support deepgram use single connection

2026-03-26 13:59:02 +00:00 · 2024-04-26 15:01:21 +07:00
4 changed files with 733 additions and 2323 deletions
--- a/package-lock.json
+++ b/package-lock.json
--- a/package.json
+++ b/package.json
@@ -1,6 +1,6 @@
 {
  "name": "@jambonz/verb-specifications",
-  "version": "0.0.113",
+  "version": "0.0.70",
  "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": "^5.7.5"
+    "tape": "^5.6.1"
  },
  "dependencies": {
    "debug": "^4.3.4",
--- a/specs.json
+++ b/specs.json
@@ -1,13 +1,4 @@
 {
-  "alert" : {
-    "properties": {
-      "id": "string",
-      "message": "string"
-    },
-    "required": [
-      "message"
-    ]
-  },
  "answer": {
    "properties": {
      "id": "string"
@@ -43,7 +34,6 @@
      "id": "string",
      "referTo": "string",
      "referredBy": "string",
-      "referredByDisplayName": "string",
      "headers": "object",
      "actionHook": "object|string",
      "eventHook": "object|string"
@@ -58,24 +48,17 @@
      "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",
-      "earlyMedia": "boolean",
-      "autoStreamTts": "boolean"
+      "boostAudioSignal": "number|string"
    },
    "required": []
  },
@@ -111,15 +94,6 @@
      "enable"
    ]
  },
-  "ttsStream": {
-    "properties": {
-      "enable": "boolean",
-      "synthesizer": "#synthesizer"
-    },
-    "required": [
-      "enable"
-    ]
-  },
  "bargeIn": {
    "properties": {
      "enable": "boolean",
@@ -224,15 +198,13 @@
    "properties": {
      "id": "string",
      "text": "string|array",
-      "instructions": "string",
-      "stream": "boolean",
      "loop": "number|string",
      "synthesizer": "#synthesizer",
      "earlyMedia": "boolean",
-      "disableTtsCache": "boolean",
-      "closeStreamOnEmpty": "boolean"
+      "disableTtsCache": "boolean"
    },
    "required": [
+      "text"
    ]
  },
  "gather": {
@@ -299,10 +271,8 @@
      "dtmfHook": "object|string",
      "headers": "object",
      "anchorMedia": "boolean",
-      "exitMediaPath": "boolean",
      "boostAudioSignal": "number|string",
      "listen": "#listen",
-      "stream": "#listen",
      "target": ["#target"],
      "timeLimit": "number",
      "timeout": "number",
@@ -310,8 +280,7 @@
      "transcribe": "#transcribe",
      "amd": "#amd",
      "dub": ["#dub"],
-      "tag": "object",
-      "forwardPAI": "boolean"
+      "tag": "object"
    },
    "required": [
      "target"
@@ -322,17 +291,15 @@
      "id": "string",
      "credentials": "object|string",
      "project": "string",
-      "agent": "string",
      "environment": "string",
-      "region": "string",
-      "model": {
+      "region": {
        "type": "string",
-        "enum": ["es", "cx"]
+        "enum": ["europe-west1", "europe-west2", "australia-southeast1", "asia-northeast1"]
      },
      "lang": "string",
      "actionHook": "object|string",
      "eventHook": "object|string",
-      "events": "array",
+      "events": "[string]",
      "welcomeEvent": "string",
      "welcomeEventParams": "object",
      "noInputTimeout": "number",
@@ -340,8 +307,7 @@
      "passDtmfAsTextInput": "boolean",
      "thinkingMusic": "string",
      "tts": "#synthesizer",
-      "bargein": "boolean",
-      "queryInput": "#queryInput"
+      "bargein": "boolean"
    },
    "required": [
      "project",
@@ -349,16 +315,6 @@
      "lang"
    ]
  },
-  "queryInput": {
-    "properties": {
-      "text": "string",
-      "intent": "string",
-      "event": "string",
-      "dtmf": "string"
-    },
-    "required": [
-    ]
-  },
  "dtmf": {
    "properties": {
      "id": "string",
@@ -395,34 +351,6 @@
    ]
  },
  "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",
@@ -449,35 +377,6 @@
      "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",
-      "llmOptions"
-    ]
-  },
-  "mcpServer": {
-    "properties": {
-      "url": "string",
-      "auth": "object",
-      "roots": ["#root"]
-    },
-    "required": [
-      "url"
-    ]
-  },
  "message": {
    "properties": {
      "id": "string",
@@ -570,9 +469,7 @@
      "timeout": "number",
      "amd": "#amd",
      "dual_streams": "boolean",
-      "sipRequestWithinDialogHook": "string",
-      "referHook": "object|string",
-      "timeLimit": "number"
+      "sipRequestWithinDialogHook": "string"
    },
    "required": [
      "call_hook",
@@ -593,12 +490,12 @@
    "properties": {
      "id": "string",
      "transcriptionHook": "string",
-      "translationHook": "string",
      "recognizer": "#recognizer",
      "earlyMedia": "boolean",
      "channel": "number"
    },
    "required": [
+      "transcriptionHook"
    ]
  },
  "target": {
@@ -658,7 +555,7 @@
      "fallbackVoice": "string|object",
      "engine": {
        "type": "string",
-        "enum": ["standard", "neural", "generative", "long-form"]
+        "enum": ["standard", "neural"]
      },
      "gender": {
        "type": "string",
@@ -741,7 +638,6 @@
      "asrDtmfTerminationDigit": "string",
      "asrTimeout": "number",
      "fastRecognitionTimeout": "number",
-      "minConfidence": "number",
      "nuanceOptions": "#nuanceOptions",
      "deepgramOptions": "#deepgramOptions",
      "ibmOptions": "#ibmOptions",
@@ -752,10 +648,7 @@
      "azureOptions": "#azureOptions",
      "assemblyAiOptions": "#assemblyAiOptions",
      "googleOptions": "#googleOptions",
-      "customOptions": "#customOptions",
-      "verbioOptions": "#verbioOptions",
-      "speechmaticsOptions": "#speechmaticsOptions",
-      "openaiOptions": "#openaiOptions"
+      "customOptions": "#customOptions"
    },
    "required": [
      "vendor"
@@ -765,7 +658,6 @@
    "properties": {
      "authToken": "string",
      "uri": "string",
-      "sampleRate": "number",
      "options": "object"
    },
    "required": [
@@ -797,21 +689,12 @@
  "azureOptions": {
    "properties": {
      "speechSegmentationSilenceTimeoutMs": "number",
-      "postProcessing" : "string",
      "languageIdMode": {
        "type": "string",
        "enum": [
          "AtStart",
          "Continuous"
        ]
-      },
-      "speechRecognitionMode": {
-        "type": "string",
-        "enum": [
-          "CONVERSATION",
-          "DICTATION",
-          "INTERACTIVE"
-        ]
      }
    },
    "required": [
@@ -886,7 +769,6 @@
      "version": "string",
      "punctuate": "boolean",
      "smartFormatting": "boolean",
-      "noDelay": "boolean",
      "profanityFilter": "boolean",
      "redact": {
        "type": "string",
@@ -906,18 +788,12 @@
      "search": "array",
      "replace": "array",
      "keywords": "array",
-      "keyterms": "array",
      "endpointing": "boolean | number",
      "utteranceEndMs": "number",
      "shortUtterance": "boolean",
      "vadTurnoff": "number",
-      "tag": "string",
-      "fillerWords" : "boolean",
-      "preflightThreshold": "number",
-      "eotThreshold": "number",
-      "eotTimeoutMs": "number",
-      "mipOptOut": "boolean",
-      "entityPrompt": "string"
+      "useSingleConnection": "boolean",
+      "tag": "string"
    }
  },
  "sonioxOptions": {
@@ -933,165 +809,6 @@
    "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",
@@ -1151,18 +868,7 @@
  },
  "assemblyAiOptions": {
    "properties": {
-      "apiKey": "string",
-      "serviceVersion": {
-        "type": "string",
-        "enum": [
-          "v2",
-          "v3"
-        ]
-      },
-      "formatTurns": "boolean",
-      "endOfTurnConfidenceThreshold": "number",
-      "minEndOfTurnSilenceWhenConfident": "number",
-      "maxTurnSilence": "number"
+      "apiKey": "string"
    }
  },
  "resource": {
@@ -1241,27 +947,16 @@
    "properties": {
      "enable": "boolean",
      "voiceMs": "number",
-      "silenceMs": "number",
-      "strategy": "string",
-      "mode": "number",
-      "vendor": {
-        "type": "string",
-        "enum": [
-          "webrtc",
-          "silero"
-        ]
-      },
-      "threshold": "number",
-      "speechPadMs": "number"
+      "mode": "number"
    },
    "required": [
+      "enable"
    ]
  },
  "amd": {
    "properties": {
      "actionHook": "object|string",
      "thresholdWordCount": "number",
-      "digitCount": "number",
      "timers": "#amdTimers",
      "recognizer": "#recognizer"
    },
@@ -1283,8 +978,7 @@
      "noResponseTimeout": "number",
      "noResponseGiveUpTimeout": "number",
      "retries": "number",
-      "actions": "array",
-      "giveUpActions": "array"
+      "actions": "array"
    }
  },
  "bidirectionalAudio" : {
@@ -1293,36 +987,5 @@
      "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",
-      "preflightThreshold": "number"
-    },
-    "required": [
-      "vendor"
-    ]
  }
 }
--- a/test/jambonz-verb-test.js
+++ b/test/jambonz-verb-test.js
@@ -72,31 +72,15 @@ test("validate correct verbs", async (t) => {
    },
    {
      "verb": "config",
-      "notifySttLatency": true,
      "recognizer": {
        "vendor": "google",
        "language": "de-DE",
        "label": "label1",
        "assemblyAiOptions": {
-          "apiKey": "apikey",
-          "serviceVersion": "v3",
-          "formatTurns": true,
-          "endOfTurnConfidenceThreshold": 0.5,
-          "minEndOfTurnSilenceWhenConfident": 500,
-          "maxTurnSilence": 2000
+          "apiKey": "apikey"
        }
      }
    },
-    {
-      "verb": "config",
-      "referHook": "https://referhook.com"
-    },
-    {
-      "verb": "config",
-      "referHook": {
-        "url": "https://referhook.com"
-      }
-    },
    {
      "verb": "config",
      "recognizer": {
@@ -145,7 +129,6 @@ test("validate correct verbs", async (t) => {
      "callerName": "Tom",
      "answerOnBridge": true,
      "dtmfCapture": ["*2", "*3"],
-      "timeLimit": 10,
      "dtmfHook": {
        "url": "/dtmf",
        "method": "GET"
@@ -212,11 +195,7 @@ test("validate correct verbs", async (t) => {
        "fastRecognitionTimeout": 2000,
        "deepgramOptions": {
          "endpointing": 500,
-          "noDelay": true,
-          preflightThreshold: 1000,
-          eotThreshold: 500,
-          eotTimeoutMs: 5000,
-          mipOptOut: true
+          "useSingleConnection": true
        }
      },
      "say": {
@@ -231,10 +210,6 @@ 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": {
@@ -254,31 +229,6 @@ 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",
@@ -388,23 +338,6 @@ test("validate correct verbs", async (t) => {
       }
      }
    },
-    {
-      "verb": "config",
-      "autoStreamTts": true
-    },
-    {
-      "verb": "config",
-      "vad": {
-        "enable": true,
-        "voiceMs": 250,
-        "silenceMs": 150,
-        "strategy": "one-shot",
-        "mode": 2,
-        "vendor": "webrtc",
-        "threshold": 0.5,
-        "speechPadMs": 1000
-      }
-    },
    {
      "verb": "message",
      "to": "15083084809",
@@ -463,12 +396,6 @@ test("validate correct verbs", async (t) => {
      "referTo": "+15083084809",
      "actionHook": "/action"
    },
-    {
-      "verb": "sip:refer",
-      "referTo": "+15083084809",
-      "referredByDisplayName": "Alice",
-      "actionHook": "/action"
-    },
    {
      "verb": "tag",
      "data": {
@@ -486,17 +413,6 @@ 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",
@@ -541,28 +457,7 @@ test("validate correct verbs", async (t) => {
        "actionHook": "/answeringMachineDetection",
    
      },
-      "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'
-          }
-        }
-      ]
+      "dual_streams": true
    }
  ];
  try {