@fonoster/apiserver 0.16.9 → 0.16.11

package/LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2024 Fonoster Inc
+Copyright (c) 2026 Fonoster Inc
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal
@@ -18,4 +18,4 @@ FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-SOFTWARE.
+SOFTWARE.

package/dist/applications/createListApplications.js

@@ -56,7 +56,7 @@ function createListApplications(prisma) {
         const items = result.map(applicationWithEncodedStruct_1.applicationWithEncodedStruct);
         callback(null, {
             items,
-            nextPageToken: (_a = result[result.length - 1]) === null || _a === void 0 ? void 0 : _a.ref
+            nextPageToken: items.length < pageSize ? undefined : (_a = result[result.length - 1]) === null || _a === void 0 ? void 0 : _a.ref
         });
     });
     return (0, common_1.withErrorHandlingAndValidation)(listApplications, common_1.Validators.listRequestSchema);

package/dist/secrets/listSecrets.js

@@ -49,7 +49,7 @@ function listSecrets(prisma) {
         })).map(common_1.datesMapper);
         callback(null, {
             items: result,
-            nextPageToken: (_a = result[result.length - 1]) === null || _a === void 0 ? void 0 : _a.ref
+            nextPageToken: result.length < pageSize ? undefined : (_a = result[result.length - 1]) === null || _a === void 0 ? void 0 : _a.ref
         });
     });
     return (0, common_1.withErrorHandlingAndValidation)(fn, common_1.Validators.listRequestSchema);

package/dist/voice/stt/Deepgram.js

@@ -83,6 +83,16 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
     streamTranscribe(stream) {
         const connection = this.client.listen.live(buildTranscribeConfig(this.engineConfig.config));
         const out = new stream_1.Stream();
+        // Track last transcript for UtteranceEnd fallback
+        // According to Deepgram docs: "If you receive an UtteranceEnd event without a
+        // preceding speech_final: true, it's advisable to process the last-received
+        // transcript as a complete utterance."
+        // UtteranceEnd fires after finalized words, so we store the last finalized transcript
+        // but also keep any transcript as a fallback
+        let lastFinalizedTranscript = null;
+        let lastFinalizedTranscriptTime = 0;
+        let lastAnyTranscript = null;
+        let lastAnyTranscriptTime = 0;
         // Add error handler immediately to catch any connection errors
         connection.on(LiveTranscriptionEvents.Error, (err) => {
             logger.error("error on Deepgram connection", { err });
@@ -106,8 +116,23 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
             });
             connection.on(LiveTranscriptionEvents.Transcript, (data) => {
                 var _a, _b, _c;
-                if (!((_c = (_b = (_a = data.channel) === null || _a === void 0 ? void 0 : _a.alternatives) === null || _b === void 0 ? void 0 : _b[0]) === null || _c === void 0 ? void 0 : _c.transcript) ||
-                    !data.speech_final) {
+                const transcript = (_c = (_b = (_a = data.channel) === null || _a === void 0 ? void 0 : _a.alternatives) === null || _b === void 0 ? void 0 : _b[0]) === null || _c === void 0 ? void 0 : _c.transcript;
+                const hasTranscript = !!transcript;
+                const isFinal = data.is_final === true;
+                const speechFinal = data.speech_final === true;
+                // Store any transcript for UtteranceEnd fallback
+                if (hasTranscript) {
+                    lastAnyTranscript = transcript;
+                    lastAnyTranscriptTime = Date.now();
+                    // Store finalized transcripts separately (preferred for UtteranceEnd)
+                    if (isFinal || speechFinal) {
+                        lastFinalizedTranscript = transcript;
+                        lastFinalizedTranscriptTime = Date.now();
+                    }
+                }
+                // Process transcript if it has content and is final
+                // Check both speech_final (primary) and is_final (backup)
+                if (!hasTranscript || (!speechFinal && !isFinal)) {
                     return;
                 }
                 const words = data.channel.alternatives[0].words || [];
@@ -117,13 +142,55 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
                         words.length
                     : 0;
                 logger.verbose("transcribe result", {
-                    speech: data.channel.alternatives[0].transcript,
-                    responseTime
+                    speech: transcript,
+                    responseTime,
+                    isFinal,
+                    speechFinal
                 });
                 out.emit("data", {
-                    speech: data.channel.alternatives[0].transcript,
+                    speech: transcript,
                     responseTime
                 });
+                // Clear transcripts after processing (they've been emitted)
+                lastFinalizedTranscript = null;
+                lastAnyTranscript = null;
+            });
+            // CRITICAL: Handle UtteranceEnd events (fallback when speech_final never becomes true)
+            // This is Deepgram's recommended fallback mechanism for noisy environments
+            // UtteranceEnd requires: interim_results=true and utterance_end_ms parameter
+            // UtteranceEnd fires after finalized words, so prefer lastFinalizedTranscript
+            connection.on(LiveTranscriptionEvents.UtteranceEnd, (data) => {
+                // Prefer finalized transcript, fall back to any transcript
+                const transcriptToUse = lastFinalizedTranscript || lastAnyTranscript;
+                const transcriptTime = lastFinalizedTranscript
+                    ? lastFinalizedTranscriptTime
+                    : lastAnyTranscriptTime;
+                if (transcriptToUse) {
+                    // Use last_word_end from UtteranceEnd event if available for more accurate timing
+                    // Otherwise fall back to time since last transcript
+                    const lastWordEnd = data === null || data === void 0 ? void 0 : data.last_word_end;
+                    const responseTime = lastWordEnd
+                        ? lastWordEnd * 1000 // Convert seconds to milliseconds
+                        : transcriptTime
+                            ? Date.now() - transcriptTime
+                            : 0;
+                    logger.info("Deepgram UtteranceEnd - processing last transcript", {
+                        speech: transcriptToUse,
+                        responseTime,
+                        lastWordEnd: lastWordEnd,
+                        wasFinalized: !!lastFinalizedTranscript
+                    });
+                    out.emit("data", {
+                        speech: transcriptToUse,
+                        responseTime
+                    });
+                    // Clear transcripts after processing
+                    lastFinalizedTranscript = null;
+                    lastAnyTranscript = null;
+                }
+                else {
+                    logger.warn("Deepgram UtteranceEnd received but no last transcript available");
+                }
             });
         });
         // Handle stream errors and cleanup
@@ -212,7 +279,9 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
                 .optional(),
             model: z
                 .nativeEnum(types_1.DeepgramModel, { message: "Invalid Deepgram model" })
-                .optional()
+                .optional(),
+            interimResults: z.boolean().optional(),
+            utteranceEndMs: z.number().int().min(1000).max(5000).optional()
         });
     }
     static getCredentialsValidationSchema() {
@@ -223,7 +292,17 @@ class Deepgram extends AbstractSpeechToText_1.AbstractSpeechToText {
 }
 exports.Deepgram = Deepgram;
 function buildTranscribeConfig(config) {
-    return Object.assign(Object.assign({}, config), { model: config.model || types_1.DeepgramModel.NOVA_2_PHONECALL, language: config.languageCode || common_1.VoiceLanguage.EN_US, encoding: "linear16", sample_rate: 16000, channels: 1, smart_format: config.smartFormat || true, 
+    // UtteranceEnd requires interim_results to be enabled
+    // Default to true to enable UtteranceEnd fallback mechanism
+    const interimResults = config.interimResults !== false;
+    // Default utterance_end_ms to 1000ms (minimum required value)
+    // This enables UtteranceEnd events as a fallback when speech_final never becomes true
+    const utteranceEndMs = config.utteranceEndMs || 1000;
+    return Object.assign(Object.assign({}, config), { model: config.model || types_1.DeepgramModel.NOVA_2_PHONECALL, language: config.languageCode || common_1.VoiceLanguage.EN_US, encoding: "linear16", sample_rate: 16000, channels: 1, smart_format: config.smartFormat !== false, 
         // This needs to be set to true to avoid delays while using smart_format
-        no_delay: config.noDelay || true });
+        no_delay: config.noDelay !== false, 
+        // REQUIRED for UtteranceEnd: interim_results must be true
+        interim_results: interimResults, 
+        // REQUIRED for UtteranceEnd: utterance_end_ms parameter
+        utterance_end_ms: utteranceEndMs });
 }

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

@@ -51,6 +51,8 @@ type DeepgramSttConfig = {
         model: DeepgramModel;
         smartFormat: boolean;
         noDelay: boolean;
+        interimResults?: boolean;
+        utteranceEndMs?: number;
     };
     credentials: {
         apiKey: string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@fonoster/apiserver",
-  "version": "0.16.9",
+  "version": "0.16.11",
   "description": "APIServer for Fonoster",
   "author": "Pedro Sanders <psanders@fonoster.com>",
   "homepage": "https://github.com/fonoster/fonoster#readme",
@@ -22,12 +22,12 @@
   },
   "dependencies": {
     "@deepgram/sdk": "^3.5.1",
-    "@fonoster/authz": "^0.16.8",
-    "@fonoster/autopilot": "^0.16.9",
+    "@fonoster/authz": "^0.16.10",
+    "@fonoster/autopilot": "^0.16.11",
     "@fonoster/common": "^0.16.8",
-    "@fonoster/identity": "^0.16.8",
+    "@fonoster/identity": "^0.16.10",
     "@fonoster/logger": "^0.16.7",
-    "@fonoster/sipnet": "^0.16.8",
+    "@fonoster/sipnet": "^0.16.10",
     "@fonoster/streams": "^0.16.7",
     "@fonoster/types": "^0.16.7",
     "@google-cloud/speech": "^6.6.0",
@@ -76,5 +76,5 @@
     "@types/uuid": "^10.0.0",
     "@types/validator": "^13.15.10"
   },
-  "gitHead": "115c6d190293c1a5f366cf1997be7e84433f9e05"
+  "gitHead": "680281a11296cb509f64823461a30de237960a05"
 }