botium-connector-voip 0.0.26 → 0.0.27

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

@@ -50,6 +50,7 @@ const Capabilities = {
   VOIP_STT_MESSAGE_HANDLING: 'VOIP_STT_MESSAGE_HANDLING',
   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_MESSAGE_HANDLING_DELIMITER: 'VOIP_STT_MESSAGE_HANDLING_DELIMITER',
   VOIP_STT_MESSAGE_HANDLING_PUNCTUATION: 'VOIP_STT_MESSAGE_HANDLING_PUNCTUATION',
   VOIP_TTS_URL: 'VOIP_TTS_URL',
@@ -250,19 +251,9 @@ class BotiumConnectorVoip {
     const armJoinSilenceTimer = () => {
       if (!this.botMsgs || this.botMsgs.length === 0) return;
       const sttHandling = this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING];
-      const joinLogicHook = this._getJoinLogicHook(this.convoStep);
-      const isJoinMethod = sttHandling === 'JOIN' || sttHandling === 'PSST' || sttHandling === 'CONCAT' || !lodash__default["default"].isNil(joinLogicHook);
+      const isJoinMethod = sttHandling === 'JOIN' || sttHandling === 'PSST' || sttHandling === 'CONCAT' || this._hasJoinLogicHookOrRule(this.convoStep);
       if (!isJoinMethod) return;
-      const isPsst = sttHandling === 'PSST';
-      const toJoinTimeoutMs = ms => {
-        const parsed = parseInt(ms, 10);
-        if (!lodash__default["default"].isFinite(parsed) || parsed <= 0) return null;
-        return isPsst ? Math.max(0, parsed - 500) : parsed;
-      };
-      let joinTimeoutMs = toJoinTimeoutMs(lodash__default["default"].get(joinLogicHook, 'args[0]'));
-      if (!lodash__default["default"].isFinite(joinTimeoutMs) || joinTimeoutMs <= 0) {
-        joinTimeoutMs = toJoinTimeoutMs(this._getEffectiveMessageHandlingTimeout());
-      }
+      const joinTimeoutMs = this._getEffectiveJoinTimeoutMs(this.convoStep, this.botMsgs);
       if (this.silenceTimeout) {
         clearTimeout(this.silenceTimeout);
         this.silenceTimeout = null;
@@ -505,19 +496,7 @@ class BotiumConnectorVoip {
           // For PSST: send join silence duration per step to VOIP worker (controls PSST silence trigger)
           try {
             if (this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'PSST' && this.ws && this.ws.readyState === ws__default["default"].OPEN) {
-              const joinLogicHook = this._getJoinLogicHook(convoStep);
-              let silenceMs = null;
-              if (joinLogicHook && joinLogicHook.args && joinLogicHook.args.length > 0) {
-                silenceMs = parseInt(joinLogicHook.args[0], 10);
-              }
-              // Fallback to effective timeout (subsequent if past first flush, else global)
-              if (!lodash__default["default"].isFinite(silenceMs) || silenceMs <= 0) {
-                silenceMs = parseInt(this._getEffectiveMessageHandlingTimeout(), 10);
-              }
-              // PSST: treat like JOIN, but subtract 500ms from timeout
-              if (lodash__default["default"].isFinite(silenceMs) && silenceMs > 0) {
-                silenceMs = Math.max(0, silenceMs - 500);
-              }
+              const silenceMs = this._getEffectiveJoinTimeoutMs(convoStep, this.botMsgs);
               if (lodash__default["default"].isFinite(silenceMs) && silenceMs > 0 && this.sessionId) {
                 debug$3(`PSST: sending silenceDurationMs=${silenceMs} for sessionId=${this.sessionId}`);
                 this.ws.send(JSON.stringify({
@@ -826,7 +805,7 @@ class BotiumConnectorVoip {
           }
           if (parsedData && parsedData.type === 'silence') {
             if (lodash__default["default"].isNil(this._getIgnoreSilenceDurationAsserterLogicHook(this.convoStep))) {
-              if (lodash__default["default"].isNil(this._getJoinLogicHook(this.convoStep)) && parsedData.data.silence.length > 0) {
+              if (!this._hasJoinLogicHookOrRule(this.convoStep) && parsedData.data.silence.length > 0) {
                 flushPendingBotMsgs('silence_exceeded');
                 this.end = true;
                 sendBotMsg(new Error(`Silence Duration of ${parsedData.data.silence[0][2].toFixed(2)}s exceeded General Silence Duration Timeout of ${this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT] / 1000}s`));
@@ -907,7 +886,7 @@ class BotiumConnectorVoip {
               }
             }
             if (lodash__default["default"].isNil(this._getIgnoreSilenceDurationAsserterLogicHook(this.convoStep))) {
-              if (!this.firstSttInfoReceived && lodash__default["default"].isNil(this._getJoinLogicHook(this.convoStep)) && this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT_START_ENABLE] && parsedData.data.start && parsedData.data.start * 1000 > this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT_START]) {
+              if (!this.firstSttInfoReceived && !this._hasJoinLogicHookOrRule(this.convoStep) && this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT_START_ENABLE] && parsedData.data.start && parsedData.data.start * 1000 > this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT_START]) {
                 this.end = true;
                 sendBotMsg(new Error(`Silence Duration of ${parsedData.data.start}s exceeded Initial Silence Duration Timeout of ${this.caps[Capabilities.VOIP_SILENCE_DURATION_TIMEOUT_START] / 1000}s`));
               }
@@ -916,23 +895,11 @@ class BotiumConnectorVoip {
             if (this.prevData && this.prevData.data && !lodash__default["default"].isNil(this.prevData.data.end)) {
               if (!lodash__default["default"].isNil(parsedData.data.start)) {
                 const silenceDuration = parsedData.data.start - this.prevData.data.end;
-                const joinLogicHook = this._getJoinLogicHook(this.convoStep);
                 const sttHandling = this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING];
-                const isPsst = sttHandling === 'PSST';
-                const isJoinMethod = sttHandling === 'JOIN' || isPsst;
-                const toJoinTimeoutMs = ms => {
-                  const parsed = parseInt(ms, 10);
-                  if (!lodash__default["default"].isFinite(parsed) || parsed <= 0) return null;
-                  return isPsst ? Math.max(0, parsed - 500) : parsed;
-                };
+                const isJoinMethod = sttHandling === 'JOIN' || sttHandling === 'PSST';
                 let matched = false;
-                if (!lodash__default["default"].isNil(joinLogicHook) && isJoinMethod) {
-                  const joinTimeoutMs = toJoinTimeoutMs(joinLogicHook && joinLogicHook.args && joinLogicHook.args[0]);
-                  if (lodash__default["default"].isFinite(joinTimeoutMs) && joinTimeoutMs > 0 && silenceDuration > joinTimeoutMs / 1000) {
-                    matched = true;
-                  }
-                } else if (lodash__default["default"].isNil(joinLogicHook) && isJoinMethod) {
-                  const joinTimeoutMs = toJoinTimeoutMs(this._getEffectiveMessageHandlingTimeout());
+                if (isJoinMethod) {
+                  const joinTimeoutMs = this._getEffectiveJoinTimeoutMs(this.convoStep, this.botMsgs);
                   if (lodash__default["default"].isFinite(joinTimeoutMs) && joinTimeoutMs > 0 && silenceDuration > joinTimeoutMs / 1000) {
                     matched = true;
                   }
@@ -963,8 +930,8 @@ class BotiumConnectorVoip {
               threshold: confidenceThreshold,
               accepted: successfulConfidenceScore
             });
-            // ORIGINAL: emit final message immediately, unless join hooks are active.
-            if (this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING] === 'ORIGINAL' && lodash__default["default"].isNil(this._getJoinLogicHook(this.convoStep))) {
+            // 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
               };
@@ -1023,7 +990,7 @@ class BotiumConnectorVoip {
               }
               this.botMsgs = [];
             }
-            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' || !lodash__default["default"].isNil(this._getJoinLogicHook(this.convoStep))) {
+            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
@@ -1102,58 +1069,35 @@ class BotiumConnectorVoip {
     }
     return new Promise((resolve, reject) => {
       setTimeout(async () => {
-        let duration = 0;
-        const preferVoiceCapRaw = this.caps[Capabilities.VOIP_USER_INPUT_PREFER_VOICE];
-        const preferVoice = !!preferVoiceCapRaw;
-        const skipTtsForMixedInput = preferVoice && hasText && hasVoiceMedia;
-        debug$3(`UserSays routing: hasText=${hasText} hasVoiceMedia=${hasVoiceMedia} preferVoice=${preferVoice} preferVoiceRaw=${JSON.stringify(preferVoiceCapRaw)} skipTtsForMixedInput=${skipTtsForMixedInput}`);
+        try {
+          let duration = 0;
+          const preferVoiceCapRaw = this.caps[Capabilities.VOIP_USER_INPUT_PREFER_VOICE];
+          const preferVoice = !!preferVoiceCapRaw;
+          const skipTtsForMixedInput = preferVoice && hasText && hasVoiceMedia;
+          debug$3(`UserSays routing: hasText=${hasText} hasVoiceMedia=${hasVoiceMedia} preferVoice=${preferVoice} preferVoiceRaw=${JSON.stringify(preferVoiceCapRaw)} skipTtsForMixedInput=${skipTtsForMixedInput}`);
 
-        // Stamp `msg.voipAgent` at the moment bytes leave the WebSocket so
-        // the coach can place the agent turn on the recording timeline.
-        // `requestedDurationMs` is the best estimate of on-wire playback
-        // length (DTMF tones × digits, TTS synth output, parsed media duration).
-        const stampAgentWire = (wireKind, requestedDurationMs, extras = {}) => {
-          msg.voipAgent = {
-            wireSentAtMs: Date.now(),
-            inputType,
-            wireKind,
-            requestedDurationMs: Math.max(0, Math.round(requestedDurationMs || 0)),
-            ...extras
+          // Stamp `msg.voipAgent` at the moment bytes leave the WebSocket so
+          // the coach can place the agent turn on the recording timeline.
+          // `requestedDurationMs` is the best estimate of on-wire playback
+          // length (DTMF tones × digits, TTS synth output, parsed media duration).
+          const stampAgentWire = (wireKind, requestedDurationMs, extras = {}) => {
+            msg.voipAgent = {
+              wireSentAtMs: Date.now(),
+              inputType,
+              wireKind,
+              requestedDurationMs: Math.max(0, Math.round(requestedDurationMs || 0)),
+              ...extras
+            };
           };
-        };
-        // Twilio default: 100 ms tone + 100 ms gap per digit. Drives agent-bar width only.
-        const DTMF_MS_PER_DIGIT = 200;
-        if (msg && msg.buttons && msg.buttons.length > 0) {
-          const digits = sanitizeDtmfDigits(msg.buttons[0].payload);
-          if (!digits) {
-            debug$3('sendDtmf skipped: no valid DTMF digits after sanitizing button payload');
-            return resolve();
-          }
-          debug$3(`Sending DTMF digits: ${digits}`);
-          const request = JSON.stringify({
-            METHOD: 'sendDtmf',
-            digits,
-            sessionId: this.sessionId
-          });
-          stampAgentWire('dtmf', digits.length * DTMF_MS_PER_DIGIT, {
-            digitCount: digits.length
-          });
-          this.ws.send(request);
-        } else if (msg && msg.messageText) {
-          // Check for DTMF tag in messageText: <DTMF>1234</DTMF>
-          const dtmfMatch = msg.messageText.match(/<DTMF>([^<]+)<\/DTMF>/i);
-          if (dtmfMatch && dtmfMatch[1]) {
-            const rawDigits = dtmfMatch[1];
-            const digits = sanitizeDtmfDigits(rawDigits);
+          // Twilio default: 100 ms tone + 100 ms gap per digit. Drives agent-bar width only.
+          const DTMF_MS_PER_DIGIT = 200;
+          if (msg && msg.buttons && msg.buttons.length > 0) {
+            const digits = sanitizeDtmfDigits(msg.buttons[0].payload);
             if (!digits) {
-              debug$3(`sendDtmf skipped: no valid DTMF digits after sanitizing <DTMF> content (raw length=${String(rawDigits).length})`);
+              debug$3('sendDtmf skipped: no valid DTMF digits after sanitizing button payload');
               return resolve();
             }
-            if (digits !== String(rawDigits).replace(/\s/g, '')) {
-              debug$3(`Sending DTMF from messageText (sanitized): "${rawDigits}" -> "${digits}"`);
-            } else {
-              debug$3(`Sending DTMF from messageText: ${digits}`);
-            }
+            debug$3(`Sending DTMF digits: ${digits}`);
             const request = JSON.stringify({
               METHOD: 'sendDtmf',
               digits,
@@ -1162,114 +1106,150 @@ class BotiumConnectorVoip {
             stampAgentWire('dtmf', digits.length * DTMF_MS_PER_DIGIT, {
               digitCount: digits.length
             });
-            this.ws.send(request);
-            return resolve();
-          }
-          if (!skipTtsForMixedInput) {
-            if (!this.axiosTtsParams) {
-              if (!(msg.media && msg.media.length > 0 && msg.media[0].buffer)) {
-                return reject(new Error('TTS not configured, only audio input supported'));
-              }
-            } else {
-              debug$3('UserSays routing: executing TTS branch');
-              const ttsRequest = this._buildTtsRequest(msg.messageText);
-              if (!ttsRequest) return reject(new Error('TTS not configured, only audio input supported'));
-              msg.sourceData = ttsRequest;
-              let ttsResult = null;
-              const ttsStartedAt = Date.now();
-              let ttsSynthMs = 0;
-              try {
-                ttsResult = await this._getTtsAudio(ttsRequest, msg.messageText);
-                ttsSynthMs = Date.now() - ttsStartedAt;
-              } catch (err) {
-                return reject(new Error(`TTS "${msg.messageText}" failed - ${this._getAxiosErrOutput(err)}`));
+            this._sendUserSaysWs(request);
+          } else if (msg && msg.messageText) {
+            // Check for DTMF tag in messageText: <DTMF>1234</DTMF>
+            const dtmfMatch = msg.messageText.match(/<DTMF>([^<]+)<\/DTMF>/i);
+            if (dtmfMatch && dtmfMatch[1]) {
+              const rawDigits = dtmfMatch[1];
+              const digits = sanitizeDtmfDigits(rawDigits);
+              if (!digits) {
+                debug$3(`sendDtmf skipped: no valid DTMF digits after sanitizing <DTMF> content (raw length=${String(rawDigits).length})`);
+                return resolve();
               }
-              if (msg && msg.messageText && msg.messageText.length > 0) {
-                const apiKey = this._extractApiKey(this._getBody(Capabilities.VOIP_TTS_BODY));
-                this.eventEmitter.emit('CONSUMPTION_METADATA', this.container, {
-                  type: lodash__default["default"].isNil(apiKey) ? 'INBUILT' : 'THIRD_PARTY',
-                  metricName: 'consumption.e2e.voip.tts.characters',
-                  transactions: msg.messageText.length,
-                  apiKey
-                });
+              if (digits !== String(rawDigits).replace(/\s/g, '')) {
+                debug$3(`Sending DTMF from messageText (sanitized): "${rawDigits}" -> "${digits}"`);
+              } else {
+                debug$3(`Sending DTMF from messageText: ${digits}`);
               }
-              if (ttsResult && Buffer.isBuffer(ttsResult.buffer)) {
-                duration = parseFloat(ttsResult.duration) || 0;
-                const b64Buffer = ttsResult.buffer.toString('base64');
-                const request = JSON.stringify({
-                  METHOD: 'sendAudio',
-                  PESQ: false,
-                  sessionId: this.sessionId,
-                  b64_buffer: b64Buffer
-                });
-                if (this._lastBotSaysQueuedAt) {
-                  const latencyMs = Date.now() - this._lastBotSaysQueuedAt;
-                  const botText = this._lastBotSaysText ? this._lastBotSaysText.substring(0, 120) : '<unknown>';
-                  debug$3(`Latency (bot says -> sendAudio/TTS): ${latencyMs} ms (last bot: ${botText})`);
+              const request = JSON.stringify({
+                METHOD: 'sendDtmf',
+                digits,
+                sessionId: this.sessionId
+              });
+              stampAgentWire('dtmf', digits.length * DTMF_MS_PER_DIGIT, {
+                digitCount: digits.length
+              });
+              this._sendUserSaysWs(request);
+              return resolve();
+            }
+            if (!skipTtsForMixedInput) {
+              if (!this.axiosTtsParams) {
+                if (!(msg.media && msg.media.length > 0 && msg.media[0].buffer)) {
+                  return reject(new Error('TTS not configured, only audio input supported'));
                 }
-                msg.attachments.push({
-                  name: 'tts.wav',
-                  mimeType: 'audio/wav',
-                  base64: b64Buffer
-                });
-                stampAgentWire('tts', (duration || 0) * 1000, {
-                  ttsSynthMs,
-                  textLength: msg.messageText ? msg.messageText.length : 0
-                });
-                this.ws.send(request);
               } else {
-                return reject(new Error('TTS failed, response is empty'));
+                debug$3('UserSays routing: executing TTS branch');
+                const ttsRequest = this._buildTtsRequest(msg.messageText);
+                if (!ttsRequest) return reject(new Error('TTS not configured, only audio input supported'));
+                msg.sourceData = ttsRequest;
+                let ttsResult = null;
+                const ttsStartedAt = Date.now();
+                let ttsSynthMs = 0;
+                try {
+                  ttsResult = await this._getTtsAudio(ttsRequest, msg.messageText);
+                  ttsSynthMs = Date.now() - ttsStartedAt;
+                } catch (err) {
+                  return reject(new Error(`TTS "${msg.messageText}" failed - ${this._getAxiosErrOutput(err)}`));
+                }
+                if (msg && msg.messageText && msg.messageText.length > 0) {
+                  const apiKey = this._extractApiKey(this._getBody(Capabilities.VOIP_TTS_BODY));
+                  this.eventEmitter.emit('CONSUMPTION_METADATA', this.container, {
+                    type: lodash__default["default"].isNil(apiKey) ? 'INBUILT' : 'THIRD_PARTY',
+                    metricName: 'consumption.e2e.voip.tts.characters',
+                    transactions: msg.messageText.length,
+                    apiKey
+                  });
+                }
+                if (ttsResult && Buffer.isBuffer(ttsResult.buffer)) {
+                  duration = parseFloat(ttsResult.duration) || 0;
+                  const b64Buffer = ttsResult.buffer.toString('base64');
+                  const request = JSON.stringify({
+                    METHOD: 'sendAudio',
+                    PESQ: false,
+                    sessionId: this.sessionId,
+                    b64_buffer: b64Buffer
+                  });
+                  if (this._lastBotSaysQueuedAt) {
+                    const latencyMs = Date.now() - this._lastBotSaysQueuedAt;
+                    const botText = this._lastBotSaysText ? this._lastBotSaysText.substring(0, 120) : '<unknown>';
+                    debug$3(`Latency (bot says -> sendAudio/TTS): ${latencyMs} ms (last bot: ${botText})`);
+                  }
+                  msg.attachments.push({
+                    name: 'tts.wav',
+                    mimeType: 'audio/wav',
+                    base64: b64Buffer
+                  });
+                  stampAgentWire('tts', (duration || 0) * 1000, {
+                    ttsSynthMs,
+                    textLength: msg.messageText ? msg.messageText.length : 0
+                  });
+                  this._sendUserSaysWs(request);
+                } else {
+                  return reject(new Error('TTS failed, response is empty'));
+                }
               }
+            } else {
+              debug$3('UserSays routing: skipping TTS for mixed input because VOIP_USER_INPUT_PREFER_VOICE is enabled');
             }
-          } else {
-            debug$3('UserSays routing: skipping TTS for mixed input because VOIP_USER_INPUT_PREFER_VOICE is enabled');
-          }
-        }
-        if (msg && msg.media && msg.media.length > 0 && msg.media[0].buffer) {
-          debug$3('UserSays routing: executing MEDIA branch');
-          const request = JSON.stringify({
-            METHOD: 'sendAudio',
-            sessionId: this.sessionId,
-            b64_buffer: msg.media[0].buffer.toString('base64')
-          });
-          if (this._lastBotSaysQueuedAt) {
-            const latencyMs = Date.now() - this._lastBotSaysQueuedAt;
-            const botText = this._lastBotSaysText ? this._lastBotSaysText.substring(0, 120) : '<unknown>';
-            debug$3(`Latency (bot says -> sendAudio/MEDIA): ${latencyMs} ms (last bot: ${botText})`);
           }
-          // Stamp now; `requestedDurationMs` is backfilled once media metadata is parsed.
-          stampAgentWire('media', 0, {
-            mediaUri: msg.media[0].mediaUri || null
-          });
-          this.ws.send(request);
-          msg.attachments.push({
-            name: msg.media[0].mediaUri,
-            mimeType: msg.media[0].mimeType,
-            base64: msg.media[0].buffer.toString('base64')
-          });
-          try {
-            const metadata = await musicMetadata__default["default"].parseBuffer(msg.media[0].buffer);
-            if (metadata && metadata.format && metadata.format.duration) {
-              debug$3('Audio duration of user audio:', metadata.format.duration, 'seconds');
-              duration = Math.round(metadata.format.duration);
-              if (msg.voipAgent) {
-                msg.voipAgent.requestedDurationMs = Math.max(0, Math.round((duration || 0) * 1000));
+          if (msg && msg.media && msg.media.length > 0 && msg.media[0].buffer) {
+            debug$3('UserSays routing: executing MEDIA branch');
+            const request = JSON.stringify({
+              METHOD: 'sendAudio',
+              sessionId: this.sessionId,
+              b64_buffer: msg.media[0].buffer.toString('base64')
+            });
+            if (this._lastBotSaysQueuedAt) {
+              const latencyMs = Date.now() - this._lastBotSaysQueuedAt;
+              const botText = this._lastBotSaysText ? this._lastBotSaysText.substring(0, 120) : '<unknown>';
+              debug$3(`Latency (bot says -> sendAudio/MEDIA): ${latencyMs} ms (last bot: ${botText})`);
+            }
+            // Stamp now; `requestedDurationMs` is backfilled once media metadata is parsed.
+            stampAgentWire('media', 0, {
+              mediaUri: msg.media[0].mediaUri || null
+            });
+            this._sendUserSaysWs(request);
+            msg.attachments.push({
+              name: msg.media[0].mediaUri,
+              mimeType: msg.media[0].mimeType,
+              base64: msg.media[0].buffer.toString('base64')
+            });
+            try {
+              const metadata = await musicMetadata__default["default"].parseBuffer(msg.media[0].buffer);
+              if (metadata && metadata.format && metadata.format.duration) {
+                debug$3('Audio duration of user audio:', metadata.format.duration, 'seconds');
+                duration = Math.round(metadata.format.duration);
+                if (msg.voipAgent) {
+                  msg.voipAgent.requestedDurationMs = Math.max(0, Math.round((duration || 0) * 1000));
+                }
+              } else {
+                reject(new Error('Could not determine audio duration from metadata'));
               }
-            } else {
-              reject(new Error('Could not determine audio duration from metadata'));
+            } catch (err) {
+              reject(new Error(`Getting audio duration failed: ${err.message}`));
             }
-          } catch (err) {
-            reject(new Error(`Getting audio duration failed: ${err.message}`));
           }
+          const requestedDurationMs = Math.max(0, Math.round((duration || 0) * 1000));
+          if (requestedDurationMs <= 0) {
+            return resolve();
+          }
+          setTimeout(resolve, requestedDurationMs);
+        } catch (err) {
+          reject(err);
         }
-        const requestedDurationMs = Math.max(0, Math.round((duration || 0) * 1000));
-        if (requestedDurationMs <= 0) {
-          return resolve();
-        }
-        setTimeout(resolve, requestedDurationMs);
       }, 0);
     });
   }
+  _voipWsCanSend() {
+    return !this.stopCalled && this.ws && this.ws.readyState === ws__default["default"].OPEN;
+  }
+  _sendUserSaysWs(request) {
+    if (!this._voipWsCanSend()) {
+      throw new Error('VoIP session stopped');
+    }
+    this.ws.send(request);
+  }
   async Stop() {
     debug$3(`${this.sessionId} - Stop called`);
     this.stopCalled = true;
@@ -1487,6 +1467,75 @@ class BotiumConnectorVoip {
     if (lodash__default["default"].isNil(convoStep.logicHooks)) return null;
     return convoStep && convoStep.logicHooks && convoStep.logicHooks.find(lh => lh.name === 'VOIP_JOIN_SILENCE_DURATION');
   }
+  _normalizeJoinRulesBySubstring() {
+    const rawRules = this.caps[Capabilities.VOIP_JOIN_SILENCE_DURATION_BY_SUBSTRING];
+    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_JOIN_SILENCE_DURATION_BY_SUBSTRING} JSON: ${err.message || err}`);
+        return [];
+      }
+    }
+    if (!lodash__default["default"].isArray(parsedRules)) {
+      debug$3(`Invalid ${Capabilities.VOIP_JOIN_SILENCE_DURATION_BY_SUBSTRING}: expected array`);
+      return [];
+    }
+    return parsedRules.map(rule => {
+      if (!rule || typeof rule !== 'object') return null;
+      const substring = typeof rule.substring === 'string' ? rule.substring.trim() : '';
+      const timeoutMs = parseInt(rule.timeoutMs, 10);
+      if (!substring || !lodash__default["default"].isFinite(timeoutMs) || timeoutMs <= 0) return null;
+      return {
+        substring,
+        timeoutMs
+      };
+    }).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;
+    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 null;
+  }
+  _hasJoinLogicHookOrRule(convoStep) {
+    return !lodash__default["default"].isNil(this._getJoinLogicHook(convoStep)) || !lodash__default["default"].isNil(this._getJoinRuleBySubstring(convoStep, this.botMsgs));
+  }
+  _toJoinTimeoutMs(ms, isPsst) {
+    const parsed = parseInt(ms, 10);
+    if (!lodash__default["default"].isFinite(parsed) || parsed <= 0) return null;
+    return isPsst ? Math.max(0, parsed - 500) : parsed;
+  }
+  _getEffectiveJoinTimeoutMs(convoStep, botMsgs) {
+    const sttHandling = this.caps[Capabilities.VOIP_STT_MESSAGE_HANDLING];
+    const isPsst = sttHandling === 'PSST';
+    const joinLogicHook = this._getJoinLogicHook(convoStep);
+    if (joinLogicHook && joinLogicHook.args && joinLogicHook.args.length > 0) {
+      const joinHookTimeoutMs = this._toJoinTimeoutMs(joinLogicHook.args[0], isPsst);
+      if (lodash__default["default"].isFinite(joinHookTimeoutMs) && joinHookTimeoutMs > 0) return joinHookTimeoutMs;
+    }
+    const joinRule = this._getJoinRuleBySubstring(convoStep, botMsgs);
+    if (joinRule) {
+      const joinRuleTimeoutMs = this._toJoinTimeoutMs(joinRule.timeoutMs, isPsst);
+      if (lodash__default["default"].isFinite(joinRuleTimeoutMs) && joinRuleTimeoutMs > 0) return joinRuleTimeoutMs;
+    }
+    return this._toJoinTimeoutMs(this._getEffectiveMessageHandlingTimeout(), isPsst);
+  }
   _getIgnoreSilenceDurationAsserterLogicHook(convoStep) {
     if (lodash__default["default"].isNil(convoStep)) return null;
     if (lodash__default["default"].isNil(convoStep.logicHooks)) return null;
@@ -1572,6 +1621,12 @@ var botiumConnectorVoip = {
       type: 'boolean',
       required: false,
       advanced: true
+    }, {
+      name: 'VOIP_JOIN_SILENCE_DURATION_BY_SUBSTRING',
+      label: 'Join silence timeout rules by bot message substring',
+      type: 'json',
+      required: false,
+      advanced: true
     }]
   },
   PluginLogicHooks: {