support houndify ws (#1540)

lib/tasks/stt-task.js

@@ -427,6 +427,7 @@ class SttTask extends Task {
       'google',
       'speechmatics',
       'openai',
+      'houndify',
     ].includes(vendor));
   }
 

lib/utils/transcription-utils.js

@@ -138,35 +138,7 @@ const stickyVars = {
     'HOUNDIFY_MAX_SILENCE_SECONDS',
     'HOUNDIFY_MAX_SILENCE_AFTER_FULL_QUERY_SECONDS',
     'HOUNDIFY_MAX_SILENCE_AFTER_PARTIAL_QUERY_SECONDS',
-    'HOUNDIFY_VAD_SENSITIVITY',
+    'HOUNDIFY_MODEL'
-    'HOUNDIFY_VAD_TIMEOUT',
-    'HOUNDIFY_VAD_MODE',
-    'HOUNDIFY_VAD_VOICE_MS',
-    'HOUNDIFY_VAD_SILENCE_MS',
-    'HOUNDIFY_VAD_DEBUG',
-    'HOUNDIFY_AUDIO_FORMAT',
-    'HOUNDIFY_ENABLE_NOISE_REDUCTION',
-    'HOUNDIFY_AUDIO_ENDPOINT',
-    'HOUNDIFY_ENABLE_PROFANITY_FILTER',
-    'HOUNDIFY_ENABLE_PUNCTUATION',
-    'HOUNDIFY_ENABLE_CAPITALIZATION',
-    'HOUNDIFY_CONFIDENCE_THRESHOLD',
-    'HOUNDIFY_ENABLE_DISFLUENCY_FILTER',
-    'HOUNDIFY_MAX_RESULTS',
-    'HOUNDIFY_ENABLE_WORD_TIMESTAMPS',
-    'HOUNDIFY_MAX_ALTERNATIVES',
-    'HOUNDIFY_PARTIAL_TRANSCRIPT_INTERVAL',
-    'HOUNDIFY_SESSION_TIMEOUT',
-    'HOUNDIFY_CONNECTION_TIMEOUT',
-    'HOUNDIFY_LATITUDE',
-    'HOUNDIFY_LONGITUDE',
-    'HOUNDIFY_CITY',
-    'HOUNDIFY_STATE',
-    'HOUNDIFY_COUNTRY',
-    'HOUNDIFY_TIMEZONE',
-    'HOUNDIFY_DOMAIN',
-    'HOUNDIFY_CUSTOM_VOCABULARY',
-    'HOUNDIFY_LANGUAGE_MODEL'
   ],
 };
 
@@ -646,15 +618,50 @@ const normalizeAssemblyAi = (evt, channel, language) => {
 const normalizeHoundify = (evt, channel, language) => {
   const copy = JSON.parse(JSON.stringify(evt));
   const alternatives = [];
-  const is_final = evt.ResultsAreFinal && evt.ResultsAreFinal[0] === true;
+  let is_final = false;
-  if (evt.Disambiguation && evt.Disambiguation.ChoiceData && evt.Disambiguation.ChoiceData.length > 0) {
+
-    // Handle Houndify Voice Search Result format
+  if (evt.type) {
-    const choiceData = evt.Disambiguation.ChoiceData[0];
+    /* WS API format: has "type" field (PartialTranscript, FinalSegmentTranscript, FinalTranscript, VadMessage) */
-    alternatives.push({
+    if (evt.type === 'VadMessage') {
-      confidence: choiceData.ConfidenceScore || choiceData.ASRConfidence || 0.0,
+      /* VadMessage is not a transcription result - return empty alternatives so callers skip it */
-      transcript: choiceData.FormattedTranscription || choiceData.Transcription || '',
+      return {
-    });
+        language_code: language,
+        channel_tag: channel,
+        is_final: false,
+        alternatives: [],
+        vendor: {name: 'houndify', evt: copy}
+      };
+    }
+
+    is_final = evt.type === 'FinalSegmentTranscript' || evt.type === 'FinalTranscript';
+
+    if (evt.hypotheses && evt.hypotheses.length > 0) {
+      /* FinalSegmentTranscript / FinalTranscript with hypotheses */
+      const best = evt.hypotheses[0];
+      alternatives.push({
+        confidence: best.confidence || 0.0,
+        transcript: best.text || evt.text || '',
+      });
+    } else if (evt.text !== undefined) {
+      /* PartialTranscript */
+      alternatives.push({
+        confidence: evt.confidence || (evt.eoq !== undefined ? 1.0 - evt.eoq : 0.8),
+        transcript: evt.text || '',
+      });
+    }
+  } else {
+    /* SDK format: has Disambiguation/ResultsAreFinal fields */
+    is_final = evt.ResultsAreFinal && evt.ResultsAreFinal[0] === true;
+    if (evt.Disambiguation && evt.Disambiguation.ChoiceData && evt.Disambiguation.ChoiceData.length > 0) {
+      // Handle Houndify Voice Search Result format
+      const choiceData = evt.Disambiguation.ChoiceData[0];
+      alternatives.push({
+        confidence: choiceData.ConfidenceScore || choiceData.ASRConfidence || 0.0,
+        transcript: choiceData.FormattedTranscription || choiceData.Transcription || '',
+      });
+    }
   }
+
   return {
     language_code: language,
     channel_tag: channel,
@@ -1220,8 +1227,10 @@ module.exports = (logger) => {
         enableCapitalization, confidenceThreshold, enableDisfluencyFilter,
         maxResults, enableWordTimestamps, maxAlternatives, partialTranscriptInterval,
         sessionTimeout, connectionTimeout, customVocabulary, languageModel,
-        requestInfo, sampleRate
+        requestInfo, sampleRate, audioQueryAbsoluteTimeout,
+        eoqThreshold, vadStopThreshold
       } = rOpts.houndifyOptions || {};
+      const {model} = rOpts;
       const audioEndpointUri = audioEndpoint || sttCredentials.houndify_server_uri;
 
       opts = {
@@ -1232,6 +1241,7 @@ module.exports = (logger) => {
         HOUNDIFY_MAX_SILENCE_SECONDS: maxSilenceSeconds || 5,
         HOUNDIFY_MAX_SILENCE_AFTER_FULL_QUERY_SECONDS: maxSilenceAfterFullQuerySeconds || 1,
         HOUNDIFY_MAX_SILENCE_AFTER_PARTIAL_QUERY_SECONDS: maxSilenceAfterPartialQuerySeconds || 1.5,
+        HOUNDIFY_MODEL: model || 'base_8k',
         ...(vadSensitivity && {HOUNDIFY_VAD_SENSITIVITY: vadSensitivity}),
         ...(vadTimeout && {HOUNDIFY_VAD_TIMEOUT: vadTimeout}),
         ...(vadMode && {HOUNDIFY_VAD_MODE: vadMode}),
@@ -1264,6 +1274,9 @@ module.exports = (logger) => {
         ...(languageModel && {HOUNDIFY_LANGUAGE_MODEL: languageModel}),
         ...(requestInfo && {HOUNDIFY_REQUEST_INFO: JSON.stringify(requestInfo)}),
         ...(sampleRate && {HOUNDIFY_SAMPLING_RATE: sampleRate}),
+        ...(audioQueryAbsoluteTimeout && {HOUNDIFY_AUDIO_QUERY_ABSOLUTE_TIMEOUT: audioQueryAbsoluteTimeout}),
+        ...(eoqThreshold !== undefined && {HOUNDIFY_EOQ_THRESHOLD: eoqThreshold}),
+        ...(vadStopThreshold !== undefined && {HOUNDIFY_VAD_STOP_THRESHOLD: vadStopThreshold}),
       };
     }
     else if ('voxist' === vendor) {

package-lock.json

@@ -18,7 +18,7 @@
         "@jambonz/speech-utils": "^0.2.30",
         "@jambonz/stats-collector": "^0.1.10",
         "@jambonz/time-series": "^0.2.17",
-        "@jambonz/verb-specifications": "^0.0.125",
+        "@jambonz/verb-specifications": "^0.1.11",
         "@modelcontextprotocol/sdk": "^1.9.0",
         "@opentelemetry/api": "^1.8.0",
         "@opentelemetry/exporter-jaeger": "^1.23.0",
@@ -1647,9 +1647,9 @@
       }
     },
     "node_modules/@jambonz/verb-specifications": {
-      "version": "0.0.125",
+      "version": "0.1.11",
-      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.0.125.tgz",
+      "resolved": "https://registry.npmjs.org/@jambonz/verb-specifications/-/verb-specifications-0.1.11.tgz",
-      "integrity": "sha512-lU1fyyYyjXOdIfQ2gmOFmssZASYNu6LD066iXjqFrBJpiI7shkprcZ1qeWGibuEk9nR2k+em3/YL31Wc8L4wvA==",
+      "integrity": "sha512-2YWqBXhLUD+WLipMdysE155eOSgFVW3id0l4Ci37OYELJIG+GeJYMubReSlMXthU/WoVk91ATu+XkSsjuVO1dg==",
       "license": "MIT",
       "dependencies": {
         "debug": "^4.3.4",

package.json

@@ -34,7 +34,7 @@
     "@jambonz/speech-utils": "^0.2.30",
     "@jambonz/stats-collector": "^0.1.10",
     "@jambonz/time-series": "^0.2.17",
-    "@jambonz/verb-specifications": "^0.0.125",
+    "@jambonz/verb-specifications": "^0.1.11",
     "@modelcontextprotocol/sdk": "^1.9.0",
     "@opentelemetry/api": "^1.8.0",
     "@opentelemetry/exporter-jaeger": "^1.23.0",