fix to add custom stt endpoint to session speech credentials object

allow azure stt custom speech endpoint id to be passed as property in recognizer
gather/transcribe: support for azure custom speech models (endpoint id)
2026-02-11 17:01:30 +00:00 · 2022-10-06 07:57:14 +01:00 · 2022-10-05 21:57:14 +01:00 · 2022-10-04 22:17:28 +01:00 · 2022-10-04 12:59:58 +01:00 · 2022-10-02 22:36:27 +01:00
13 changed files with 60 additions and 29 deletions
--- a/lib/session/call-info.js
+++ b/lib/session/call-info.js
@@ -11,6 +11,7 @@ class CallInfo {
    let srf;
    this.direction = opts.direction;
    this.traceId = opts.traceId;
+    this.callTerminationBy = undefined;
    if (opts.req) {
      const u = opts.req.getParsedHeader('from');
      const uri = parseUri(u.uri);
@@ -119,7 +120,7 @@ class CallInfo {
      applicationSid: this.applicationSid,
      fsSipAddress: this.localSipAddress
    };
-    ['parentCallSid', 'originatingSipIp', 'originatingSipTrunkName'].forEach((prop) => {
+    ['parentCallSid', 'originatingSipIp', 'originatingSipTrunkName', 'callTerminationBy'].forEach((prop) => {
      if (this[prop]) obj[prop] = this[prop];
    });
    if (typeof this.duration === 'number') obj.duration = this.duration;
--- a/lib/session/call-session.js
+++ b/lib/session/call-session.js
@@ -549,7 +549,9 @@ class CallSession extends Emitter {
          return {
            speech_credential_sid: credential.speech_credential_sid,
            api_key: credential.api_key,
-            region: credential.region
+            region: credential.region,
+            use_custom_stt: credential.use_custom_stt,
+            custom_stt_endpoint: credential.custom_stt_endpoint
          };
        }
        else if ('wellsaid' === vendor) {
@@ -638,6 +640,7 @@ class CallSession extends Emitter {
    this._onTasksDone();
    this._clearResources();

+
    if (!this.isConfirmCallSession && !this.isSmsCallSession) sessionTracker.remove(this.callSid);
  }

@@ -1188,6 +1191,7 @@ class CallSession extends Emitter {
    } catch (err) {
      if (err === CALLER_CANCELLED_ERR_MSG) {
        this.logger.error(err, 'caller canceled quickly before we could respond, ending call');
+        this.callInfo.callTerminationBy = 'caller';
        this._notifyCallStatusChange({
          callStatus: CallStatus.NoAnswer,
          sipStatus: 487,
@@ -1302,7 +1306,8 @@ class CallSession extends Emitter {
      this.dlg = await this.srf.createUAS(this.req, this.res, {
        headers: {
          'X-Trace-ID': this.req.locals.traceId,
-          'X-Call-Sid': this.req.locals.callSid
+          'X-Call-Sid': this.req.locals.callSid,
+          ...(this.applicationSid && {'X-Application-Sid': this.applicationSid})
        },
        localSdp: this.ep.local.sdp
      });
@@ -1494,8 +1499,9 @@ class CallSession extends Emitter {
        dlg.connected = false;
        dlg.destroy = origDestroy;
        const duration = moment().diff(this.dlg.connectTime, 'seconds');
+        this.callInfo.callTerminationBy = 'jambonz';
        this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});
-        this.logger.debug('CallSession: call terminated by jambones');
+        this.logger.debug('CallSession: call terminated by jambonz');
        this.rootSpan.setAttributes({'call.termination': 'hangup by jambonz'});
        origDestroy().catch((err) => this.logger.info({err}, 'CallSession - error destroying dialog'));
        if (this.wakeupResolver) {
--- a/lib/session/inbound-call-session.js
+++ b/lib/session/inbound-call-session.js
@@ -34,6 +34,7 @@ class InboundCallSession extends CallSession {

  _onCancel() {
    this.rootSpan.setAttributes({'call.termination': 'caller abandoned'});
+    this.callInfo.callTerminationBy = 'caller';
    this._notifyCallStatusChange({
      callStatus: CallStatus.NoAnswer,
      sipStatus: 487,
@@ -69,6 +70,7 @@ class InboundCallSession extends CallSession {
    assert(this.dlg.connectTime);
    const duration = moment().diff(this.dlg.connectTime, 'seconds');
    this.rootSpan.setAttributes({'call.termination': 'hangup by caller'});
+    this.callInfo.callTerminationBy = 'caller';
    this.emit('callStatusChange', {
      callStatus: CallStatus.Completed,
      duration
--- a/lib/session/rest-call-session.js
+++ b/lib/session/rest-call-session.js
@@ -44,6 +44,7 @@ class RestCallSession extends CallSession {
   * This is invoked when the called party hangs up, in order to calculate the call duration.
   */
  _callerHungup() {
+    this.callInfo.callTerminationBy = 'caller';
    const duration = moment().diff(this.dlg.connectTime, 'seconds');
    this.emit('callStatusChange', {callStatus: CallStatus.Completed, duration});
    this.logger.debug('RestCallSession: called party hung up');
--- a/lib/session/siprec-call-session.js
+++ b/lib/session/siprec-call-session.js
@@ -36,7 +36,8 @@ class SipRecCallSession extends InboundCallSession {
        headers: {
          'Content-Type': 'application/sdp',
          'X-Trace-ID': this.req.locals.traceId,
-          'X-Call-Sid': this.req.locals.callSid
+          'X-Call-Sid': this.req.locals.callSid,
+          ...(this.applicationSid && {'X-Application-Sid': this.applicationSid})
        },
        localSdp: combinedSdp
      });
--- a/lib/tasks/gather.js
+++ b/lib/tasks/gather.js
@@ -82,6 +82,7 @@ class TaskGather extends Task {
      this.requestSnr = recognizer.requestSnr || false;
      this.initialSpeechTimeoutMs = recognizer.initialSpeechTimeoutMs || 0;
      this.azureServiceEndpoint = recognizer.azureServiceEndpoint;
+      this.azureSttEndpointId = recognizer.azureSttEndpointId;
    }
    else {
      this.hints = [];
@@ -301,7 +302,7 @@ class TaskGather extends Task {
    }

    if ('google' === this.vendor) {
-      this.bugname = 'google_trancribe';
+      this.bugname = 'google_transcribe';
      if (this.sttCredentials) opts.GOOGLE_APPLICATION_CREDENTIALS = JSON.stringify(this.sttCredentials.credentials);
      [
        ['enhancedModel', 'GOOGLE_SPEECH_USE_ENHANCED'],
@@ -339,7 +340,7 @@ class TaskGather extends Task {
      ep.addCustomEventListener(GoogleTranscriptionEvents.VadDetected, this._onVadDetected.bind(this, cs, ep));
    }
    else if (['aws', 'polly'].includes(this.vendor)) {
-      this.bugname = 'aws_trancribe';
+      this.bugname = 'aws_transcribe';
      if (this.vocabularyName) opts.AWS_VOCABULARY_NAME = this.vocabularyName;
      if (this.vocabularyFilterName) {
        opts.AWS_VOCABULARY_NAME = this.vocabularyFilterName;
@@ -358,10 +359,18 @@ class TaskGather extends Task {
    else if ('microsoft' === this.vendor) {
      this.bugname = 'azure_transcribe';
      if (this.sttCredentials) {
+        const {api_key, region, use_custom_stt, custom_stt_endpoint} = this.sttCredentials;
+
        Object.assign(opts, {
-          'AZURE_SUBSCRIPTION_KEY': this.sttCredentials.api_key,
-          'AZURE_REGION': this.sttCredentials.region
+          'AZURE_SUBSCRIPTION_KEY': api_key,
+          'AZURE_REGION': region
        });
+        if (this.azureSttEndpointId) {
+          Object.assign(opts, {'AZURE_SERVICE_ENDPOINT_ID': this.azureSttEndpointId});
+        }
+        else if (use_custom_stt && custom_stt_endpoint) {
+          Object.assign(opts, {'AZURE_SERVICE_ENDPOINT_ID': custom_stt_endpoint});
+        }
      }
      if (this.hints && this.hints.length > 0) {
        opts.AZURE_SPEECH_HINTS = this.hints.map((h) => h.trim()).join(',');
@@ -641,7 +650,7 @@ class TaskGather extends Task {
    }

    this.span.setAttributes({'stt.resolve': reason, 'stt.result': JSON.stringify(evt)});
-    if (this.ep && this.ep.connected) {
+    if (this.needsStt && this.ep && this.ep.connected) {
      this.ep.stopTranscription({vendor: this.vendor})
        .catch((err) => this.logger.error({err}, 'Error stopping transcription'));
    }
--- a/lib/tasks/specs.json
+++ b/lib/tasks/specs.json
@@ -497,6 +497,7 @@
      "requestSnr": "boolean",
      "initialSpeechTimeoutMs": "number",
      "azureServiceEndpoint": "string",
+      "azureSttEndpointId": "string",
      "asrDtmfTerminationDigit": "string",
      "asrTimeout": "number"
    },
--- a/lib/tasks/transcribe.js
+++ b/lib/tasks/transcribe.js
@@ -54,6 +54,7 @@ class TaskTranscribe extends Task {
    this.requestSnr = recognizer.requestSnr || false;
    this.initialSpeechTimeoutMs = recognizer.initialSpeechTimeoutMs || 0;
    this.azureServiceEndpoint = recognizer.azureServiceEndpoint;
+    this.azureSttEndpointId = recognizer.azureSttEndpointId;
  }

  get name() { return TaskName.Transcribe; }
@@ -161,7 +162,7 @@ class TaskTranscribe extends Task {
    ep.addCustomEventListener(AzureTranscriptionEvents.NoSpeechDetected, this._onNoAudio.bind(this, cs, ep, channel));

    if (this.vendor === 'google') {
-      this.bugname = 'google_trancribe';
+      this.bugname = 'google_transcribe';
      if (this.sttCredentials) opts.GOOGLE_APPLICATION_CREDENTIALS = JSON.stringify(this.sttCredentials.credentials);
      [
        ['enhancedModel', 'GOOGLE_SPEECH_USE_ENHANCED'],
@@ -199,7 +200,7 @@ class TaskTranscribe extends Task {
        .catch((err) => this.logger.info(err, 'TaskTranscribe:_startTranscribing with google'));
    }
    else if (this.vendor === 'aws') {
-      this.bugname = 'aws_trancribe';
+      this.bugname = 'aws_transcribe';
      [
        ['diarization', 'AWS_SHOW_SPEAKER_LABEL'],
        ['identifyChannels', 'AWS_ENABLE_CHANNEL_IDENTIFICATION']
@@ -231,11 +232,18 @@ class TaskTranscribe extends Task {
        .catch((err) => this.logger.info(err, 'TaskTranscribe:_startTranscribing with aws'));
    }
    else if (this.vendor === 'microsoft') {
-      this.bugname = 'azure_trancribe';
+      this.bugname = 'azure_transcribe';
+      const {api_key, region, use_custom_stt, custom_stt_endpoint} = this.sttCredentials;
      Object.assign(opts, {
-        'AZURE_SUBSCRIPTION_KEY': this.sttCredentials.api_key,
-        'AZURE_REGION': this.sttCredentials.region
+        'AZURE_SUBSCRIPTION_KEY': api_key,
+        'AZURE_REGION': region
      });
+      if (this.azureSttEndpointId) {
+        Object.assign(opts, {'AZURE_SERVICE_ENDPOINT_ID': this.azureSttEndpointId});
+      }
+      else if (use_custom_stt && custom_stt_endpoint) {
+        Object.assign(opts, {'AZURE_SERVICE_ENDPOINT_ID': custom_stt_endpoint});
+      }
      if (this.hints && this.hints.length > 0) {
        opts.AZURE_SPEECH_HINTS = this.hints.map((h) => h.trim()).join(',');
      }
--- a/lib/utils/db-utils.js
+++ b/lib/utils/db-utils.js
@@ -36,6 +36,8 @@ const speechMapper = (cred) => {
    const o = JSON.parse(decrypt(credential));
    obj.api_key = o.api_key;
    obj.region = o.region;
+    obj.use_custom_stt = o.use_custom_stt;
+    obj.custom_stt_endpoint = o.custom_stt_endpoint;
  }
  else if ('wellsaid' === obj.vendor) {
    const o = JSON.parse(decrypt(credential));
--- a/lib/utils/place-outdial.js
+++ b/lib/utils/place-outdial.js
@@ -67,7 +67,8 @@ class SingleDialer extends Emitter {
      ...opts.headers,
      ...(this.target.headers || {}),
      'X-Jambonz-Routing': this.target.type,
-      'X-Call-Sid': this.callSid
+      'X-Call-Sid': this.callSid,
+      ...(this.applicationSid && {'X-Application-Sid': this.applicationSid})
    };
    if (srf.locals.fsUUID) {
      opts.headers = {
--- a/lib/utils/ws-requestor.js
+++ b/lib/utils/ws-requestor.js
@@ -138,7 +138,7 @@ class WsRequestor extends BaseRequestor {
        success: (response) => {
          clearTimeout(timer);
          const rtt = this._roundTrip(startAt);
-          this.logger.info({response}, `WsRequestor:request ${url} succeeded in ${rtt}ms`);
+          this.logger.debug({response}, `WsRequestor:request ${url} succeeded in ${rtt}ms`);
          this.stats.histogram('app.hook.ws_response_time', rtt, ['hook_type:app']);
          resolve(response);
        },
@@ -286,7 +286,7 @@ class WsRequestor extends BaseRequestor {
      const obj = JSON.parse(content);
      const {type, msgid, command, call_sid = this.call_sid, queueCommand = false, data} = obj;

-      this.logger.debug({obj}, 'WsRequestor:request websocket: received');
+      //this.logger.debug({obj}, 'WsRequestor:request websocket: received');
      assert.ok(type, 'type property not supplied');

      switch (type) {
@@ -323,7 +323,7 @@ class WsRequestor extends BaseRequestor {

  _recvCommand(msgid, command, call_sid, queueCommand, data) {
    // TODO: validate command
-    this.logger.info({msgid, command, call_sid, queueCommand, data}, 'received command');
+    this.logger.debug({msgid, command, call_sid, queueCommand, data}, 'received command');
    this.emit('command', {msgid, command, call_sid, queueCommand, data});
  }
 }
--- a/package-lock.json
+++ b/package-lock.json
@@ -11,7 +11,7 @@
      "dependencies": {
        "@jambonz/db-helpers": "^0.6.18",
        "@jambonz/http-health-check": "^0.0.1",
-        "@jambonz/realtimedb-helpers": "^0.4.30",
+        "@jambonz/realtimedb-helpers": "^0.4.32",
        "@jambonz/stats-collector": "^0.1.6",
        "@jambonz/time-series": "^0.2.1",
        "@opentelemetry/api": "^1.1.0",
@@ -543,9 +543,9 @@
      }
    },
    "node_modules/@jambonz/realtimedb-helpers": {
-      "version": "0.4.30",
-      "resolved": "https://registry.npmjs.org/@jambonz/realtimedb-helpers/-/realtimedb-helpers-0.4.30.tgz",
-      "integrity": "sha512-bF1crcyz/sYDCnOX0odbI1KLZUo6TADc548MHX5/O2+gY6Yi6u9GdEaf2u5q5fgmpoJ+Qrf3fuiENxzro+Pq8g==",
+      "version": "0.4.32",
+      "resolved": "https://registry.npmjs.org/@jambonz/realtimedb-helpers/-/realtimedb-helpers-0.4.32.tgz",
+      "integrity": "sha512-m08rqFEg6I43JVwjiyt0ahvljxJ3YvL3xsKHMoSJw48hdy/8V4w9ovYnUNaZGhaerhJVw36fCfTk4YNlByTnVw==",
      "dependencies": {
        "@google-cloud/text-to-speech": "^3.4.0",
        "@jambonz/promisify-redis": "^0.0.6",
@@ -6669,9 +6669,9 @@
      }
    },
    "@jambonz/realtimedb-helpers": {
-      "version": "0.4.30",
-      "resolved": "https://registry.npmjs.org/@jambonz/realtimedb-helpers/-/realtimedb-helpers-0.4.30.tgz",
-      "integrity": "sha512-bF1crcyz/sYDCnOX0odbI1KLZUo6TADc548MHX5/O2+gY6Yi6u9GdEaf2u5q5fgmpoJ+Qrf3fuiENxzro+Pq8g==",
+      "version": "0.4.32",
+      "resolved": "https://registry.npmjs.org/@jambonz/realtimedb-helpers/-/realtimedb-helpers-0.4.32.tgz",
+      "integrity": "sha512-m08rqFEg6I43JVwjiyt0ahvljxJ3YvL3xsKHMoSJw48hdy/8V4w9ovYnUNaZGhaerhJVw36fCfTk4YNlByTnVw==",
      "requires": {
        "@google-cloud/text-to-speech": "^3.4.0",
        "@jambonz/promisify-redis": "^0.0.6",
--- a/package.json
+++ b/package.json
@@ -16,8 +16,7 @@
    "type": "git",
    "url": "https://github.com/jambonz/jambonz-feature-server.git"
  },
-  "bugs": {
-  },
+  "bugs": {},
  "scripts": {
    "start": "node app",
    "test": "NODE_ENV=test JAMBONES_HOSTING=1 HTTP_POOL=1 DRACHTIO_HOST=127.0.0.1 DRACHTIO_PORT=9060 DRACHTIO_SECRET=cymru JAMBONES_MYSQL_HOST=127.0.0.1 JAMBONES_MYSQL_PORT=3360 JAMBONES_MYSQL_USER=jambones_test JAMBONES_MYSQL_PASSWORD=jambones_test JAMBONES_MYSQL_DATABASE=jambones_test JAMBONES_REDIS_HOST=127.0.0.1 JAMBONES_REDIS_PORT=16379 JAMBONES_LOGLEVEL=error ENABLE_METRICS=0 HTTP_PORT=3000 JAMBONES_SBCS=172.38.0.10 JAMBONES_FREESWITCH=127.0.0.1:8022:ClueCon:docker-host JAMBONES_TIME_SERIES_HOST=127.0.0.1 JAMBONES_NETWORK_CIDR=172.38.0.0/16 node test/ ",
@@ -27,7 +26,7 @@
  "dependencies": {
    "@jambonz/db-helpers": "^0.6.18",
    "@jambonz/http-health-check": "^0.0.1",
-    "@jambonz/realtimedb-helpers": "^0.4.30",
+    "@jambonz/realtimedb-helpers": "^0.4.32",
    "@jambonz/stats-collector": "^0.1.6",
    "@jambonz/time-series": "^0.2.1",
    "@opentelemetry/api": "^1.1.0",
Author	SHA1	Message	Date
Dave Horton	2a7746537d	fix to add custom stt endpoint to session speech credentials object	2022-10-06 07:57:14 +01:00
Dave Horton	b2cac47d05	allow azure stt custom speech endpoint id to be passed as property in recognizer	2022-10-05 21:57:14 +01:00
Dave Horton	8a2b2962f8	gather/transcribe: support for azure custom speech models (endpoint id)	2022-10-04 22:17:28 +01:00
Dave Horton	90cb5e1348	bugfix: typo in bugname was causing transcripts to be ignored	2022-10-04 12:59:58 +01:00
Dave Horton	cf821569b3	minor logging changes	2022-10-02 22:36:27 +01:00
Dave Horton	218f2d6c67	bugfix: unnecessary call to stopTranscription in gather verb when only collecting digits	2022-09-30 10:27:33 +01:00
Joan	c2c8f00978	added call_termination_by on app call status (#168 ) Co-authored-by: Joan Salvatella <joan@bookline.io>	2022-09-23 09:13:55 +02:00
Dave Horton	32714d73f3	update to synthAudio with bugfix for writing TTS rtt stats for microsoft	2022-09-21 15:22:19 +02:00
Dave Horton	8da85ebd5a	include custom header X-Application-Sid to make it available to cdrs	2022-09-20 13:54:54 +02:00