       "dub": ["#dub"],
-      "tag": "object"
+      "tag": "object",
+      "forwardPAI": "boolean"
     },
     "required": [
       "target"
@@ -443,6 +458,7 @@
       "model": "string",
       "auth": "object",
       "connectOptions": "object",
+      "mcpServers": ["#mcpServer"],
       "actionHook": "object|string",
       "eventHook": "object|string",
       "toolHook": "object|string",
@@ -451,11 +467,19 @@
     },
     "required": [
       "vendor",
-      "model",
-      "auth",
       "llmOptions"
     ]
   },
+  "mcpServer": {
+    "properties": {
+      "url": "string",
+      "auth": "object",
+      "roots": ["#root"]
+    },
+    "required": [
+      "url"
+    ]
+  },
   "message": {
     "properties": {
       "id": "string",
@@ -510,6 +534,11 @@
         "type": "string",
         "enum": ["startCallRecording", "stopCallRecording", "pauseCallRecording", "resumeCallRecording"]
       },
+      "type" : {
+        "type" : "string",
+        "enum" : ["cloud", "siprec"]
+
+      },
       "recordingID": "string",
       "siprecServerURL": "string|array",
       "headers": "object"
@@ -521,7 +550,8 @@
   "redirect": {
     "properties": {
       "id": "string",
-      "actionHook": "object|string"
+      "actionHook": "object|string",
+      "statusHook": "object|string"
     },
     "required": [
       "actionHook"
@@ -577,7 +607,6 @@
       "channel": "number"
     },
     "required": [
-      "transcriptionHook"
     ]
   },
   "target": {
@@ -733,7 +762,11 @@
       "googleOptions": "#googleOptions",
       "customOptions": "#customOptions",
       "verbioOptions": "#verbioOptions",
-      "speechmaticsOptions": "#speechmaticsOptions"
+      "speechmaticsOptions": "#speechmaticsOptions",
+      "openaiOptions": "#openaiOptions",
+      "houndifyOptions": "#houndifyOptions",
+      "gladiaOptions": "object",
+      "elevenlabsOptions": "#elevenlabsOptions"
     },
     "required": [
       "vendor"
@@ -776,6 +809,7 @@
     "properties": {
       "speechSegmentationSilenceTimeoutMs": "number",
       "postProcessing" : "string",
+      "audioLogging" : "boolean",
       "languageIdMode": {
         "type": "string",
         "enum": [
@@ -811,6 +845,61 @@
       "transcriptNormalization": "array"
     }
   },
+  "houndifyOptions": {
+    "properties": {
+      "requestInfo": "object",
+      "sampleRate": "number",
+      "latitude": "number",
+      "longitude": "number",
+      "city": "string",
+      "state": "string",
+      "country": "string",
+      "timeZone": "string",
+      "domain": "string",
+      "audioEndpoint": "string",
+      "maxSilenceSeconds": "number",
+      "maxSilenceAfterFullQuerySeconds": "number",
+      "maxSilenceAfterPartialQuerySeconds": "number",
+      "vadSensitivity": "number",
+      "vadTimeout": "number",
+      "vadMode": "string",
+      "vadVoiceMs": "number",
+      "vadSilenceMs": "number",
+      "vadDebug": "boolean",
+      "audioFormat": "string",
+      "enableNoiseReduction": "boolean",
+      "enableProfanityFilter": "boolean",
+      "enablePunctuation": "boolean",
+      "enableCapitalization": "boolean",
+      "confidenceThreshold": "number",
+      "enableDisfluencyFilter": "boolean",
+      "maxResults": "number",
+      "enableWordTimestamps": "boolean",
+      "maxAlternatives": "number",
+      "partialTranscriptInterval": "number",
+      "sessionTimeout": "number",
+      "connectionTimeout": "number",
+      "customVocabulary": "array",
+      "languageModel": "string"
+    }
+  },
+  "elevenlabsOptions": {
+    "properties": {
+      "includeTimestamps": "boolean",
+      "commitStrategy": {
+        "type": "string",
+        "enum": [
+          "manual",
+          "vad"
+        ]
+      },
+      "vadSilenceThresholdSecs": "number",
+      "vadThreshold": "number",
+      "minSpeechDurationMs": "number",
+      "minSilenceDurationMs": "number",
+      "enableLogging": "boolean"
+    }
+  },
   "cobaltOptions": {
     "properties": {
       "serverUri": "string",
@@ -890,7 +979,12 @@
       "shortUtterance": "boolean",
       "vadTurnoff": "number",
       "tag": "string",
-      "fillerWords" : "boolean"
+      "fillerWords" : "boolean",
+      "eotThreshold": "number",
+      "eotTimeoutMs": "number",
+      "mipOptOut": "boolean",
+      "entityPrompt": "string",
+      "eagerEotThreshold":"number"
     }
   },
   "sonioxOptions": {
@@ -921,6 +1015,60 @@
     "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",
@@ -1070,7 +1218,18 @@
   },
   "assemblyAiOptions": {
     "properties": {
-      "apiKey": "string"
+      "apiKey": "string",
+      "serviceVersion": {
+        "type": "string",
+        "enum": [
+          "v2",
+          "v3"
+        ]
+      },
+      "formatTurns": "boolean",
+      "endOfTurnConfidenceThreshold": "number",
+      "minEndOfTurnSilenceWhenConfident": "number",
+      "maxTurnSilence": "number"
     }
   },
   "resource": {
@@ -1151,16 +1310,25 @@
       "voiceMs": "number",
       "silenceMs": "number",
       "strategy": "string",
-      "mode": "number"
+      "mode": "number",
+      "vendor": {
+        "type": "string",
+        "enum": [
+          "webrtc",
+          "silero"
+        ]
+      },
+      "threshold": "number",
+      "speechPadMs": "number"
     },
     "required": [
-      "enable"
     ]
   },
   "amd": {
     "properties": {
       "actionHook": "object|string",
       "thresholdWordCount": "number",
+      "digitCount": "number",
       "timers": "#amdTimers",
       "recognizer": "#recognizer"
     },
@@ -1192,5 +1360,36 @@
       "streaming": "boolean",
       "sampleRate": "number"
     }
+  },
+  "pipeline": {
+    "properties": {
+      "id": "string",
+      "stt": "#recognizer",
+      "tts": "#synthesizer",
+      "vad": "#vad",
+      "turnDetection": "#turnDetectionPipeline",
+      "llm": "#llm",
+      "preflightLlm": "boolean",
+      "actionHook": "object|string",
+      "eventHook": "object|string"
+    },
+    "required": [
+      "stt",
+      "llm",
+      "tts"
+    ]
+  },
+  "turnDetectionPipeline": {
+    "properties": {
+      "vendor": {
+        "type": "string",
+        "enum": ["krisp"]
+      },
+      "threshold": "number",
+      "eagerEotThreshold": "number"
+    },
+    "required": [
+      "vendor"
+    ]
   }
 }

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
         }
       }
     },
@@ -207,6 +213,10 @@ test("validate correct verbs", async (t) => {
         "deepgramOptions": {
           "endpointing": 500,
           "noDelay": true,
+          "eotThreshold": 500,
+          "eotTimeoutMs": 5000,
+          "eagerEotThreshold": 200,
+          "mipOptOut": true
         }
       },
       "say": {
@@ -221,6 +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": {
@@ -265,6 +279,69 @@ 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": "houndify",
+        "language": "en-US",
+        "houndifyOptions": {
+          "requestInfo": {
+            "Latitude": 30.6,
+            "Longitude": 30.6,
+            "City": "Huston"
+          },
+          "sampleRate": 16000,
+          "latitude": 30.6,
+          "longitude": 30.6,
+          "city": "Huston",
+          "state": "CA",
+          "country": "US",
+          "timeZone": "GMT",
+          "domain": "Voice",
+          "audioEndpoint": "https://api.houndify.com/v1/audio",
+          "maxSilenceSeconds": 5,
+          "maxSilenceAfterFullQuerySeconds": 3,
+          "maxSilenceAfterPartialQuerySeconds": 2,
+          "vadSensitivity": 0.5,
+          "vadTimeout": 1000,
+          "vadMode": "auto",
+          "vadVoiceMs": 250,
+          "vadSilenceMs": 500,
+          "vadDebug": true,
+          "audioFormat": "PCM16",
+          "enableNoiseReduction": true,
+          "enableProfanityFilter": false,
+          "enablePunctuation": true,
+          "enableCapitalization": true,
+          "confidenceThreshold": 0.7,
+          "enableDisfluencyFilter": true,
+          "maxResults": 5,
+          "enableWordTimestamps": true,
+          "maxAlternatives": 3,
+          "partialTranscriptInterval": 100,
+          "sessionTimeout": 30000,
+          "connectionTimeout": 5000,
+          "customVocabulary": ["jambonz", "telephony", "voip"],
+          "languageModel": "enhanced"
+        },
+        "gladiaOptions": {
+          "post_processing": {
+            "summarization": false,
+            "summarization_config": {
+              "type": "general"
+            },
+            "chapterization": false
+          },
+        }
+      },
+    },
     {
       "verb": "gather",
       "actionHook": "http://example.com/collect",
@@ -374,6 +451,10 @@ test("validate correct verbs", async (t) => {
        }
       }
     },
+    {
+      "verb": "config",
+      "autoStreamTts": true
+    },
     {
       "verb": "config",
       "vad": {
@@ -381,7 +462,10 @@ test("validate correct verbs", async (t) => {
         "voiceMs": 250,
         "silenceMs": 150,
         "strategy": "one-shot",
-        "mode": 2
+        "mode": 2,
+        "vendor": "webrtc",
+        "threshold": 0.5,
+        "speechPadMs": 1000
       }
     },
     {
@@ -465,6 +549,24 @@ test("validate correct verbs", async (t) => {
         "interim": true
       }
     },
+    {
+      "verb": "transcribe",
+      "transcriptionHook": "http://example.com/transcribe",
+      "recognizer": {
+        "vendor": "nvidia",
+        "language" : "en-US",
+        "interim": true,
+        "elevenlabsOptions": {
+          "includeTimestamps": true,
+          "commitStrategy": "vad",
+          "vadSilenceThresholdSecs": 0.8,
+          "vadThreshold": 0.5,
+          "minSpeechDurationMs": 300,
+          "minSilenceDurationMs": 500,
+          "enableLogging": false
+        }
+      }
+    },
     {
       "verb": "transcribe",
       "transcriptionHook": "http://example.com/transcribe",
@@ -522,6 +624,26 @@ test("validate correct verbs", async (t) => {
       },
       "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 {