botium-connector-voip 0.0.28 → 0.0.29

package/dist/botium-connector-voip-cjs.js

@@ -30,7 +30,7 @@ const _info = (event, data) => {
   const parts = Object.entries({
     event,
     ...data
-  }).filter(([, v]) => v != null && v !== '').map(([k, v]) => `${k}=${typeof v === 'string' && v.length > 80 ? `"${v.substring(0, 77)}..."` : JSON.stringify(v)}`);
+  }).filter(([, v]) => v != null && v !== '').map(([k, v]) => `${k}=${JSON.stringify(v)}`);
   console.info(`[botium-connector-voip] ${parts.join(' ')}`);
 };
 
@@ -51,6 +51,7 @@ const Capabilities = {
   VOIP_STT_MESSAGE_HANDLING_TIMEOUT: 'VOIP_STT_MESSAGE_HANDLING_TIMEOUT',
   VOIP_STT_MESSAGE_HANDLING_TIMEOUT_SUBSEQUENT: 'VOIP_STT_MESSAGE_HANDLING_TIMEOUT_SUBSEQUENT',
   VOIP_JOIN_SILENCE_DURATION_BY_SUBSTRING: 'VOIP_JOIN_SILENCE_DURATION_BY_SUBSTRING',
+  VOIP_STT_DICTIONARY_REPLACEMENTS: 'VOIP_STT_DICTIONARY_REPLACEMENTS',
   VOIP_STT_MESSAGE_HANDLING_DELIMITER: 'VOIP_STT_MESSAGE_HANDLING_DELIMITER',
   VOIP_STT_MESSAGE_HANDLING_PUNCTUATION: 'VOIP_STT_MESSAGE_HANDLING_PUNCTUATION',
   VOIP_TTS_URL: 'VOIP_TTS_URL',
@@ -90,7 +91,8 @@ const Capabilities = {
   VOIP_STT_CONFIDENCE_THRESHOLD: 'VOIP_STT_CONFIDENCE_THRESHOLD',
   VOIP_USE_GLOBAL_VOIP_WORKER: 'VOIP_USE_GLOBAL_VOIP_WORKER',
   VOIP_USER_INPUT_PREFER_VOICE: 'VOIP_USER_INPUT_PREFER_VOICE',
-  VOIP_EMIT_SPECULATIVE_TEXT: 'VOIP_EMIT_SPECULATIVE_TEXT'
+  VOIP_EMIT_SPECULATIVE_TEXT: 'VOIP_EMIT_SPECULATIVE_TEXT',
+  VOIP_SDP_MEDIA_TYPE_TEXT_ENABLE: 'VOIP_SDP_MEDIA_TYPE_TEXT_ENABLE'
 };
 const Defaults = {
   VOIP_STT_METHOD: 'POST',
@@ -113,7 +115,8 @@ const Defaults = {
   VOIP_STT_CONFIDENCE_THRESHOLD: 0.5,
   VOIP_USE_GLOBAL_VOIP_WORKER: false,
   VOIP_SIP_PROTOCOL: 'TCP',
-  VOIP_USER_INPUT_PREFER_VOICE: true
+  VOIP_USER_INPUT_PREFER_VOICE: true,
+  VOIP_SDP_MEDIA_TYPE_TEXT_ENABLE: false
 };
 const TTS_HTTP_AGENT = new http__default["default"].Agent({
   keepAlive: true
@@ -557,6 +560,7 @@ class BotiumConnectorVoip {
           ICE_TURN_PASSWORD: this.caps[Capabilities.VOIP_ICE_TURN_PASSWORD],
           ICE_TURN_PROTOCOL: this.caps[Capabilities.VOIP_ICE_TURN_PROTOCOL] || 'TCP',
           MIN_SILENCE_DURATION: this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT_ENABLE] ? this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT] : null,
+          SDP_MEDIA_TYPE_TEXT_ENABLE: !!this.caps[Capabilities.VOIP_SDP_MEDIA_TYPE_TEXT_ENABLE],
           STT_LEGACY: sttLegacy,
           STT_CONFIG: {
             stt_url: sttUrl,
@@ -846,14 +850,15 @@ class BotiumConnectorVoip {
             this.sttPartialCount++;
             const partialText = parsedData.data.message;
             if (typeof partialText === 'string' && partialText.trim().length > 0) {
+              const replacementResult = this._applySttDictionaryReplacements(partialText);
               this.lastPartialBotMsg = {
-                messageText: partialText,
-                sourceData: Object.assign({}, parsedData, {
+                messageText: replacementResult.text,
+                sourceData: Object.assign({}, this._decorateSourceDataWithSttDictionaryReplacements(parsedData, replacementResult), {
                   partialRecovery: true
                 })
               };
             }
-            const partialPreview = typeof partialText === 'string' ? partialText.trim().substring(0, 60) : '';
+            const partialPreview = typeof partialText === 'string' ? partialText.trim() : '';
             _info('stt_partial_received', {
               sessionId: this.sessionId,
               partialIndex: this.sttPartialCount,
@@ -935,11 +940,13 @@ class BotiumConnectorVoip {
             const confidenceThreshold = this._getConfidenceScoreLogicHook(this.convoStep) && this._getConfidenceScoreLogicHook(this.convoStep).args[0] || this.caps[Capabilities.VOIP_STT_CONFIDENCE_THRESHOLD];
             const successfulConfidenceScore = this._getConfidenceScore(parsedData) >= confidenceThreshold;
             const msgText = parsedData.data.message || '';
+            const replacementResult = this._applySttDictionaryReplacements(msgText);
+            const normalizedMsgText = replacementResult.text;
             const msgLen = typeof msgText === 'string' ? msgText.length : 0;
-            const msgPreview = typeof msgText === 'string' ? msgText.trim().substring(0, 80) : '';
+            const msgPreview = typeof normalizedMsgText === 'string' ? normalizedMsgText.trim() : '';
             // A final supersedes the cached interim; clear to avoid duplicate tail emission.
             this.lastPartialBotMsg = null;
-            debug$3(`Message: ${parsedData.data.message} / Confidence Score: ${this._getConfidenceScore(parsedData)} (Threshold: ${confidenceThreshold})`);
+            debug$3(`Message: ${normalizedMsgText} / Confidence Score: ${this._getConfidenceScore(parsedData)} (Threshold: ${confidenceThreshold})`);
             _info('stt_final', {
               sessionId: this.sessionId,
               message: msgPreview,
@@ -951,10 +958,10 @@ class BotiumConnectorVoip {
             // ORIGINAL: always emit final message immediately (ignore JOIN hooks/rules).
             if (this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'ORIGINAL') {
               let botMsg = {
-                messageText: parsedData.data.message
+                messageText: normalizedMsgText
               };
               if (this.firstMsg) {
-                const sourceData = parsedData;
+                const sourceData = this._decorateSourceDataWithSttDictionaryReplacements(parsedData, replacementResult);
                 sourceData.silenceDuration = parsedData.data.start;
                 sourceData.voiceDuration = parsedData.data.end - parsedData.data.start;
                 botMsg = Object.assign({}, botMsg, {
@@ -962,7 +969,7 @@ class BotiumConnectorVoip {
                 });
                 this.firstMsg = false;
               } else {
-                const sourceData = parsedData;
+                const sourceData = this._decorateSourceDataWithSttDictionaryReplacements(parsedData, replacementResult);
                 const start = lodash__default["default"].get(parsedData, 'data.start', null);
                 const prevEnd = lodash__default["default"].get(this.prevData, 'data.end', null);
                 sourceData.silenceDuration = lodash__default["default"].isFinite(start) && lodash__default["default"].isFinite(prevEnd) ? start - prevEnd : lodash__default["default"].isFinite(start) ? start : null;
@@ -980,10 +987,10 @@ class BotiumConnectorVoip {
             }
             if (this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'SPLIT' || this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'EXPAND') {
               let botMsg = {
-                messageText: parsedData.data.message
+                messageText: normalizedMsgText
               };
               if (this.firstMsg) {
-                const sourceData = parsedData;
+                const sourceData = this._decorateSourceDataWithSttDictionaryReplacements(parsedData, replacementResult);
                 sourceData.silenceDuration = parsedData.data.start;
                 sourceData.voiceDuration = parsedData.data.end - parsedData.data.start;
                 botMsg = Object.assign({}, botMsg, {
@@ -991,7 +998,7 @@ class BotiumConnectorVoip {
                 });
                 this.firstMsg = false;
               } else {
-                const sourceData = parsedData;
+                const sourceData = this._decorateSourceDataWithSttDictionaryReplacements(parsedData, replacementResult);
                 const start = lodash__default["default"].get(parsedData, 'data.start', null);
                 const prevEnd = lodash__default["default"].get(this.prevData, 'data.end', null);
                 sourceData.silenceDuration = lodash__default["default"].isFinite(start) && lodash__default["default"].isFinite(prevEnd) ? start - prevEnd : lodash__default["default"].isFinite(start) ? start : null;
@@ -1010,8 +1017,8 @@ class BotiumConnectorVoip {
             }
             if (this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'JOIN' || this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'PSST' || this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'CONCAT') {
               const botMsg = {
-                messageText: parsedData.data.message,
-                sourceData: parsedData
+                messageText: normalizedMsgText,
+                sourceData: this._decorateSourceDataWithSttDictionaryReplacements(parsedData, replacementResult)
               };
               this.prevData = parsedData;
               if (successfulConfidenceScore) {
@@ -1521,27 +1528,99 @@ class BotiumConnectorVoip {
       };
     }).filter(Boolean);
   }
-  _getJoinRuleBySubstring(convoStep, botMsgs) {
-    const texts = [];
-    const stepMessageText = lodash__default["default"].get(convoStep, 'messageText', '');
-    if (typeof stepMessageText === 'string' && stepMessageText.length > 0) {
-      texts.push(stepMessageText);
-    }
-    if (lodash__default["default"].isArray(botMsgs) && botMsgs.length > 0) {
-      const bufferedText = botMsgs.map(m => m && typeof m.messageText === 'string' ? m.messageText : '').filter(Boolean).join(' ');
-      if (bufferedText.length > 0) texts.push(bufferedText);
-    }
-    if (texts.length === 0) return null;
+
+  // Matches a substring rule against ONLY the latest buffered STT final chunk.
+  // This keeps the rule's custom timeout in effect for just the one following
+  // final: once a final no longer matches, callers fall back to the default
+  // timeout (the convoStep expected text and the cumulative buffer are
+  // deliberately not considered, so a stale match cannot stick).
+  _getJoinRuleBySubstring(botMsgs) {
+    if (!lodash__default["default"].isArray(botMsgs) || botMsgs.length === 0) return null;
+    const last = botMsgs[botMsgs.length - 1];
+    const text = last && typeof last.messageText === 'string' ? last.messageText : '';
+    if (!text) return null;
+    const loweredText = text.toLowerCase();
     const rules = this._normalizeJoinRulesBySubstring();
-    for (const text of texts) {
-      const loweredText = text.toLowerCase();
-      const match = rules.find(rule => loweredText.includes(rule.substring.toLowerCase()));
-      if (match) return match;
+    return rules.find(rule => loweredText.includes(rule.substring.toLowerCase())) || null;
+  }
+  _normalizeSttDictionaryReplacements() {
+    const rawRules = this.caps[Capabilities.VOIP_STT_DICTIONARY_REPLACEMENTS];
+    if (lodash__default["default"].isNil(rawRules) || rawRules === '') return [];
+    let parsedRules = rawRules;
+    if (lodash__default["default"].isString(rawRules)) {
+      try {
+        parsedRules = JSON.parse(rawRules);
+      } catch (err) {
+        debug$3(`Invalid ${Capabilities.VOIP_STT_DICTIONARY_REPLACEMENTS} JSON: ${err.message || err}`);
+        return [];
+      }
     }
-    return null;
+    if (!lodash__default["default"].isArray(parsedRules)) {
+      debug$3(`Invalid ${Capabilities.VOIP_STT_DICTIONARY_REPLACEMENTS}: expected array`);
+      return [];
+    }
+    return parsedRules.map(rule => {
+      if (!rule || typeof rule !== 'object') return null;
+      const fromValues = lodash__default["default"].isArray(rule.from) ? rule.from : [rule.from];
+      const from = lodash__default["default"].uniq(fromValues.map(value => value != null ? String(value).trim() : '').filter(Boolean));
+      const to = rule.to != null ? String(rule.to).trim() : '';
+      if (from.length === 0 || !to) return null;
+      return {
+        from,
+        to
+      };
+    }).filter(Boolean);
+  }
+  _applySttDictionaryReplacements(text) {
+    if (!lodash__default["default"].isString(text) || text.length === 0) return {
+      text,
+      applied: []
+    };
+    const replacements = this._normalizeSttDictionaryReplacements();
+    if (replacements.length === 0) return {
+      text,
+      applied: []
+    };
+    const replacementsByFrom = new Map();
+    const fromAlternatives = [];
+    replacements.forEach(rule => {
+      rule.from.forEach(from => {
+        const fromKey = from.toLowerCase();
+        if (!replacementsByFrom.has(fromKey)) {
+          replacementsByFrom.set(fromKey, rule.to);
+          fromAlternatives.push(from);
+        }
+      });
+    });
+    if (fromAlternatives.length === 0) return {
+      text,
+      applied: []
+    };
+    fromAlternatives.sort((a, b) => b.length - a.length);
+    const matcher = new RegExp(fromAlternatives.map(value => lodash__default["default"].escapeRegExp(value)).join('|'), 'gi');
+    const applied = [];
+    const replacedText = text.replace(matcher, match => {
+      const to = replacementsByFrom.get(match.toLowerCase());
+      applied.push({
+        from: match,
+        to
+      });
+      return to;
+    });
+    return {
+      text: replacedText,
+      applied
+    };
+  }
+  _decorateSourceDataWithSttDictionaryReplacements(sourceData, replacementResult) {
+    if (!replacementResult || !lodash__default["default"].isArray(replacementResult.applied) || replacementResult.applied.length === 0) return sourceData;
+    return Object.assign({}, sourceData, {
+      sttDictionaryOriginalMessage: replacementResult.text === sourceData?.data?.message ? undefined : sourceData?.data?.message,
+      sttDictionaryReplacements: replacementResult.applied
+    });
   }
   _hasJoinLogicHookOrRule(convoStep) {
-    return !lodash__default["default"].isNil(this._getJoinLogicHook(convoStep)) || !lodash__default["default"].isNil(this._getJoinRuleBySubstring(convoStep, this.botMsgs));
+    return !lodash__default["default"].isNil(this._getJoinLogicHook(convoStep)) || !lodash__default["default"].isNil(this._getJoinRuleBySubstring(this.botMsgs));
   }
   _toJoinTimeoutMs(ms, isPsst) {
     const parsed = parseInt(ms, 10);
@@ -1556,7 +1635,7 @@ class BotiumConnectorVoip {
       const joinHookTimeoutMs = this._toJoinTimeoutMs(joinLogicHook.args[0], isPsst);
       if (lodash__default["default"].isFinite(joinHookTimeoutMs) && joinHookTimeoutMs > 0) return joinHookTimeoutMs;
     }
-    const joinRule = this._getJoinRuleBySubstring(convoStep, botMsgs);
+    const joinRule = this._getJoinRuleBySubstring(botMsgs);
     if (joinRule) {
       const joinRuleTimeoutMs = this._toJoinTimeoutMs(joinRule.timeoutMs, isPsst);
       if (lodash__default["default"].isFinite(joinRuleTimeoutMs) && joinRuleTimeoutMs > 0) return joinRuleTimeoutMs;
@@ -1654,6 +1733,18 @@ var botiumConnectorVoip = {
       type: 'json',
       required: false,
       advanced: true
+    }, {
+      name: 'VOIP_STT_DICTIONARY_REPLACEMENTS',
+      label: 'STT dictionary replacements',
+      type: 'json',
+      required: false,
+      advanced: true
+    }, {
+      name: 'VOIP_SDP_MEDIA_TYPE_TEXT_ENABLE',
+      label: 'Enable SDP media type text',
+      type: 'boolean',
+      required: false,
+      advanced: true
     }]
   },
   PluginLogicHooks: {