@fonoster/apiserver 0.8.51 → 0.8.52

package/dist/voice/VoiceClientImpl.d.ts

@@ -32,8 +32,8 @@ declare class VoiceClientImpl implements VoiceClient {
     synthesize(text: string, options: SayOptions): Promise<string>;
     transcribe(): Promise<SpeechResult>;
     startSpeechGather(callback: (stream: {
-        speech?: string;
-        digit?: string;
+        speech: string;
+        responseTime: number;
     }) => void): Promise<void>;
     startDtmfGather(sessionRef: string, callback: (event: {
         digit: string;

package/dist/voice/handlers/createStreamGatherHandler.js

@@ -47,18 +47,20 @@ function createStreamGatherHandler(voiceClient) {
                 voiceClient.sendResponse({
                     streamGatherPayload: {
                         sessionRef,
-                        digit
+                        digit,
+                        responseTime: 0
                     }
                 });
             });
         }
         if (effectiveSource.includes(common_1.StreamGatherSource.SPEECH)) {
             voiceClient.startSpeechGather((event) => {
-                const { speech } = event;
+                const { speech, responseTime } = event;
                 voiceClient.sendResponse({
                     streamGatherPayload: {
                         sessionRef,
-                        speech
+                        speech,
+                        responseTime
                     }
                 });
             });

package/dist/voice/handlers/gather/createGatherHandler.js

@@ -60,7 +60,7 @@ function createGatherHandler(voiceClient) {
         const effectiveSource = source || common_1.GatherSource.SPEECH_AND_DTMF;
         const promises = [timeoutPromise];
         if (effectiveSource.includes(common_1.GatherSource.SPEECH)) {
-            promises.push(voiceClient.transcribe().then((result) => result.speech));
+            promises.push(voiceClient.transcribe().then((result) => result));
         }
         if (effectiveSource.includes(common_1.GatherSource.DTMF)) {
             promises.push(voiceClient
@@ -71,14 +71,15 @@ function createGatherHandler(voiceClient) {
                 timeout: effectiveTimeout,
                 onDigitReceived: timeoutPromise.cancelGlobalTimer
             })
-                .then(({ digits }) => digits));
+                .then((result) => result));
         }
-        const result = yield Promise.race(promises);
+        const result = (yield Promise.race(promises));
         voiceClient.sendResponse({
             gatherResponse: {
                 sessionRef,
-                speech: (0, utils_1.isDtmf)(result) ? null : result,
-                digits: (0, utils_1.isDtmf)(result) ? result : null
+                responseTime: result.responseTime,
+                speech: (0, utils_1.isDtmf)(result.digits) ? undefined : result.speech,
+                digits: (0, utils_1.isDtmf)(result.digits) ? result.digits : undefined
             }
         });
     }));

package/dist/voice/handlers/gather/getTimeoutPromise.d.ts

@@ -2,7 +2,7 @@ interface PromiseWithResetTimer<T> extends Promise<T> {
     cancelGlobalTimer?: () => void;
 }
 declare function getTimeoutPromise(timeout: number): {
-    timeoutPromise: PromiseWithResetTimer<string>;
+    timeoutPromise: PromiseWithResetTimer<unknown>;
     effectiveTimeout: number;
 };
 export { getTimeoutPromise };

package/dist/voice/stt/Deepgram.js

@@ -90,8 +90,15 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
                 if (!data.channel.alternatives[0].transcript || !data.speech_final) {
                     return;
                 }
+                const words = data.channel.alternatives[0].words;
+                const responseTime = words.length > 0
+                    ? (words.reduce((acc, word) => acc + (word.end - word.start), 0) *
+                        1000) /
+                        words.length
+                    : 0;
                 out.emit("data", {
-                    speech: data.channel.alternatives[0].transcript
+                    speech: data.channel.alternatives[0].transcript,
+                    responseTime
                 });
             });
             connection.on(LiveTranscriptionEvents.Error, (err) => {
@@ -104,6 +111,7 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
     transcribe(stream) {
         return __awaiter(this, void 0, void 0, function* () {
             return new Promise((resolve, reject) => {
+                const startTime = performance.now();
                 const connection = this.client.listen.live(buildTranscribeConfig(this.engineConfig.config));
                 stream.on("data", (chunk) => {
                     connection.send(chunk);
@@ -113,7 +121,8 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
                         if (data.channel.alternatives[0].transcript && data.speech_final) {
                             const result = {
                                 speech: data.channel.alternatives[0].transcript,
-                                isFinal: true
+                                isFinal: true,
+                                responseTime: performance.now() - startTime
                             };
                             resolve(result);
                             connection.destroy();

package/dist/voice/stt/Google.js

@@ -47,6 +47,7 @@ const common_1 = require("@fonoster/common");
 const speech_1 = require("@google-cloud/speech");
 const z = __importStar(require("zod"));
 const AbstractSpeechToText_1 = require("./AbstractSpeechToText");
+const perf_hooks_1 = require("perf_hooks");
 const ENGINE_NAME = "stt.google";
 exports.ENGINE_NAME = ENGINE_NAME;
 class Google extends AbstractSpeechToText_1.AbstractSpeechToText {
@@ -63,6 +64,7 @@ class Google extends AbstractSpeechToText_1.AbstractSpeechToText {
     }
     transcribe(stream) {
         return __awaiter(this, void 0, void 0, function* () {
+            const startTime = perf_hooks_1.performance.now();
             const languageCode = this.engineConfig.config.languageCode || common_1.VoiceLanguage.EN_US;
             const audioConfig = {
                 interimResults: false,
@@ -78,15 +80,17 @@ class Google extends AbstractSpeechToText_1.AbstractSpeechToText {
                     .on("error", (e) => reject(e))
                     .on("data", (data) => {
                     var _a;
+                    const responseTime = perf_hooks_1.performance.now() - startTime;
                     if ((_a = data.results[0]) === null || _a === void 0 ? void 0 : _a.alternatives[0]) {
                         const result = {
                             speech: data.results[0].alternatives[0].transcript,
-                            isFinal: true
+                            isFinal: true,
+                            responseTime
                         };
                         resolve(result);
                     }
                     else {
-                        resolve({ speech: "", isFinal: true });
+                        resolve({ speech: "", isFinal: true, responseTime });
                     }
                     recognizeStream.destroy();
                 });

package/dist/voice/stt/types.d.ts

@@ -2,6 +2,7 @@ import { VoiceLanguage } from "@fonoster/common";
 type SpeechResult = {
     speech: string;
     isFinal: boolean;
+    responseTime: number;
 };
 type StreamSpeech = {
     on(events: string, callback: (result: SpeechResult) => void): void;

package/dist/voice/tts/ElevenLabs.js

@@ -125,7 +125,9 @@ class ElevenLabs extends AbstractTextToSpeech_1.AbstractTextToSpeech {
                 voice,
                 text,
                 model_id: model !== null && model !== void 0 ? model : "eleven_flash_v2_5",
-                output_format: "pcm_16000"
+                output_format: "pcm_16000",
+                // TODO: Make this configurable
+                optimize_streaming_latency: 2
             });
             return (yield (0, streamToBuffer_1.streamToBuffer)(response));
         });

package/dist/voice/types/voice.d.ts

@@ -11,7 +11,8 @@ type VoiceClient = {
     synthesize: (text: string, options: SayOptions) => Promise<string>;
     transcribe: () => Promise<SpeechResult>;
     startSpeechGather: (callback: (stream: {
-        speech?: string;
+        speech: string;
+        responseTime: number;
     }) => void) => void;
     startDtmfGather: (sessionRef: string, callback: (event: {
         digit: string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@fonoster/apiserver",
-  "version": "0.8.51",
+  "version": "0.8.52",
   "description": "APIServer for Fonoster",
   "author": "Pedro Sanders <psanders@fonoster.com>",
   "homepage": "https://github.com/fonoster/fonoster#readme",
@@ -21,11 +21,11 @@
   },
   "dependencies": {
     "@deepgram/sdk": "^3.5.1",
-    "@fonoster/authz": "^0.8.51",
-    "@fonoster/common": "^0.8.51",
-    "@fonoster/identity": "^0.8.51",
+    "@fonoster/authz": "^0.8.52",
+    "@fonoster/common": "^0.8.52",
+    "@fonoster/identity": "^0.8.52",
     "@fonoster/logger": "^0.8.51",
-    "@fonoster/sipnet": "^0.8.51",
+    "@fonoster/sipnet": "^0.8.52",
     "@fonoster/streams": "^0.8.51",
     "@fonoster/types": "^0.8.51",
     "@google-cloud/speech": "^6.6.0",
@@ -73,5 +73,5 @@
     "@types/uuid": "^10.0.0",
     "@types/validator": "^13.12.0"
   },
-  "gitHead": "67a0f500926304550c2698943ba6e7a7aa4b9f06"
+  "gitHead": "772398ea115078012f8e270943ec36b61343ea36"
 }