support dial translate with s2s services

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.115",
+  "version": "0.0.95",
   "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",

specs.json

@@ -1,13 +1,4 @@
 {
-  "alert" : {
-    "properties": {
-      "id": "string",
-      "message": "string"
-    },
-    "required": [
-      "message"
-    ]
-  },
   "answer": {
     "properties": {
       "id": "string"
@@ -66,7 +57,6 @@
       "amd": "#amd",
       "fillerNoise": "#fillerNoise",
       "notifyEvents": "boolean",
-      "notifySttLatency": "boolean",
       "reset": "string|array",
       "onHoldMusic": "string",
       "actionHookDelayAction": "#actionHookDelayAction",
@@ -74,8 +64,7 @@
       "boostAudioSignal": "number|string",
       "vad":"#vad",
       "referHook": "object|string",
-      "earlyMedia": "boolean",
-      "autoStreamTts": "boolean"
+      "earlyMedia": "boolean"
     },
     "required": []
   },
@@ -224,7 +213,6 @@
     "properties": {
       "id": "string",
       "text": "string|array",
-      "instructions": "string",
       "stream": "boolean",
       "loop": "number|string",
       "synthesizer": "#synthesizer",
@@ -311,7 +299,7 @@
       "amd": "#amd",
       "dub": ["#dub"],
       "tag": "object",
-      "forwardPAI": "boolean"
+      "translate": "#translate"
     },
     "required": [
       "target"
@@ -456,7 +444,6 @@
       "model": "string",
       "auth": "object",
       "connectOptions": "object",
-      "mcpServers": ["#mcpServer"],
       "actionHook": "object|string",
       "eventHook": "object|string",
       "toolHook": "object|string",
@@ -465,19 +452,11 @@
     },
     "required": [
       "vendor",
+      "model",
+      "auth",
       "llmOptions"
     ]
   },
-  "mcpServer": {
-    "properties": {
-      "url": "string",
-      "auth": "object",
-      "roots": ["#root"]
-    },
-    "required": [
-      "url"
-    ]
-  },
   "message": {
     "properties": {
       "id": "string",
@@ -599,6 +578,7 @@
       "channel": "number"
     },
     "required": [
+      "transcriptionHook"
     ]
   },
   "target": {
@@ -754,8 +734,7 @@
       "googleOptions": "#googleOptions",
       "customOptions": "#customOptions",
       "verbioOptions": "#verbioOptions",
-      "speechmaticsOptions": "#speechmaticsOptions",
-      "openaiOptions": "#openaiOptions"
+      "speechmaticsOptions": "#speechmaticsOptions"
     },
     "required": [
       "vendor"
@@ -886,7 +865,6 @@
       "version": "string",
       "punctuate": "boolean",
       "smartFormatting": "boolean",
-      "noDelay": "boolean",
       "profanityFilter": "boolean",
       "redact": {
         "type": "string",
@@ -906,19 +884,12 @@
       "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",
-      "eagerEotThreshold":"number"
+      "fillerWords" : "boolean"
     }
   },
   "sonioxOptions": {
@@ -949,60 +920,6 @@
     "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",
@@ -1152,18 +1069,7 @@
   },
   "assemblyAiOptions": {
     "properties": {
-      "apiKey": "string",
-      "serviceVersion": {
-        "type": "string",
-        "enum": [
-          "v2",
-          "v3"
-        ]
-      },
-      "formatTurns": "boolean",
-      "endOfTurnConfidenceThreshold": "number",
-      "minEndOfTurnSilenceWhenConfident": "number",
-      "maxTurnSilence": "number"
+      "apiKey": "string"
     }
   },
   "resource": {
@@ -1244,25 +1150,16 @@
       "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"
     },
@@ -1295,35 +1192,18 @@
       "sampleRate": "number"
     }
   },
-  "pipeline": {
+  "translate": {
     "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"
-    ]
+      "callerLanguage": "string",
+      "calleeLanguage": "string",
+      "gain": "#gain",
+      "llm": "#llm"
+    }
   },
-  "turnDetectionPipeline": {
+  "gain": {
     "properties": {
-      "vendor": {
-        "type": "string",
-        "enum": ["krisp"]
-      },
-      "threshold": "number",
-      "preflightThreshold": "number"
-    },
-    "required": [
-      "vendor"
-    ]
+      "naturalVoice": "string",
+      "translatedVoice": "string"
+    }
   }
 }

test/jambonz-verb-test.js

@@ -72,18 +72,12 @@ 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"
         }
       }
     },
@@ -172,7 +166,23 @@ test("validate correct verbs", async (t) => {
           "type": "user",
           "name": "spike@sip.example.com"
         }
-      ]
+      ],
+      "translate": {
+        "callerLanguage": "en-US",
+        "calleeLanguage": "vn-VN",
+        "gain": {
+            "naturalVoice": "-10db",
+            "translatedVoice": "+2dB"
+        },
+        "llm": {
+          "model": 'openai',
+          "vendor": "openai",
+          "auth": {
+            "apiKey": "API key"
+          },
+          "llmOptions": {}
+        }
+      }
     },
     {
       "verb": "dialogflow",
@@ -211,13 +221,7 @@ test("validate correct verbs", async (t) => {
         "hintsBoost": 10,
         "fastRecognitionTimeout": 2000,
         "deepgramOptions": {
-          "endpointing": 500,
-          "noDelay": true,
-          preflightThreshold: 1000,
-          eotThreshold: 500,
-          eotTimeoutMs: 5000,
-          eagerEotThreshold: 200,
-          mipOptOut: true
+          "endpointing": 500
         }
       },
       "say": {
@@ -232,10 +236,6 @@ 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": {
@@ -389,10 +389,6 @@ test("validate correct verbs", async (t) => {
        }
       }
     },
-    {
-      "verb": "config",
-      "autoStreamTts": true
-    },
     {
       "verb": "config",
       "vad": {
@@ -400,10 +396,7 @@ test("validate correct verbs", async (t) => {
         "voiceMs": 250,
         "silenceMs": 150,
         "strategy": "one-shot",
-        "mode": 2,
-        "vendor": "webrtc",
-        "threshold": 0.5,
-        "speechPadMs": 1000
+        "mode": 2
       }
     },
     {
@@ -544,26 +537,6 @@ 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 {