wip

* support noise isolation to config verb

* wip

* wip

* wip

* wip

* add vendor to turnTaking

jambonz-app-json-validation.js

@@ -4,6 +4,27 @@ const _specData = require('./specs');
 const specs = new Map();
 for (const key in _specData) { specs.set(key, _specData[key]); }
 
+/* verb synonyms and shortcuts: maps alias verb names to their canonical form,
+   optionally injecting properties (e.g. vendor) into the verb data */
+const verbTransforms = new Map([
+  ['stream', {verb: 'listen'}],
+  ['s2s', {verb: 'llm'}],
+  ['openai_s2s', {verb: 'llm', properties: {vendor: 'openai'}}],
+  ['microsoft_s2s', {verb: 'llm', properties: {vendor: 'microsoft'}}],
+  ['google_s2s', {verb: 'llm', properties: {vendor: 'google'}}],
+  ['elevenlabs_s2s', {verb: 'llm', properties: {vendor: 'elevenlabs'}}],
+  ['deepgram_s2s', {verb: 'llm', properties: {vendor: 'deepgram'}}],
+  ['voiceagent_s2s', {verb: 'llm', properties: {vendor: 'voiceagent'}}],
+  ['ultravox_s2s', {verb: 'llm', properties: {vendor: 'ultravox'}}],
+]);
+
+function applyVerbTransform(name, data) {
+  const transform = verbTransforms.get(name);
+  if (!transform) return {name, data};
+  const newData = transform.properties ? {...transform.properties, ...data} : data;
+  return {name: transform.verb, data: newData};
+}
+
 function normalizeJambones(logger, obj) {
   if (!Array.isArray(obj)) {
     throw new Error('malformed jambonz payload: must be array');
@@ -13,18 +34,22 @@ function normalizeJambones(logger, obj) {
     if (typeof tdata !== 'object') throw new Error('malformed jambonz payload: must be array of objects');
     if ('verb' in tdata) {
       // {verb: 'say', text: 'foo..bar'..}
-      const name = tdata.verb;
       const o = {};
       Object.keys(tdata)
         .filter((k) => k !== 'verb')
         .forEach((k) => o[k] = tdata[k]);
+      const {name, data} = applyVerbTransform(tdata.verb, o);
       const o2 = {};
-      o2[name] = o;
+      o2[name] = data;
       document.push(o2);
     }
     else if (Object.keys(tdata).length === 1) {
       // {'say': {..}}
-      document.push(tdata);
+      const key = Object.keys(tdata)[0];
+      const {name, data} = applyVerbTransform(key, tdata[key]);
+      const o2 = {};
+      o2[name] = data;
+      document.push(o2);
     }
     else {
       logger.info(tdata, 'malformed jambonz payload: missing verb property');

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.122",
+  "version": "0.1.2",
   "lockfileVersion": 2,
   "requires": true,
   "packages": {
     "": {
       "name": "@jambonz/verb-specifications",
-      "version": "0.0.122",
+      "version": "0.1.2",
       "license": "MIT",
       "dependencies": {
         "debug": "^4.3.4",

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@jambonz/verb-specifications",
-  "version": "0.0.122",
+  "version": "0.1.2",
   "description": "Jambonz Verb Specification Utilities",
   "main": "index.js",
   "scripts": {

specs.json

@@ -76,7 +76,9 @@
       "referHook": "object|string",
       "earlyMedia": "boolean",
       "autoStreamTts": "boolean",
-      "disableTtsCache": "boolean"
+      "disableTtsCache": "boolean",
+      "noiseIsolation": "#noiseIsolation",
+      "turnTaking": "#turnTaking"
     },
     "required": []
   },
@@ -280,6 +282,7 @@
       "statusHook": "object|string",
       "enterHook": "object|string",
       "record": "#record",
+      "listen": "#listen",
       "distributeDtmf": "boolean"
     },
     "required": [
@@ -329,7 +332,7 @@
       "region": "string",
       "model": {
         "type": "string",
-        "enum": ["es", "cx"]
+        "enum": ["es", "cx", "ces"]
       },
       "lang": "string",
       "actionHook": "object|string",
@@ -424,33 +427,6 @@
       "url"
     ]
   },
-  "stream": {
-    "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"
-    },
-    "required": [
-      "url"
-    ]
-  },
   "llm": {
     "properties": {
       "id": "string",
@@ -534,6 +510,11 @@
         "type": "string",
         "enum": ["startCallRecording", "stopCallRecording", "pauseCallRecording", "resumeCallRecording"]
       },
+      "type" : {
+        "type" : "string",
+        "enum" : ["cloud", "siprec"]
+
+      },
       "recordingID": "string",
       "siprecServerURL": "string|array",
       "headers": "object"
@@ -545,7 +526,8 @@
   "redirect": {
     "properties": {
       "id": "string",
-      "actionHook": "object|string"
+      "actionHook": "object|string",
+      "statusHook": "object|string"
     },
     "required": [
       "actionHook"
@@ -681,6 +663,7 @@
       "fallbackLabel": "string",
       "fallbackLanguage": "string",
       "vad": "#vad",
+      "autogeneratePrompt": "boolean",
       "hints": "array",
       "hintsBoost": "number",
       "altLanguages": "array",
@@ -874,7 +857,8 @@
       "sessionTimeout": "number",
       "connectionTimeout": "number",
       "customVocabulary": "array",
-      "languageModel": "string"
+      "languageModel": "string",
+      "audioQueryAbsoluteTimeout": "number"
     }
   },
   "elevenlabsOptions": {
@@ -1079,6 +1063,7 @@
       "additional_vocab": "array",
       "diarization": "string",
       "speaker_diarization_config": "#sm_speakerDiarizationConfig",
+      "conversation_config": "#sm_conversationConfig",
       "enable_partials": "boolean",
       "max_delay": "number",
       "max_delay_mode": {
@@ -1106,6 +1091,13 @@
     "required": [
     ]
   },
+  "sm_conversationConfig": {
+    "properties": {
+      "end_of_utterance_silence_trigger": "number"
+    },
+    "required": [
+    ]
+  },
   "sm_puctuationOverrides": {
     "properties": {
       "permitted_marks": "array",
@@ -1220,10 +1212,17 @@
           "v3"
         ]
       },
+      "speechModel": "string",
       "formatTurns": "boolean",
       "endOfTurnConfidenceThreshold": "number",
       "minEndOfTurnSilenceWhenConfident": "number",
-      "maxTurnSilence": "number"
+      "maxTurnSilence": "number",
+      "minTurnSilence": "number",
+      "keyterms": "array",
+      "prompt": "string",
+      "languageDetection": "boolean",
+      "vadThreshold": "number",
+      "inactivityTimeout": "number"
     }
   },
   "resource": {
@@ -1362,10 +1361,12 @@
       "tts": "#synthesizer",
       "vad": "#vad",
       "turnDetection": "#turnDetectionPipeline",
+      "turnGate": "#turnGatePipeline",
       "llm": "#llm",
       "preflightLlm": "boolean",
       "actionHook": "object|string",
-      "eventHook": "object|string"
+      "eventHook": "object|string",
+      "toolHook": "object|string"
     },
     "required": [
       "stt",
@@ -1385,5 +1386,32 @@
     "required": [
       "vendor"
     ]
+  },
+  "turnGatePipeline": {
+    "properties": {
+      "enabled": "boolean",
+      "vendor": "string",
+      "model": "string",
+      "auth": "object"
+    },
+    "required": [
+      "enabled"
+    ]
+  },
+  "noiseIsolation" : {
+    "properties": {
+      "enable": "boolean",
+      "vendor": "string",
+      "level": "number",
+      "model": "string"
+    }
+  },
+  "turnTaking": {
+    "properties": {
+      "enable": "boolean",
+      "vendor": "string",
+      "threshold": "number",
+      "model": "string"
+    }
   }
 }

test/jambonz-verb-test.js

@@ -1,6 +1,6 @@
 const test = require('tape');
 const logger = require('pino')({level: process.env.JAMBONES_LOGLEVEL || 'error'});
-const { validate } = require('..');
+const { validate, normalizeJambones } = require('..');
 
 test("validate correct verbs", async (t) => {
 
@@ -329,7 +329,8 @@ test("validate correct verbs", async (t) => {
           "sessionTimeout": 30000,
           "connectionTimeout": 5000,
           "customVocabulary": ["jambonz", "telephony", "voip"],
-          "languageModel": "enhanced"
+          "languageModel": "enhanced",
+          "audioQueryAbsoluteTimeout": 5
         },
         "gladiaOptions": {
           "post_processing": {
@@ -468,6 +469,24 @@ test("validate correct verbs", async (t) => {
         "speechPadMs": 1000
       }
     },
+    {
+      "verb": "config",
+      "noiseIsolation": {
+        "enable": true,
+        "vendor": "krisp",
+        "level": 3,
+        "model": "custom-model"
+      }
+    },
+    {
+      "verb": "config",
+      "turnTaking": {
+        "enable": true,
+        "vendor": "krisp",
+        "threshold": 0.5,
+        "model": "turn-taking-model"
+      }
+    },
     {
       "verb": "message",
       "to": "15083084809",
@@ -644,6 +663,74 @@ test("validate correct verbs", async (t) => {
           }
         }
       ]
+    },
+    {
+      "verb": "s2s",
+      "vendor": "openai",
+      "llmOptions": {
+        "model": "gpt-4o-realtime"
+      }
+    },
+    {
+      "verb": "openai_s2s",
+      "llmOptions": {
+        "model": "gpt-4o-realtime"
+      }
+    },
+    {
+      "verb": "google_s2s",
+      "llmOptions": {
+        "model": "gemini-2.0-flash"
+      }
+    },
+    {
+      "verb": "elevenlabs_s2s",
+      "llmOptions": {
+        "agentId": "agent-123"
+      }
+    },
+    {
+      "verb": "stream",
+      "url": "wss://myrecorder.example.com/calls",
+      "mixType": "stereo"
+    },
+    {
+      "verb": "pipeline",
+      "stt": {
+        "vendor": "google",
+        "language": "en-US"
+      },
+      "tts": {
+        "vendor": "google",
+        "language": "en-US"
+      },
+      "llm": {
+        "vendor": "openai",
+        "llmOptions": {
+          "model": "gpt-4o"
+        }
+      },
+      "actionHook": "/pipeline/action",
+      "eventHook": "/pipeline/event",
+      "toolHook": "/pipeline/tool"
+    },
+    {
+      "verb": "transcribe",
+      "transcriptionHook": "http://example.com/transcribe",
+      "recognizer": {
+        "vendor": "speechmatics",
+        "language": "en",
+        "speechmaticsOptions": {
+          "transcription_config": {
+            "language": "en",
+            "enable_partials": true,
+            "max_delay": 2,
+            "conversation_config": {
+              "end_of_utterance_silence_trigger": 0.5
+            }
+          }
+        }
+      }
     }
   ];
   try {
@@ -671,6 +758,110 @@ test('invalid test', async (t) => {
   } catch(err) {
     t.ok(1 == 1,'successfully validate verbs');
   }
-  
+
   t.end();
-})
+});
+
+test('verb synonyms: stream is synonym for listen', async (t) => {
+  // "verb" format
+  const result1 = normalizeJambones(logger, [
+    {"verb": "stream", "url": "wss://example.com/calls", "mixType": "stereo"}
+  ]);
+  t.equal(Object.keys(result1[0])[0], 'listen', 'stream verb is rewritten to listen');
+  t.equal(result1[0].listen.url, 'wss://example.com/calls', 'data is preserved');
+
+  // object-key format
+  const result2 = normalizeJambones(logger, [
+    {"stream": {"url": "wss://example.com/calls"}}
+  ]);
+  t.equal(Object.keys(result2[0])[0], 'listen', 'stream key is rewritten to listen');
+
+  // validate passes
+  try {
+    validate(logger, [{"verb": "stream", "url": "wss://example.com/calls"}]);
+    t.pass('stream verb validates successfully');
+  } catch (err) {
+    t.fail('stream verb should validate: ' + err);
+  }
+  t.end();
+});
+
+test('verb synonyms: s2s is synonym for llm', async (t) => {
+  const result = normalizeJambones(logger, [
+    {"verb": "s2s", "vendor": "openai", "llmOptions": {"model": "gpt-4o"}}
+  ]);
+  t.equal(Object.keys(result[0])[0], 'llm', 's2s verb is rewritten to llm');
+  t.equal(result[0].llm.vendor, 'openai', 'vendor is preserved');
+  t.equal(result[0].llm.llmOptions.model, 'gpt-4o', 'llmOptions preserved');
+
+  try {
+    validate(logger, [{"verb": "s2s", "vendor": "openai", "llmOptions": {"model": "gpt-4o"}}]);
+    t.pass('s2s verb validates successfully');
+  } catch (err) {
+    t.fail('s2s verb should validate: ' + err);
+  }
+  t.end();
+});
+
+test('vendor shortcuts: openai_s2s injects vendor', async (t) => {
+  const result = normalizeJambones(logger, [
+    {"verb": "openai_s2s", "llmOptions": {"model": "gpt-4o-realtime"}}
+  ]);
+  t.equal(Object.keys(result[0])[0], 'llm', 'openai_s2s is rewritten to llm');
+  t.equal(result[0].llm.vendor, 'openai', 'vendor is injected');
+  t.equal(result[0].llm.llmOptions.model, 'gpt-4o-realtime', 'llmOptions preserved');
+
+  try {
+    validate(logger, [{"verb": "openai_s2s", "llmOptions": {"model": "gpt-4o-realtime"}}]);
+    t.pass('openai_s2s validates successfully');
+  } catch (err) {
+    t.fail('openai_s2s should validate: ' + err);
+  }
+  t.end();
+});
+
+test('vendor shortcuts: all vendors work', async (t) => {
+  const vendors = [
+    'openai', 'microsoft', 'google', 'elevenlabs', 'deepgram', 'voiceagent', 'ultravox'
+  ];
+  for (const vendor of vendors) {
+    const verbName = `${vendor}_s2s`;
+    const result = normalizeJambones(logger, [
+      {"verb": verbName, "llmOptions": {}}
+    ]);
+    t.equal(Object.keys(result[0])[0], 'llm', `${verbName} rewrites to llm`);
+    t.equal(result[0].llm.vendor, vendor, `${verbName} injects vendor=${vendor}`);
+  }
+  t.end();
+});
+
+test('vendor shortcuts: object-key format works', async (t) => {
+  const result = normalizeJambones(logger, [
+    {"google_s2s": {"llmOptions": {"model": "gemini-2.0-flash"}}}
+  ]);
+  t.equal(Object.keys(result[0])[0], 'llm', 'google_s2s key is rewritten to llm');
+  t.equal(result[0].llm.vendor, 'google', 'vendor is injected');
+  t.equal(result[0].llm.llmOptions.model, 'gemini-2.0-flash', 'llmOptions preserved');
+  t.end();
+});
+
+test('vendor shortcuts: explicit vendor in data overrides injected vendor', async (t) => {
+  const result = normalizeJambones(logger, [
+    {"verb": "openai_s2s", "vendor": "custom", "llmOptions": {}}
+  ]);
+  t.equal(result[0].llm.vendor, 'custom', 'explicit vendor takes precedence');
+  t.end();
+});
+
+test('non-synonym verbs are unchanged', async (t) => {
+  const result = normalizeJambones(logger, [
+    {"verb": "say", "text": "hello"}
+  ]);
+  t.equal(Object.keys(result[0])[0], 'say', 'say verb is not transformed');
+
+  const result2 = normalizeJambones(logger, [
+    {"llm": {"vendor": "openai", "llmOptions": {}}}
+  ]);
+  t.equal(Object.keys(result2[0])[0], 'llm', 'llm verb is not transformed');
+  t.end();
+});