@myscheme/voice-navigation-sdk 0.1.9 → 0.1.10

package/dist/microphone-handler.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"microphone-handler.d.ts","sourceRoot":"","sources":["../src/microphone-handler.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,GAAG,MAAM,wCAAwC,CAAC;AAC9D,OAAO,KAAK,EAAE,mBAAmB,EAAE,mBAAmB,EAAE,MAAM,YAAY,CAAC;AAC3E,OAAO,EAAE,kBAAkB,EAAE,MAAM,4BAA4B,CAAC;AAChE,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AAEvD,UAAU,gBAAgB;IACxB,kBAAkB,EAAE,kBAAkB,CAAC;IACvC,cAAc,EAAE,cAAc,CAAC;IAC/B,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,qBAAa,iBAAiB;IAC5B,OAAO,CAAC,kBAAkB,CAAqB;IAC/C,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,cAAc,CAAS;IAC/B,OAAO,CAAC,KAAK,CAAuB;IACpC,OAAO,CAAC,MAAM,CAAuB;IACrC,OAAO,CAAC,UAAU,CAAqC;IACvD,OAAO,CAAC,mBAAmB,CAAgB;IAC3C,OAAO,CAAC,cAAc,CAA8C;IACpE,OAAO,CAAC,WAAW,CAAkB;IACrC,OAAO,CAAC,aAAa,CAAkB;IACvC,OAAO,CAAC,YAAY,CAA6B;IACjD,OAAO,CAAC,oBAAoB,CAAkB;IAC9C,OAAO,CAAC,eAAe,CAA2B;IAClD,OAAO,CAAC,mBAAmB,CAAa;IACxC,OAAO,CAAC,8BAA8B,CAEtB;IAChB,OAAO,CAAC,YAAY,CAAkB;IACtC,OAAO,CAAC,uBAAuB,CAAkB;gBAErC,MAAM,EAAE,gBAAgB;IAUpC,uBAAuB,IAAI,OAAO;IAIlC,uBAAuB,IAAI,OAAO;IAO5B,kBAAkB,IAAI,OAAO,CAAC,eAAe,GAAG,SAAS,CAAC;IAoB1D,qBAAqB,IAAI,OAAO,CAAC,OAAO,CAAC;IAgBzC,kBAAkB,CAAC,aAAa,EAAE,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC;IAwDzD,2BAA2B,IAAI,OAAO,CAAC,OAAO,CAAC;IA2BrD,WAAW,CAAC,QAAQ,EAAE,MAAM,GAAG,IAAI;IAYnC,wBAAwB,IAAI,IAAI;IAO1B,eAAe,IAAI,OAAO,CAAC,OAAO,CAAC;IAgBnC,eAAe,IAAI,OAAO,CAAC,OAAO,CAAC;IAkBzC,iBAAiB,IAAI,IAAI;IAUzB,iBAAiB,IAAI,MAAM;IAO3B,oBAAoB,CAAC,SAAS,GAAE,mBAAwB,GAAG,IAAI;IAiB/D,sBAAsB,IAAI,IAAI;IAmB9B,OAAO,CAAC,sBAAsB;IA4B9B,OAAO,CAAC,gBAAgB;IAyBxB,OAAO,CAAC,eAAe;YAUT,4BAA4B;IAoCpC,kBAAkB,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,mBAAmB,CAAC;IAiBpE,iBAAiB,IAAI,IAAI;IAqCnB,gBAAgB,IAAI,OAAO,CAAC,GAAG,CAAC,gBAAgB,CAAC;IAmBjD,cAAc,CAAC,SAAS,GAAE,mBAAwB,GAAG,OAAO,CAAC,OAAO,CAAC;IA6JrE,aAAa,IAAI,OAAO,CAAC,MAAM,CAAC;WA6CzB,qBAAqB,IAAI,OAAO,CAAC,OAAO,CAAC;CAYvD"}
+{"version":3,"file":"microphone-handler.d.ts","sourceRoot":"","sources":["../src/microphone-handler.ts"],"names":[],"mappings":"AAKA,OAAO,KAAK,GAAG,MAAM,wCAAwC,CAAC;AAC9D,OAAO,KAAK,EACV,mBAAmB,EACnB,mBAAmB,EACnB,cAAc,EACf,MAAM,YAAY,CAAC;AACpB,OAAO,EAAE,kBAAkB,EAAE,MAAM,4BAA4B,CAAC;AAChE,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EAAE,mBAAmB,EAAE,MAAM,6BAA6B,CAAC;AAElE,UAAU,gBAAgB;IACxB,cAAc,CAAC,EAAE,cAAc,CAAC;IAChC,kBAAkB,CAAC,EAAE,kBAAkB,CAAC;IACxC,mBAAmB,CAAC,EAAE,mBAAmB,CAAC;IAC1C,cAAc,EAAE,cAAc,CAAC;IAC/B,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,cAAc,CAAC,EAAE,MAAM,CAAC;CACzB;AAED,qBAAa,iBAAiB;IAC5B,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,kBAAkB,CAA4B;IACtD,OAAO,CAAC,mBAAmB,CAA6B;IACxD,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,QAAQ,CAAS;IACzB,OAAO,CAAC,cAAc,CAAS;IAC/B,OAAO,CAAC,KAAK,CAAuB;IACpC,OAAO,CAAC,MAAM,CAAuB;IACrC,OAAO,CAAC,UAAU,CAAqC;IACvD,OAAO,CAAC,mBAAmB,CAAgB;IAC3C,OAAO,CAAC,cAAc,CAA8C;IACpE,OAAO,CAAC,WAAW,CAAkB;IACrC,OAAO,CAAC,aAAa,CAAkB;IACvC,OAAO,CAAC,YAAY,CAA6B;IACjD,OAAO,CAAC,oBAAoB,CAAkB;IAC9C,OAAO,CAAC,eAAe,CAA2B;IAClD,OAAO,CAAC,mBAAmB,CAAa;IACxC,OAAO,CAAC,8BAA8B,CAEtB;IAChB,OAAO,CAAC,YAAY,CAAkB;IACtC,OAAO,CAAC,uBAAuB,CAAkB;gBAErC,MAAM,EAAE,gBAAgB;IAqBpC,uBAAuB,IAAI,OAAO;IAIlC,uBAAuB,IAAI,OAAO;IAO5B,kBAAkB,IAAI,OAAO,CAAC,eAAe,GAAG,SAAS,CAAC;IAoB1D,qBAAqB,IAAI,OAAO,CAAC,OAAO,CAAC;IAgBzC,kBAAkB,CAAC,aAAa,EAAE,OAAO,GAAG,OAAO,CAAC,IAAI,CAAC;IAwDzD,2BAA2B,IAAI,OAAO,CAAC,OAAO,CAAC;IA2BrD,WAAW,CAAC,QAAQ,EAAE,MAAM,GAAG,IAAI;IAYnC,wBAAwB,IAAI,IAAI;IAO1B,eAAe,IAAI,OAAO,CAAC,OAAO,CAAC;IAoBnC,eAAe,IAAI,OAAO,CAAC,OAAO,CAAC;IAkBzC,iBAAiB,IAAI,IAAI;IAUzB,iBAAiB,IAAI,MAAM;IAO3B,oBAAoB,CAAC,SAAS,GAAE,mBAAwB,GAAG,IAAI;IAiB/D,sBAAsB,IAAI,IAAI;IAmB9B,OAAO,CAAC,sBAAsB;IA4B9B,OAAO,CAAC,gBAAgB;IAyBxB,OAAO,CAAC,eAAe;YAUT,4BAA4B;IAwCpC,kBAAkB,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAAC,mBAAmB,CAAC;IAiBpE,iBAAiB,IAAI,IAAI;IAiDnB,gBAAgB,IAAI,OAAO,CAAC,GAAG,CAAC,gBAAgB,CAAC;IAmBjD,cAAc,CAAC,SAAS,GAAE,mBAAwB,GAAG,OAAO,CAAC,OAAO,CAAC;YA8J7D,oBAAoB;IAqD5B,aAAa,IAAI,OAAO,CAAC,MAAM,CAAC;WAgEzB,qBAAqB,IAAI,OAAO,CAAC,OAAO,CAAC;CAYvD"}

package/dist/microphone-handler.js

@@ -15,7 +15,15 @@ export class MicrophoneHandler {
         this.recognitionHealthCheckInterval = null;
         this.isRecovering = false;
         this.isStoppingIntentionally = false;
-        this.azureSpeechService = config.azureSpeechService;
+        this.speechProvider = config.speechProvider ?? "sarvam";
+        this.azureSpeechService = config.azureSpeechService ?? null;
+        this.sarvamSpeechService = config.sarvamSpeechService ?? null;
+        if (this.speechProvider === "azure" && !this.azureSpeechService) {
+            throw new Error("Azure speech service is required for azure provider");
+        }
+        if (this.speechProvider === "sarvam" && !this.sarvamSpeechService) {
+            throw new Error("Sarvam speech service is required for sarvam provider");
+        }
         this.bedrockService = config.bedrockService;
         this.language = config.language || "en-IN";
         this.silenceTimeout = config.silenceTimeout || 2000;
@@ -127,6 +135,9 @@ export class MicrophoneHandler {
         this.transcriptionBuffer = [];
     }
     async fetchAzureToken() {
+        if (!this.azureSpeechService) {
+            throw new Error("Azure speech service is not configured");
+        }
         try {
             const response = await this.azureSpeechService.fetchToken();
             this.token = response.token;
@@ -174,7 +185,7 @@ export class MicrophoneHandler {
             console.warn("[MicrophoneHandler] Cannot reschedule - not recording");
             return;
         }
-        if (!this.recognizer) {
+        if (this.speechProvider === "azure" && !this.recognizer) {
             console.error("[MicrophoneHandler] Cannot reschedule - no recognizer");
             return;
         }
@@ -220,6 +231,9 @@ export class MicrophoneHandler {
         }
         this.isRecovering = true;
         console.log("[MicrophoneHandler] Attempting to recover from stalled recognizer");
+        if (this.speechProvider === "sarvam" && this.sarvamSpeechService) {
+            await this.sarvamSpeechService.stopListening();
+        }
         this.disposeRecognizer();
         this.isRecording = false;
         await new Promise((resolve) => setTimeout(resolve, 500));
@@ -254,6 +268,14 @@ export class MicrophoneHandler {
     }
     disposeRecognizer() {
         this.stopHealthCheck();
+        if (this.speechProvider === "sarvam") {
+            if (this.sarvamSpeechService) {
+                void this.sarvamSpeechService.stopListening().catch((error) => {
+                    console.error("[MicrophoneHandler] Error stopping Sarvam STT:", error);
+                });
+            }
+            return;
+        }
         if (this.recognizer) {
             try {
                 this.recognizer.stopContinuousRecognitionAsync(() => {
@@ -308,6 +330,9 @@ export class MicrophoneHandler {
                 return false;
             }
         }
+        if (this.speechProvider === "sarvam") {
+            return this.startSarvamRecording(callbacks);
+        }
         const sdkReady = await this.ensureSpeechSdk();
         if (!sdkReady) {
             if (callbacks.onError) {
@@ -398,10 +423,72 @@ export class MicrophoneHandler {
         });
         return true;
     }
+    async startSarvamRecording(callbacks) {
+        if (!this.sarvamSpeechService) {
+            callbacks.onError?.(new Error("Sarvam speech service not configured"));
+            return false;
+        }
+        this.resetTranscriptionBuffer();
+        try {
+            await this.sarvamSpeechService.startListening(this.language, {
+                shouldSendAudio: () => this.isRecording,
+                onPartial: (text) => {
+                    this.lastRecognitionTime = Date.now();
+                    this.activeCallbacks.onPartial?.(text);
+                },
+                onFinal: (text) => {
+                    this.lastRecognitionTime = Date.now();
+                    this.transcriptionBuffer.push(text);
+                    this.activeCallbacks.onSegment?.(text, {});
+                    this.scheduleSilenceCheck(this.activeCallbacks);
+                },
+                onSpeechStart: () => {
+                    this.lastRecognitionTime = Date.now();
+                },
+                onSpeechEnd: () => {
+                    this.lastRecognitionTime = Date.now();
+                    this.scheduleSilenceCheck(this.activeCallbacks);
+                },
+                onError: (error) => {
+                    this.isRecording = false;
+                    this.stopHealthCheck();
+                    this.activeCallbacks.onError?.(error);
+                },
+            });
+            this.isRecording = true;
+            this.audioContextUnlocked = true;
+            this.lastRecognitionTime = Date.now();
+            this.startHealthCheck();
+            this.scheduleSilenceCheck(this.activeCallbacks);
+            return true;
+        }
+        catch (error) {
+            this.isRecording = false;
+            callbacks.onError?.(error);
+            return false;
+        }
+    }
     async stopRecording() {
         this.clearSilenceTimer();
         this.stopHealthCheck();
         this.isStoppingIntentionally = true;
+        if (this.speechProvider === "sarvam") {
+            if (!this.isRecording || !this.sarvamSpeechService) {
+                console.warn("[MicrophoneHandler] Not currently recording");
+                this.isStoppingIntentionally = false;
+                return this.getAggregatedText();
+            }
+            try {
+                await this.sarvamSpeechService.stopListening();
+            }
+            catch (error) {
+                console.error("[MicrophoneHandler] Error stopping Sarvam STT:", error);
+            }
+            this.isRecording = false;
+            this.isStoppingIntentionally = false;
+            const finalText = this.getAggregatedText();
+            return finalText;
+        }
         if (!this.isRecording || !this.recognizer) {
             console.warn("[MicrophoneHandler] Not currently recording");
             this.isStoppingIntentionally = false;

package/dist/navigation-controller.d.ts

@@ -1,8 +1,10 @@
 import type { NavigationConfig } from "./types.js";
 export declare class VoiceNavigationController {
     private config;
+    private speechProvider;
     private microphoneHandler;
     private azureSpeechService;
+    private sarvamSpeechService;
     private bedrockService;
     private voiceFeedbackService;
     private pageRegistry;

package/dist/navigation-controller.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"navigation-controller.d.ts","sourceRoot":"","sources":["../src/navigation-controller.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EACV,gBAAgB,EAMjB,MAAM,YAAY,CAAC;AA4DpB,qBAAa,yBAAyB;IACpC,OAAO,CAAC,MAAM,CAAmB;IACjC,OAAO,CAAC,iBAAiB,CAAoB;IAC7C,OAAO,CAAC,kBAAkB,CAAqB;IAC/C,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,oBAAoB,CAAuB;IACnD,OAAO,CAAC,YAAY,CAAe;IACnC,OAAO,CAAC,EAAE,CAAc;IACxB,OAAO,CAAC,cAAc,CAAc;IACpC,OAAO,CAAC,YAAY,CAAkB;IACtC,OAAO,CAAC,qBAAqB,CAAkB;IAC/C,OAAO,CAAC,eAAe,CAA2B;IAClD,OAAO,CAAC,+BAA+B,CAAkB;IACzD,OAAO,CAAC,wBAAwB,CAA6B;IAC7D,OAAO,CAAC,mBAAmB,CAAoC;IAC/D,OAAO,CAAC,gBAAgB,CAAgB;IACxC,OAAO,CAAC,aAAa,CAAgB;IACrC,OAAO,CAAC,oBAAoB,CAAkB;IAC9C,OAAO,CAAC,gBAAgB,CAAkB;IAC1C,OAAO,CAAC,wBAAwB,CAA6B;gBAEjD,MAAM,EAAE,gBAAgB;YA0JtB,iBAAiB;IAiD/B,OAAO,CAAC,eAAe;IAoBhB,YAAY,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAepC,oBAAoB,CAAC,OAAO,GAAE;QAAE,MAAM,CAAC,EAAE,MAAM,CAAA;KAAO,GAAG,IAAI;YAOtD,eAAe;IAmE7B,OAAO,CAAC,kBAAkB;YASZ,eAAe;IA8BhB,KAAK,CAChB,OAAO,GAAE;QAAE,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,CAAA;KAAO,GACzC,OAAO,CAAC,IAAI,CAAC;IAuFH,IAAI,CACf,OAAO,GAAE;QAAE,eAAe,CAAC,EAAE,OAAO,CAAA;KAAO,GAC1C,OAAO,CAAC,IAAI,CAAC;IAqEhB,OAAO,CAAC,mBAAmB;IAoD3B,OAAO,CAAC,qBAAqB;IA4C7B,OAAO,CAAC,aAAa;IAUrB,OAAO,CAAC,aAAa;YAWP,aAAa;YAOb,wBAAwB;IA2CtC,OAAO,CAAC,WAAW;YAuBL,iBAAiB;YAwBjB,iBAAiB;IA6RxB,WAAW,CAAC,QAAQ,EAAE,MAAM,GAAG,IAAI;IAQnC,gBAAgB,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAOxC,sBAAsB,IAAI,OAAO;YAO1B,qBAAqB;YAgFrB,yBAAyB;IAmMvC,OAAO,CAAC,oBAAoB;IAqD5B,OAAO,CAAC,yBAAyB;IAoCjC,OAAO,CAAC,UAAU;IAQlB,OAAO,CAAC,YAAY;IAUpB,OAAO,CAAC,YAAY;IASpB,OAAO,CAAC,4BAA4B;IAiB7B,OAAO,IAAI,IAAI;IAgBtB,OAAO,CAAC,kBAAkB;IA8B1B,OAAO,CAAC,kBAAkB;IAwD1B,OAAO,CAAC,aAAa;IAYrB,OAAO,CAAC,uBAAuB;IAqD/B,OAAO,CAAC,sBAAsB;IAW9B,OAAO,CAAC,wBAAwB;CAmBjC"}
+{"version":3,"file":"navigation-controller.d.ts","sourceRoot":"","sources":["../src/navigation-controller.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EACV,gBAAgB,EAOjB,MAAM,YAAY,CAAC;AA8DpB,qBAAa,yBAAyB;IACpC,OAAO,CAAC,MAAM,CAAmB;IACjC,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,iBAAiB,CAAoB;IAC7C,OAAO,CAAC,kBAAkB,CAA4B;IACtD,OAAO,CAAC,mBAAmB,CAA6B;IACxD,OAAO,CAAC,cAAc,CAAiB;IACvC,OAAO,CAAC,oBAAoB,CAAuB;IACnD,OAAO,CAAC,YAAY,CAAe;IACnC,OAAO,CAAC,EAAE,CAAc;IACxB,OAAO,CAAC,cAAc,CAAc;IACpC,OAAO,CAAC,YAAY,CAAkB;IACtC,OAAO,CAAC,qBAAqB,CAAkB;IAC/C,OAAO,CAAC,eAAe,CAA2B;IAClD,OAAO,CAAC,+BAA+B,CAAkB;IACzD,OAAO,CAAC,wBAAwB,CAA6B;IAC7D,OAAO,CAAC,mBAAmB,CAAoC;IAC/D,OAAO,CAAC,gBAAgB,CAAgB;IACxC,OAAO,CAAC,aAAa,CAAgB;IACrC,OAAO,CAAC,oBAAoB,CAAkB;IAC9C,OAAO,CAAC,gBAAgB,CAAkB;IAC1C,OAAO,CAAC,wBAAwB,CAA6B;gBAEjD,MAAM,EAAE,gBAAgB;YAgMtB,iBAAiB;IAoD/B,OAAO,CAAC,eAAe;IAoBhB,YAAY,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAepC,oBAAoB,CAAC,OAAO,GAAE;QAAE,MAAM,CAAC,EAAE,MAAM,CAAA;KAAO,GAAG,IAAI;YAOtD,eAAe;IAmE7B,OAAO,CAAC,kBAAkB;YASZ,eAAe;IA8BhB,KAAK,CAChB,OAAO,GAAE;QAAE,MAAM,CAAC,EAAE,MAAM,GAAG,MAAM,CAAA;KAAO,GACzC,OAAO,CAAC,IAAI,CAAC;IAuFH,IAAI,CACf,OAAO,GAAE;QAAE,eAAe,CAAC,EAAE,OAAO,CAAA;KAAO,GAC1C,OAAO,CAAC,IAAI,CAAC;IAqEhB,OAAO,CAAC,mBAAmB;IAoD3B,OAAO,CAAC,qBAAqB;IA4C7B,OAAO,CAAC,aAAa;IAUrB,OAAO,CAAC,aAAa;YAWP,aAAa;YAOb,wBAAwB;IA2CtC,OAAO,CAAC,WAAW;YAuBL,iBAAiB;YAwBjB,iBAAiB;IA6RxB,WAAW,CAAC,QAAQ,EAAE,MAAM,GAAG,IAAI;IASnC,gBAAgB,CAAC,OAAO,EAAE,OAAO,GAAG,IAAI;IAOxC,sBAAsB,IAAI,OAAO;YAO1B,qBAAqB;YAgFrB,yBAAyB;IAmMvC,OAAO,CAAC,oBAAoB;IAqD5B,OAAO,CAAC,yBAAyB;IAoCjC,OAAO,CAAC,UAAU;IAQlB,OAAO,CAAC,YAAY;IAUpB,OAAO,CAAC,YAAY;IASpB,OAAO,CAAC,4BAA4B;IAiB7B,OAAO,IAAI,IAAI;IAgBtB,OAAO,CAAC,kBAAkB;IA8B1B,OAAO,CAAC,kBAAkB;IAwD1B,OAAO,CAAC,aAAa;IAYrB,OAAO,CAAC,uBAAuB;IAqD/B,OAAO,CAAC,sBAAsB;IAW9B,OAAO,CAAC,wBAAwB;CAmBjC"}

package/dist/navigation-controller.js

@@ -1,9 +1,11 @@
 import { MicrophoneHandler } from "./microphone-handler.js";
 import { AzureSpeechService } from "./services/azure-speech.js";
 import { BedrockService } from "./services/bedrock.js";
+import { SarvamSpeechService } from "./services/sarvam-speech.js";
 import { PageRegistry } from "./services/page-registry.js";
 import { VoiceFeedbackService } from "./services/voice-feedback.js";
 import { loadNavigationPages } from "./services/xml-parser.js";
+import { normalizeLanguageCode } from "./languages.js";
 import { VectorSearchService, } from "./services/vector-search.js";
 import { performAgentAction, formatActionLabel, extractAgentAction, setPageRegistry, } from "./actions.js";
 import { createFloatingControl, attachToCustomButton, setState, updateStatus, updateTranscript, updateActionIndicator, showError, emitEvent, } from "./ui.js";
@@ -45,17 +47,28 @@ export class VoiceNavigationController {
         this.config = {
             ...config,
             opensearch: openSearchConfig,
+            speechProvider: config.speechProvider ?? "sarvam",
         };
+        this.speechProvider = this.config.speechProvider ?? "sarvam";
         console.log("[Voice Navigation Controller] Constructor called with UI config:", this.config.ui);
-        const hasAzureEndpoint = !!this.config.azure?.tokenEndpoint;
-        const hasAzureCredentials = !!(this.config.azure?.subscriptionKey && this.config.azure?.region);
-        if (!this.config.azure) {
-            this.config.azure = { tokenEndpoint: "/api/speech/token" };
-            console.log("[VoiceNavigation] Auto-configured Azure endpoint: /api/speech/token");
+        if (this.speechProvider === "azure") {
+            const hasAzureEndpoint = !!this.config.azure?.tokenEndpoint;
+            const hasAzureCredentials = !!(this.config.azure?.subscriptionKey && this.config.azure?.region);
+            if (!this.config.azure) {
+                this.config.azure = { tokenEndpoint: "/api/speech/token" };
+                console.log("[VoiceNavigation] Auto-configured Azure endpoint: /api/speech/token");
+            }
+            else if (!hasAzureEndpoint && !hasAzureCredentials) {
+                this.config.azure.tokenEndpoint = "/api/speech/token";
+                console.log("[VoiceNavigation] Auto-configured Azure endpoint: /api/speech/token");
+            }
         }
-        else if (!hasAzureEndpoint && !hasAzureCredentials) {
-            this.config.azure.tokenEndpoint = "/api/speech/token";
-            console.log("[VoiceNavigation] Auto-configured Azure endpoint: /api/speech/token");
+        if (this.speechProvider === "sarvam") {
+            this.config.sarvam = {
+                ...(this.config.sarvam ?? {}),
+                ttsUrl: this.config.sarvam?.ttsUrl ?? "/api/speech/sarvam",
+                sttUrl: this.config.sarvam?.sttUrl ?? "/api/speech/sarvam-stt",
+            };
         }
         const hasAwsEndpoint = !!this.config.aws?.bedrockEndpoint;
         const hasAwsCredentials = !!(this.config.aws?.accessKeyId &&
@@ -80,11 +93,29 @@ export class VoiceNavigationController {
             openSearchConfig.suggestEndpoint = "/api/opensearch/suggest";
             console.log("[VoiceNavigation] Auto-configured OpenSearch endpoints: /api/opensearch/search, /api/opensearch/suggest");
         }
-        this.azureSpeechService = new AzureSpeechService({
-            tokenEndpoint: this.config.azure.tokenEndpoint,
-            subscriptionKey: this.config.azure.subscriptionKey,
-            region: this.config.azure.region,
-        });
+        this.azureSpeechService = null;
+        this.sarvamSpeechService = null;
+        if (this.speechProvider === "azure") {
+            if (!this.config.azure) {
+                throw new Error("Azure configuration is required for azure provider");
+            }
+            this.azureSpeechService = new AzureSpeechService({
+                tokenEndpoint: this.config.azure.tokenEndpoint,
+                subscriptionKey: this.config.azure.subscriptionKey,
+                region: this.config.azure.region,
+            });
+        }
+        else {
+            this.sarvamSpeechService = new SarvamSpeechService({
+                ttsUrl: this.config.sarvam?.ttsUrl,
+                sttUrl: this.config.sarvam?.sttUrl,
+                sttModel: this.config.sarvam?.sttModel,
+                ttsModel: this.config.sarvam?.ttsModel,
+                speaker: this.config.sarvam?.speaker,
+                sttSampleRate: this.config.sarvam?.sttSampleRate,
+                ttsSampleRate: this.config.sarvam?.ttsSampleRate,
+            });
+        }
         this.bedrockService = new BedrockService({
             bedrockEndpoint: this.config.aws.bedrockEndpoint,
             embeddingEndpoint: this.config.aws.embeddingEndpoint,
@@ -99,16 +130,20 @@ export class VoiceNavigationController {
         this.pageRegistry = new PageRegistry([]);
         this.pagesInitialized = this.initializePages();
         this.microphoneHandler = new MicrophoneHandler({
-            azureSpeechService: this.azureSpeechService,
+            speechProvider: this.speechProvider,
+            azureSpeechService: this.azureSpeechService ?? undefined,
+            sarvamSpeechService: this.sarvamSpeechService ?? undefined,
             bedrockService: this.bedrockService,
-            language: this.config.language || "en-IN",
+            language: normalizeLanguageCode(this.config.language),
         });
         const voiceFeedbackEnabled = this.config.voiceFeedback?.enabled !== false;
         console.log("[VoiceNavigation] Initializing voice feedback service - enabled:", voiceFeedbackEnabled);
         this.voiceFeedbackService = new VoiceFeedbackService({
             enabled: voiceFeedbackEnabled,
-            language: this.config.language || "en-IN",
-            azureSpeechService: this.azureSpeechService,
+            language: normalizeLanguageCode(this.config.language),
+            speechProvider: this.speechProvider,
+            azureSpeechService: this.azureSpeechService ?? undefined,
+            sarvamSpeechService: this.sarvamSpeechService ?? undefined,
             bedrockService: this.bedrockService,
         });
         console.log("[VoiceNavigation] Voice feedback service initialized successfully");
@@ -143,7 +178,7 @@ export class VoiceNavigationController {
             customButtonSelector,
             buttonPlacement,
         });
-        if (!showDefaultButton && customButtonSelector) {
+        if (customButtonSelector) {
             console.log(`[Voice Navigation] Attempting to use custom button: ${customButtonSelector}`);
             try {
                 this.ui = await attachToCustomButton(customButtonSelector);
@@ -151,11 +186,11 @@ export class VoiceNavigationController {
             }
             catch (error) {
                 console.error("Failed to attach to custom button:", error);
-                console.warn("Falling back to default button due to custom button error");
+                console.warn("Falling back to default button - custom button selector was not found in the DOM");
                 this.ui = createFloatingControl(buttonPlacement);
             }
         }
-        else if (!showDefaultButton && !customButtonSelector) {
+        else if (!showDefaultButton) {
             console.warn("showDefaultButton is false but no customButtonSelector provided. Creating default button.");
             this.ui = createFloatingControl(buttonPlacement);
         }
@@ -719,8 +754,9 @@ export class VoiceNavigationController {
         });
     }
     setLanguage(language) {
-        this.microphoneHandler.setLanguage(language);
-        this.voiceFeedbackService.setLanguage(language);
+        const normalized = normalizeLanguageCode(language);
+        this.microphoneHandler.setLanguage(normalized);
+        this.voiceFeedbackService.setLanguage(normalized);
     }
     setVoiceFeedback(enabled) {
         this.voiceFeedbackService.setEnabled(enabled);

package/dist/server/azure-speech-handler.d.ts

@@ -1,4 +1,5 @@
-type NextApiRequest = {
+import { type GateRequestLike } from "./security.js";
+type NextApiRequest = GateRequestLike & {
     method?: string;
     body?: unknown;
 };
@@ -8,6 +9,6 @@ type NextApiResponse = {
     setHeader: (name: string, value: string | string[]) => void;
 };
 export declare function azureSpeechTokenHandler(req: NextApiRequest, res: NextApiResponse): Promise<void>;
-export declare function POST(_request: Request): Promise<Response>;
+export declare function POST(request: Request): Promise<Response>;
 export {};
 //# sourceMappingURL=azure-speech-handler.d.ts.map

package/dist/server/azure-speech-handler.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"azure-speech-handler.d.ts","sourceRoot":"","sources":["../../src/server/azure-speech-handler.ts"],"names":[],"mappings":"AAwBA,KAAK,cAAc,GAAG;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE,OAAO,CAAC;CAChB,CAAC;AAEF,KAAK,eAAe,GAAG;IACrB,MAAM,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,eAAe,CAAC;IAC1C,IAAI,EAAE,CAAC,IAAI,EAAE,OAAO,KAAK,IAAI,CAAC;IAC9B,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,KAAK,IAAI,CAAC;CAC7D,CAAC;AA6EF,wBAAsB,uBAAuB,CAC3C,GAAG,EAAE,cAAc,EACnB,GAAG,EAAE,eAAe,GACnB,OAAO,CAAC,IAAI,CAAC,CAqBf;AAMD,wBAAsB,IAAI,CAAC,QAAQ,EAAE,OAAO,GAAG,OAAO,CAAC,QAAQ,CAAC,CAmB/D"}
+{"version":3,"file":"azure-speech-handler.d.ts","sourceRoot":"","sources":["../../src/server/azure-speech-handler.ts"],"names":[],"mappings":"AAuBA,OAAO,EAAgB,KAAK,eAAe,EAAE,MAAM,eAAe,CAAC;AAGnE,KAAK,cAAc,GAAG,eAAe,GAAG;IACtC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE,OAAO,CAAC;CAChB,CAAC;AAEF,KAAK,eAAe,GAAG;IACrB,MAAM,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,eAAe,CAAC;IAC1C,IAAI,EAAE,CAAC,IAAI,EAAE,OAAO,KAAK,IAAI,CAAC;IAC9B,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,KAAK,IAAI,CAAC;CAC7D,CAAC;AA6EF,wBAAsB,uBAAuB,CAC3C,GAAG,EAAE,cAAc,EACnB,GAAG,EAAE,eAAe,GACnB,OAAO,CAAC,IAAI,CAAC,CAgCf;AAMD,wBAAsB,IAAI,CAAC,OAAO,EAAE,OAAO,GAAG,OAAO,CAAC,QAAQ,CAAC,CAqC9D"}

package/dist/server/azure-speech-handler.js

@@ -1,3 +1,4 @@
+import { evaluateGate } from "./security.js";
 async function getAzureSpeechToken() {
     const key = process.env.AZURE_SPEECH_KEY;
     const region = process.env.AZURE_SPEECH_REGION;
@@ -49,6 +50,12 @@ export async function azureSpeechTokenHandler(req, res) {
         res.status(405).json({ error: "Method not allowed" });
         return;
     }
+    const gate = evaluateGate(req, { maxRequests: 20, windowMs: 60 * 1000, bucket: "aztoken" });
+    if (!gate.ok) {
+        res.setHeader("Cache-Control", "no-store, no-cache, max-age=0, must-revalidate");
+        res.status(gate.status ?? 403).json({ error: gate.error });
+        return;
+    }
     const result = await getAzureSpeechToken();
     res.setHeader("Cache-Control", "no-store, no-cache, max-age=0, must-revalidate");
     if (result.success && result.data) {
@@ -58,12 +65,29 @@ export async function azureSpeechTokenHandler(req, res) {
         res.status(result.statusCode).json(result.error);
     }
 }
-export async function POST(_request) {
-    const result = await getAzureSpeechToken();
+export async function POST(request) {
     const headers = {
         "Content-Type": "application/json",
         "Cache-Control": "no-store, no-cache, max-age=0, must-revalidate",
     };
+    const gateReq = {
+        method: request.method,
+        headers: {
+            origin: request.headers.get("origin") ?? undefined,
+            referer: request.headers.get("referer") ?? undefined,
+            host: request.headers.get("host") ?? undefined,
+            "x-forwarded-for": request.headers.get("x-forwarded-for") ?? undefined,
+            "x-real-ip": request.headers.get("x-real-ip") ?? undefined,
+        },
+    };
+    const gate = evaluateGate(gateReq, { maxRequests: 20, windowMs: 60 * 1000, bucket: "aztoken" });
+    if (!gate.ok) {
+        return new Response(JSON.stringify({ error: gate.error }), {
+            status: gate.status ?? 403,
+            headers,
+        });
+    }
+    const result = await getAzureSpeechToken();
     if (result.success && result.data) {
         return new Response(JSON.stringify(result.data), {
             status: result.statusCode,

package/dist/server/bedrock-embedding-handler.d.ts

@@ -1,4 +1,5 @@
-type NextApiRequest = {
+import { type GateRequestLike } from "./security.js";
+type NextApiRequest = GateRequestLike & {
     method?: string;
     body?: {
         text?: string | string[];

package/dist/server/bedrock-embedding-handler.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"bedrock-embedding-handler.d.ts","sourceRoot":"","sources":["../../src/server/bedrock-embedding-handler.ts"],"names":[],"mappings":"AA4BA,KAAK,cAAc,GAAG;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE;QACL,IAAI,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;QACzB,SAAS,CAAC,EAAE,MAAM,CAAC;KACpB,CAAC;CACH,CAAC;AAEF,KAAK,eAAe,GAAG;IACrB,MAAM,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,eAAe,CAAC;IAC1C,IAAI,EAAE,CAAC,IAAI,EAAE,OAAO,KAAK,IAAI,CAAC;IAC9B,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,KAAK,IAAI,CAAC;CAC7D,CAAC;AA0MF,wBAAsB,uBAAuB,CAC3C,GAAG,EAAE,cAAc,EACnB,GAAG,EAAE,eAAe,GACnB,OAAO,CAAC,IAAI,CAAC,CAiBf;AAMD,wBAAsB,oBAAoB,CACxC,OAAO,EAAE,OAAO,GACf,OAAO,CAAC,QAAQ,CAAC,CAkCnB"}
+{"version":3,"file":"bedrock-embedding-handler.d.ts","sourceRoot":"","sources":["../../src/server/bedrock-embedding-handler.ts"],"names":[],"mappings":"AA2BA,OAAO,EAAgB,KAAK,eAAe,EAAE,MAAM,eAAe,CAAC;AAEnE,KAAK,cAAc,GAAG,eAAe,GAAG;IACtC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE;QACL,IAAI,CAAC,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;QACzB,SAAS,CAAC,EAAE,MAAM,CAAC;KACpB,CAAC;CACH,CAAC;AAEF,KAAK,eAAe,GAAG;IACrB,MAAM,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,eAAe,CAAC;IAC1C,IAAI,EAAE,CAAC,IAAI,EAAE,OAAO,KAAK,IAAI,CAAC;IAC9B,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,KAAK,IAAI,CAAC;CAC7D,CAAC;AA0OF,wBAAsB,uBAAuB,CAC3C,GAAG,EAAE,cAAc,EACnB,GAAG,EAAE,eAAe,GACnB,OAAO,CAAC,IAAI,CAAC,CA8Bf;AAMD,wBAAsB,oBAAoB,CACxC,OAAO,EAAE,OAAO,GACf,OAAO,CAAC,QAAQ,CAAC,CAsDnB"}

package/dist/server/bedrock-embedding-handler.js

@@ -1,4 +1,29 @@
 import crypto from "crypto";
+import { evaluateGate } from "./security.js";
+const MAX_EMBEDDING_TEXTS = 16;
+const MAX_EMBEDDING_TEXT_CHARS = 2000;
+const ALLOWED_INPUT_TYPES = new Set([
+    "search_query",
+    "search_document",
+    "classification",
+    "clustering",
+]);
+function sanitizeEmbeddingInput(text, inputType) {
+    const rawTexts = Array.isArray(text) ? text : text != null ? [text] : [];
+    const texts = [];
+    for (const entry of rawTexts.slice(0, MAX_EMBEDDING_TEXTS)) {
+        if (typeof entry === "string" && entry.trim().length > 0) {
+            texts.push(entry.slice(0, MAX_EMBEDDING_TEXT_CHARS));
+        }
+    }
+    if (texts.length === 0) {
+        return { ok: false };
+    }
+    const resolvedType = typeof inputType === "string" && ALLOWED_INPUT_TYPES.has(inputType)
+        ? inputType
+        : "search_query";
+    return { ok: true, texts, inputType: resolvedType };
+}
 function sign(key, msg) {
     return crypto.createHmac("sha256", key).update(msg, "utf8").digest();
 }
@@ -141,43 +166,72 @@ export async function bedrockEmbeddingHandler(req, res) {
         res.status(405).json({ error: "Method not allowed" });
         return;
     }
-    const { text, inputType = "search_query" } = req.body || {};
-    const result = await generateBedrockEmbedding(text, inputType);
+    const gate = evaluateGate(req, { maxRequests: 30, windowMs: 60 * 1000, bucket: "embed" });
+    if (!gate.ok) {
+        res.status(gate.status ?? 403).json({ error: gate.error });
+        return;
+    }
+    const { text, inputType } = req.body || {};
+    const sanitized = sanitizeEmbeddingInput(text, inputType);
+    if (!sanitized.ok) {
+        res.status(400).json({ error: "Invalid text parameter" });
+        return;
+    }
+    const result = await generateBedrockEmbedding(sanitized.texts, sanitized.inputType);
     if (result.success && result.data) {
         res.status(result.statusCode).json(result.data);
     }
     else {
-        res.status(result.statusCode).json(result.error);
+        res
+            .status(result.statusCode >= 500 ? 502 : result.statusCode)
+            .json({ error: "Embedding request failed" });
     }
 }
 export async function BedrockEmbeddingPOST(request) {
+    const headers = { "Content-Type": "application/json" };
+    const gateReq = {
+        method: request.method,
+        headers: {
+            origin: request.headers.get("origin") ?? undefined,
+            referer: request.headers.get("referer") ?? undefined,
+            host: request.headers.get("host") ?? undefined,
+            "x-forwarded-for": request.headers.get("x-forwarded-for") ?? undefined,
+            "x-real-ip": request.headers.get("x-real-ip") ?? undefined,
+        },
+    };
+    const gate = evaluateGate(gateReq, { maxRequests: 30, windowMs: 60 * 1000, bucket: "embed" });
+    if (!gate.ok) {
+        return new Response(JSON.stringify({ error: gate.error }), {
+            status: gate.status ?? 403,
+            headers,
+        });
+    }
     try {
         const body = await request.json();
-        const { text, inputType = "search_query" } = body || {};
-        const result = await generateBedrockEmbedding(text, inputType);
-        const headers = {
-            "Content-Type": "application/json",
-        };
-        if (result.success && result.data) {
-            return new Response(JSON.stringify(result.data), {
-                status: result.statusCode,
+        const { text, inputType } = body || {};
+        const sanitized = sanitizeEmbeddingInput(text, inputType);
+        if (!sanitized.ok) {
+            return new Response(JSON.stringify({ error: "Invalid text parameter" }), {
+                status: 400,
                 headers,
             });
         }
-        else {
-            return new Response(JSON.stringify(result.error), {
+        const result = await generateBedrockEmbedding(sanitized.texts, sanitized.inputType);
+        if (result.success && result.data) {
+            return new Response(JSON.stringify(result.data), {
                 status: result.statusCode,
                 headers,
             });
         }
+        return new Response(JSON.stringify({ error: "Embedding request failed" }), {
+            status: result.statusCode >= 500 ? 502 : result.statusCode,
+            headers,
+        });
     }
-    catch (error) {
-        return new Response(JSON.stringify({
-            error: "Invalid request",
-            message: error instanceof Error ? error.message : "Unknown error",
-        }), {
+    catch {
+        return new Response(JSON.stringify({ error: "Invalid request" }), {
             status: 400,
-            headers: { "Content-Type": "application/json" },
+            headers,
         });
     }
 }

package/dist/server/bedrock-handler.d.ts

@@ -1,4 +1,5 @@
-type NextApiRequest = {
+import { type GateRequestLike } from "./security.js";
+type NextApiRequest = GateRequestLike & {
     method?: string;
     body?: {
         body?: unknown;

package/dist/server/bedrock-handler.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"bedrock-handler.d.ts","sourceRoot":"","sources":["../../src/server/bedrock-handler.ts"],"names":[],"mappings":"AA4BA,KAAK,cAAc,GAAG;IACpB,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE;QACL,IAAI,CAAC,EAAE,OAAO,CAAC;KAChB,CAAC;CACH,CAAC;AAEF,KAAK,eAAe,GAAG;IACrB,MAAM,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,eAAe,CAAC;IAC1C,IAAI,EAAE,CAAC,IAAI,EAAE,OAAO,KAAK,IAAI,CAAC;IAC9B,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,KAAK,IAAI,CAAC;CAC7D,CAAC;AAqLF,wBAAsB,cAAc,CAClC,GAAG,EAAE,cAAc,EACnB,GAAG,EAAE,eAAe,GACnB,OAAO,CAAC,IAAI,CAAC,CAkBf;AAMD,wBAAsB,WAAW,CAAC,OAAO,EAAE,OAAO,GAAG,OAAO,CAAC,QAAQ,CAAC,CAkCrE"}
+{"version":3,"file":"bedrock-handler.d.ts","sourceRoot":"","sources":["../../src/server/bedrock-handler.ts"],"names":[],"mappings":"AA2BA,OAAO,EAAgB,KAAK,eAAe,EAAE,MAAM,eAAe,CAAC;AAEnE,KAAK,cAAc,GAAG,eAAe,GAAG;IACtC,MAAM,CAAC,EAAE,MAAM,CAAC;IAChB,IAAI,CAAC,EAAE;QACL,IAAI,CAAC,EAAE,OAAO,CAAC;KAChB,CAAC;CACH,CAAC;AAEF,KAAK,eAAe,GAAG;IACrB,MAAM,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,eAAe,CAAC;IAC1C,IAAI,EAAE,CAAC,IAAI,EAAE,OAAO,KAAK,IAAI,CAAC;IAC9B,SAAS,EAAE,CAAC,IAAI,EAAE,MAAM,EAAE,KAAK,EAAE,MAAM,GAAG,MAAM,EAAE,KAAK,IAAI,CAAC;CAC7D,CAAC;AAyRF,wBAAsB,cAAc,CAClC,GAAG,EAAE,cAAc,EACnB,GAAG,EAAE,eAAe,GACnB,OAAO,CAAC,IAAI,CAAC,CA8Bf;AAMD,wBAAsB,WAAW,CAAC,OAAO,EAAE,OAAO,GAAG,OAAO,CAAC,QAAQ,CAAC,CAmDrE"}