@myscheme/voice-navigation-sdk 0.1.9 → 0.1.10

package/README.md

@@ -1,10 +1,11 @@
 # Voice Navigation SDK
 
-A TypeScript SDK for voice-controlled navigation using Azure Speech-to-Text and AWS Bedrock for intent understanding.
+A TypeScript SDK for voice-controlled navigation using Sarvam or Azure Speech, and AWS Bedrock for intent understanding.
 
 ## 🚀 Features
 
-- 🎤 **Real-time speech recognition** using Azure Speech SDK
+- 🎤 **Real-time speech recognition** using Sarvam (default) or Azure Speech SDK
+- 🔊 **Text-to-speech feedback** using Sarvam (default) or Azure Speech SDK
 - 🤖 **AI-powered intent extraction** using AWS Bedrock (Claude)
 - 🌐 **Dynamic page navigation** via XML configuration
 - 🧭 **Voice-controlled navigation** actions
@@ -48,9 +49,23 @@ The guide covers:
 ### Basic Setup
 
 ```typescript
-import { initNavigationOnMicrophone } from "@myscheme/voice-navigation-sdk";
+import {
+  initNavigationOnMicrophone,
+  buildSpeechProviderConfig,
+} from "@myscheme/voice-navigation-sdk";
 
 const controller = initNavigationOnMicrophone({
+  ...buildSpeechProviderConfig({
+    provider: "sarvam",
+    sarvam: {
+      apiKey: "your-sarvam-api-key",
+      sttModel: "saaras:v3",
+      ttsModel: "bulbul:v3",
+      speaker: "manan",
+    },
+  }),
+
+  // Azure Speech configuration (required only when speechProvider is "azure")
   // Azure Speech configuration
   azure: {
     subscriptionKey: "your-azure-subscription-key",
@@ -73,6 +88,97 @@ const controller = initNavigationOnMicrophone({
 });
 ```
 
+### Runtime Provider Switch Helper
+
+Switch between Sarvam and Azure without rebuilding your app:
+
+```typescript
+import { switchVoiceProvider } from "@myscheme/voice-navigation-sdk";
+
+// Switch to Azure at runtime
+switchVoiceProvider("azure", {
+  azure: {
+    tokenEndpoint: "/api/speech/token",
+    // or: subscriptionKey + region for direct mode
+  },
+});
+
+// Switch back to Sarvam at runtime
+switchVoiceProvider("sarvam", {
+  sarvam: {
+    apiKey: process.env.NEXT_PUBLIC_SARVAM_API_KEY,
+    sttModel: "saaras:v3",
+    ttsModel: "bulbul:v3",
+    speaker: "manan",
+  },
+});
+```
+
+### Ready-to-Copy Config
+
+Sarvam (default provider):
+
+```typescript
+import { initNavigationOnMicrophone } from "@myscheme/voice-navigation-sdk";
+
+const controller = initNavigationOnMicrophone({
+  speechProvider: "sarvam",
+  sarvam: {
+    apiKey: process.env.NEXT_PUBLIC_SARVAM_API_KEY,
+    sttModel: "saaras:v3",
+    ttsModel: "bulbul:v3",
+    speaker: "manan",
+    sttSampleRate: 16000,
+    ttsSampleRate: 22050,
+  },
+  aws: {
+    bedrockEndpoint: "/api/speech/bedrock",
+    embeddingEndpoint: "/api/speech/bedrock-embedding",
+  },
+  language: "en-IN",
+  autoStart: false,
+});
+```
+
+Azure (proxy mode recommended):
+
+```typescript
+import { initNavigationOnMicrophone } from "@myscheme/voice-navigation-sdk";
+
+const controller = initNavigationOnMicrophone({
+  speechProvider: "azure",
+  azure: {
+    tokenEndpoint: "/api/speech/token",
+  },
+  aws: {
+    bedrockEndpoint: "/api/speech/bedrock",
+    embeddingEndpoint: "/api/speech/bedrock-embedding",
+  },
+  language: "en-IN",
+  autoStart: false,
+});
+```
+
+Azure (direct credentials, use only in trusted/server contexts):
+
+```typescript
+import { initNavigationOnMicrophone } from "@myscheme/voice-navigation-sdk";
+
+const controller = initNavigationOnMicrophone({
+  speechProvider: "azure",
+  azure: {
+    subscriptionKey: process.env.NEXT_PUBLIC_AZURE_SPEECH_KEY,
+    region: process.env.NEXT_PUBLIC_AZURE_SPEECH_REGION,
+  },
+  aws: {
+    bedrockEndpoint: "/api/speech/bedrock",
+    embeddingEndpoint: "/api/speech/bedrock-embedding",
+  },
+  language: "en-IN",
+  autoStart: false,
+});
+```
+
 ### With Dynamic Pages
 
 ```typescript
@@ -96,20 +202,27 @@ const controller = initNavigationOnMicrophone({
 
 ### NavigationConfig Options
 
-| Property                | Type      | Required | Description                                    |
-| ----------------------- | --------- | -------- | ---------------------------------------------- |
-| `azure.subscriptionKey` | `string`  | ✅       | Azure Speech subscription key                  |
-| `azure.region`          | `string`  | ✅       | Azure Speech region                            |
-| `aws.accessKeyId`       | `string`  | ✅       | AWS access key ID                              |
-| `aws.secretAccessKey`   | `string`  | ✅       | AWS secret access key                          |
-| `aws.modelId`           | `string`  | ✅       | AWS Bedrock model ID                           |
-| `aws.region`            | `string`  | ❌       | AWS region (default: 'ap-south-1')             |
-| `language`              | `string`  | ❌       | Speech recognition language (default: 'en-IN') |
-| `autoStart`             | `boolean` | ❌       | Auto-start voice control (default: false)      |
-| `actionHandlers`        | `object`  | ❌       | Custom action callbacks                        |
-| `pages`                 | `object`  | ❌       | Dynamic page navigation configuration          |
-| `opensearch`            | `object`  | ❌       | OpenSearch vector search configuration         |
-| `ui`                    | `object`  | ❌       | UI customization (button placement & style)    |
+| Property                | Type      | Required | Description                                      |
+| ----------------------- | --------- | -------- | ------------------------------------------------ |
+| `speechProvider`        | `string`  | ❌       | Speech provider: `sarvam` (default) or `azure`   |
+| `sarvam.apiKey`         | `string`  | ✅\*     | Sarvam API key (\*required when provider=sarvam) |
+| `sarvam.sttModel`       | `string`  | ❌       | STT model (default: `saaras:v3`)                 |
+| `sarvam.ttsModel`       | `string`  | ❌       | TTS model (default: `bulbul:v3`)                 |
+| `sarvam.speaker`        | `string`  | ❌       | Sarvam speaker name                              |
+| `azure.subscriptionKey` | `string`  | ✅       | Azure Speech subscription key                    |
+| `azure.region`          | `string`  | ✅       | Azure Speech region                              |
+| `aws.accessKeyId`       | `string`  | ✅       | AWS access key ID                                |
+| `aws.secretAccessKey`   | `string`  | ✅       | AWS secret access key                            |
+| `aws.modelId`           | `string`  | ✅       | AWS Bedrock model ID                             |
+| `aws.region`            | `string`  | ❌       | AWS region (default: 'ap-south-1')               |
+| `language`              | `string`  | ❌       | Speech recognition language (default: 'en-IN')   |
+| `autoStart`             | `boolean` | ❌       | Auto-start voice control (default: false)        |
+| `actionHandlers`        | `object`  | ❌       | Custom action callbacks                          |
+| `pages`                 | `object`  | ❌       | Dynamic page navigation configuration            |
+| `opensearch`            | `object`  | ❌       | OpenSearch vector search configuration           |
+| `ui`                    | `object`  | ❌       | UI customization (button placement & style)      |
+
+Note: Bhashini is not integrated in this release.
 
 ### OpenSearch Configuration
 

package/dist/actions.d.ts

@@ -4,6 +4,7 @@ export declare const setPageRegistry: (registry: PageRegistry) => void;
 export declare const getPageRegistry: () => PageRegistry | null;
 export declare const setZoomScale: (scale: number) => number;
 export declare const adjustZoom: (delta: number) => number;
+export declare const ACTION_LABELS: Record<string, string>;
 export declare const formatActionLabel: (action: string) => string;
 export declare const performAgentAction: (action: NavigationAction, context?: ActionContext) => ActionResult;
 export declare const extractAgentAction: (result: any) => AgentActionResponse | null;

package/dist/actions.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"actions.d.ts","sourceRoot":"","sources":["../src/actions.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EACV,gBAAgB,EAEhB,aAAa,EACb,YAAY,EACZ,mBAAmB,EACpB,MAAM,YAAY,CAAC;AACpB,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,6BAA6B,CAAC;AA4DhE,eAAO,MAAM,eAAe,GAAI,UAAU,YAAY,KAAG,IAOxD,CAAC;AAKF,eAAO,MAAM,eAAe,QAAO,YAAY,GAAG,IAEjD,CAAC;AAgGF,eAAO,MAAM,YAAY,GAAI,OAAO,MAAM,KAAG,MAS5C,CAAC;AAKF,eAAO,MAAM,UAAU,GAAI,OAAO,MAAM,KAAG,MAM1C,CAAC;AAKF,eAAO,MAAM,iBAAiB,GAAI,QAAQ,MAAM,KAAG,MAElD,CAAC;AAKF,eAAO,MAAM,kBAAkB,GAC7B,QAAQ,gBAAgB,EACxB,UAAS,aAAkB,KAC1B,YAg7BF,CAAC;AAKF,eAAO,MAAM,kBAAkB,GAAI,QAAQ,GAAG,KAAG,mBAAmB,GAAG,IAiCtE,CAAC;AAKF,eAAO,MAAM,eAAe,GAAI,QAAQ,MAAM,KAAG,MAAM,IAAI,gBAY1D,CAAC"}
+{"version":3,"file":"actions.d.ts","sourceRoot":"","sources":["../src/actions.ts"],"names":[],"mappings":"AAIA,OAAO,KAAK,EACV,gBAAgB,EAEhB,aAAa,EACb,YAAY,EACZ,mBAAmB,EACpB,MAAM,YAAY,CAAC;AACpB,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,6BAA6B,CAAC;AA4DhE,eAAO,MAAM,eAAe,GAAI,UAAU,YAAY,KAAG,IAOxD,CAAC;AAKF,eAAO,MAAM,eAAe,QAAO,YAAY,GAAG,IAEjD,CAAC;AAgGF,eAAO,MAAM,YAAY,GAAI,OAAO,MAAM,KAAG,MAS5C,CAAC;AAKF,eAAO,MAAM,UAAU,GAAI,OAAO,MAAM,KAAG,MAM1C,CAAC;AAOF,eAAO,MAAM,aAAa,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAiDhD,CAAC;AAQF,eAAO,MAAM,iBAAiB,GAAI,QAAQ,MAAM,KAAG,MAElD,CAAC;AAKF,eAAO,MAAM,kBAAkB,GAC7B,QAAQ,gBAAgB,EACxB,UAAS,aAAkB,KAC1B,YAg7BF,CAAC;AAKF,eAAO,MAAM,kBAAkB,GAAI,QAAQ,GAAG,KAAG,mBAAmB,GAAG,IAiCtE,CAAC;AAKF,eAAO,MAAM,eAAe,GAAI,QAAQ,MAAM,KAAG,MAAM,IAAI,gBAY1D,CAAC"}

package/dist/actions.js

@@ -131,8 +131,58 @@ export const adjustZoom = (delta) => {
     const current = parseFloat(document.documentElement.dataset.navigateZoom || "1");
     return setZoomScale(current + delta);
 };
+export const ACTION_LABELS = {
+    zoom_in: "Zoomed in",
+    zoom_out: "Zoomed out",
+    scroll_up: "Scrolled up",
+    scroll_down: "Scrolled down",
+    scroll_left: "Scrolled left",
+    scroll_right: "Scrolled right",
+    page_up: "Moved up a page",
+    page_down: "Moved down a page",
+    scroll_top: "Jumped to the top of the page",
+    scroll_bottom: "Jumped to the bottom of the page",
+    go_back: "Went back to the previous page",
+    go_forward: "Went forward to the next page",
+    reload_page: "Reloaded the page",
+    print_page: "Opened the print dialog",
+    copy_url: "Copied the page link",
+    open_menu: "Opened the menu",
+    close_menu: "Closed the menu",
+    focus_search: "Moved to the search box",
+    toggle_fullscreen: "Turned on full screen",
+    exit_fullscreen: "Turned off full screen",
+    play_media: "Playing the video",
+    pause_media: "Paused the video",
+    mute_media: "Muted the sound",
+    unmute_media: "Unmuted the sound",
+    search_content: "Searching",
+    stop: "Stopped listening",
+    quit: "Stopped voice navigation",
+    tab_next: "Moved to the next item",
+    tab_back: "Moved to the previous item",
+    focus_next: "Moved focus to the next item",
+    focus_previous: "Moved focus to the previous item",
+    click: "Clicked",
+    enter: "Pressed enter",
+    submit: "Submitted the form",
+    activate: "Activated",
+    move_left: "Moved left",
+    move_right: "Moved right",
+    move_up: "Moved up",
+    move_down: "Moved down",
+    move_beginning: "Moved to the beginning",
+    move_end: "Moved to the end",
+    select_all: "Selected all text",
+    select_line: "Selected the line",
+    type_text: "Typed the text",
+    cut: "Cut the selection",
+    copy: "Copied the selection",
+    paste: "Pasted",
+    clear: "Cleared the field",
+};
 export const formatActionLabel = (action) => {
-    return action.replace(/_/g, " ");
+    return ACTION_LABELS[action] ?? action.replace(/_/g, " ");
 };
 export const performAgentAction = (action, context = {}) => {
     if (!action) {

package/dist/index.d.ts

@@ -1,15 +1,52 @@
 export { VoiceNavigationController } from "./navigation-controller.js";
 export { MicrophoneHandler } from "./microphone-handler.js";
 export { AzureSpeechService } from "./services/azure-speech.js";
+export { SarvamSpeechService } from "./services/sarvam-speech.js";
 export { BedrockService } from "./services/bedrock.js";
 export { PageRegistry } from "./services/page-registry.js";
 export type { NavigablePage } from "./services/page-registry.js";
 export { parseNavigationXML, fetchNavigationXML, loadNavigationPages, } from "./services/xml-parser.js";
 export type { PageXMLConfig } from "./services/xml-parser.js";
+export { SARVAM_LANGUAGES, SARVAM_STT_LANGUAGES, SARVAM_TTS_LANGUAGES, DEFAULT_LANGUAGE, isSttSupported, isTtsSupported, resolveTtsLanguage, normalizeLanguageCode, } from "./languages.js";
+export type { SarvamLanguage } from "./languages.js";
 export * from "./types.js";
 export * from "./actions.js";
 export { createFloatingControl, attachToCustomButton, setState, updateStatus, updateTranscript, updateActionIndicator, showError, emitEvent, } from "./ui.js";
 import { VoiceNavigationController } from "./navigation-controller.js";
-import type { NavigationConfig } from "./types.js";
+import type { NavigationConfig, SpeechProvider } from "./types.js";
+type SpeechProviderConfigOptions = {
+    provider: SpeechProvider;
+    azure?: NavigationConfig["azure"];
+    sarvam?: NavigationConfig["sarvam"];
+};
+export type ProviderSwitchOptions = {
+    azure?: NavigationConfig["azure"];
+    sarvam?: NavigationConfig["sarvam"];
+    language?: NavigationConfig["language"];
+    voiceFeedback?: NavigationConfig["voiceFeedback"];
+};
+export type VoiceProviderDropdownOption = {
+    provider: SpeechProvider;
+    label: string;
+    disabled?: boolean;
+};
+export type VoiceProviderDropdownConfig = {
+    container: string | HTMLElement;
+    initialProvider?: SpeechProvider;
+    options?: VoiceProviderDropdownOption[];
+    className?: string;
+    selectId?: string;
+    resolveSwitchOptions?: (provider: SpeechProvider) => ProviderSwitchOptions | Promise<ProviderSwitchOptions>;
+    onProviderChanged?: (provider: SpeechProvider) => void;
+    onError?: (error: unknown, provider: SpeechProvider) => void;
+};
+export type VoiceProviderDropdownHandle = {
+    element: HTMLSelectElement;
+    destroy: () => void;
+    setProvider: (provider: SpeechProvider) => void;
+};
+export declare function buildSpeechProviderConfig(options: SpeechProviderConfigOptions): Pick<NavigationConfig, "speechProvider" | "azure" | "sarvam">;
+export declare function switchVoiceProvider(provider: SpeechProvider, options?: ProviderSwitchOptions): VoiceNavigationController;
+export declare function createVoiceProviderDropdown(config: VoiceProviderDropdownConfig): VoiceProviderDropdownHandle;
 export declare function initNavigationOnMicrophone(config: NavigationConfig): VoiceNavigationController;
 //# sourceMappingURL=index.d.ts.map

package/dist/index.d.ts.map

@@ -1 +1 @@
-{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,yBAAyB,EAAE,MAAM,4BAA4B,CAAC;AACvE,OAAO,EAAE,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AAC5D,OAAO,EAAE,kBAAkB,EAAE,MAAM,4BAA4B,CAAC;AAChE,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EAAE,YAAY,EAAE,MAAM,6BAA6B,CAAC;AAC3D,YAAY,EAAE,aAAa,EAAE,MAAM,6BAA6B,CAAC;AACjE,OAAO,EACL,kBAAkB,EAClB,kBAAkB,EAClB,mBAAmB,GACpB,MAAM,0BAA0B,CAAC;AAClC,YAAY,EAAE,aAAa,EAAE,MAAM,0BAA0B,CAAC;AAC9D,cAAc,YAAY,CAAC;AAC3B,cAAc,cAAc,CAAC;AAC7B,OAAO,EACL,qBAAqB,EACrB,oBAAoB,EACpB,QAAQ,EACR,YAAY,EACZ,gBAAgB,EAChB,qBAAqB,EACrB,SAAS,EACT,SAAS,GACV,MAAM,SAAS,CAAC;AAEjB,OAAO,EAAE,yBAAyB,EAAE,MAAM,4BAA4B,CAAC;AACvE,OAAO,KAAK,EAAE,gBAAgB,EAAE,MAAM,YAAY,CAAC;AA+JnD,wBAAgB,0BAA0B,CACxC,MAAM,EAAE,gBAAgB,GACvB,yBAAyB,CAoG3B"}
+{"version":3,"file":"index.d.ts","sourceRoot":"","sources":["../src/index.ts"],"names":[],"mappings":"AAMA,OAAO,EAAE,yBAAyB,EAAE,MAAM,4BAA4B,CAAC;AACvE,OAAO,EAAE,iBAAiB,EAAE,MAAM,yBAAyB,CAAC;AAC5D,OAAO,EAAE,kBAAkB,EAAE,MAAM,4BAA4B,CAAC;AAChE,OAAO,EAAE,mBAAmB,EAAE,MAAM,6BAA6B,CAAC;AAClE,OAAO,EAAE,cAAc,EAAE,MAAM,uBAAuB,CAAC;AACvD,OAAO,EAAE,YAAY,EAAE,MAAM,6BAA6B,CAAC;AAC3D,YAAY,EAAE,aAAa,EAAE,MAAM,6BAA6B,CAAC;AACjE,OAAO,EACL,kBAAkB,EAClB,kBAAkB,EAClB,mBAAmB,GACpB,MAAM,0BAA0B,CAAC;AAClC,YAAY,EAAE,aAAa,EAAE,MAAM,0BAA0B,CAAC;AAC9D,OAAO,EACL,gBAAgB,EAChB,oBAAoB,EACpB,oBAAoB,EACpB,gBAAgB,EAChB,cAAc,EACd,cAAc,EACd,kBAAkB,EAClB,qBAAqB,GACtB,MAAM,gBAAgB,CAAC;AACxB,YAAY,EAAE,cAAc,EAAE,MAAM,gBAAgB,CAAC;AACrD,cAAc,YAAY,CAAC;AAC3B,cAAc,cAAc,CAAC;AAC7B,OAAO,EACL,qBAAqB,EACrB,oBAAoB,EACpB,QAAQ,EACR,YAAY,EACZ,gBAAgB,EAChB,qBAAqB,EACrB,SAAS,EACT,SAAS,GACV,MAAM,SAAS,CAAC;AAEjB,OAAO,EAAE,yBAAyB,EAAE,MAAM,4BAA4B,CAAC;AACvE,OAAO,KAAK,EAAE,gBAAgB,EAAE,cAAc,EAAE,MAAM,YAAY,CAAC;AAiJnE,KAAK,2BAA2B,GAAG;IACjC,QAAQ,EAAE,cAAc,CAAC;IACzB,KAAK,CAAC,EAAE,gBAAgB,CAAC,OAAO,CAAC,CAAC;IAClC,MAAM,CAAC,EAAE,gBAAgB,CAAC,QAAQ,CAAC,CAAC;CACrC,CAAC;AAEF,MAAM,MAAM,qBAAqB,GAAG;IAClC,KAAK,CAAC,EAAE,gBAAgB,CAAC,OAAO,CAAC,CAAC;IAClC,MAAM,CAAC,EAAE,gBAAgB,CAAC,QAAQ,CAAC,CAAC;IACpC,QAAQ,CAAC,EAAE,gBAAgB,CAAC,UAAU,CAAC,CAAC;IACxC,aAAa,CAAC,EAAE,gBAAgB,CAAC,eAAe,CAAC,CAAC;CACnD,CAAC;AAEF,MAAM,MAAM,2BAA2B,GAAG;IACxC,QAAQ,EAAE,cAAc,CAAC;IACzB,KAAK,EAAE,MAAM,CAAC;IACd,QAAQ,CAAC,EAAE,OAAO,CAAC;CACpB,CAAC;AAEF,MAAM,MAAM,2BAA2B,GAAG;IACxC,SAAS,EAAE,MAAM,GAAG,WAAW,CAAC;IAChC,eAAe,CAAC,EAAE,cAAc,CAAC;IACjC,OAAO,CAAC,EAAE,2BAA2B,EAAE,CAAC;IACxC,SAAS,CAAC,EAAE,MAAM,CAAC;IACnB,QAAQ,CAAC,EAAE,MAAM,CAAC;IAClB,oBAAoB,CAAC,EAAE,CACrB,QAAQ,EAAE,cAAc,KACrB,qBAAqB,GAAG,OAAO,CAAC,qBAAqB,CAAC,CAAC;IAC5D,iBAAiB,CAAC,EAAE,CAAC,QAAQ,EAAE,cAAc,KAAK,IAAI,CAAC;IACvD,OAAO,CAAC,EAAE,CAAC,KAAK,EAAE,OAAO,EAAE,QAAQ,EAAE,cAAc,KAAK,IAAI,CAAC;CAC9D,CAAC;AAEF,MAAM,MAAM,2BAA2B,GAAG;IACxC,OAAO,EAAE,iBAAiB,CAAC;IAC3B,OAAO,EAAE,MAAM,IAAI,CAAC;IACpB,WAAW,EAAE,CAAC,QAAQ,EAAE,cAAc,KAAK,IAAI,CAAC;CACjD,CAAC;AAuDF,wBAAgB,yBAAyB,CACvC,OAAO,EAAE,2BAA2B,GACnC,IAAI,CAAC,gBAAgB,EAAE,gBAAgB,GAAG,OAAO,GAAG,QAAQ,CAAC,CAK/D;AAMD,wBAAgB,mBAAmB,CACjC,QAAQ,EAAE,cAAc,EACxB,OAAO,GAAE,qBAA0B,GAClC,yBAAyB,CAmB3B;AAKD,wBAAgB,2BAA2B,CACzC,MAAM,EAAE,2BAA2B,GAClC,2BAA2B,CAkH7B;AA4BD,wBAAgB,0BAA0B,CACxC,MAAM,EAAE,gBAAgB,GACvB,yBAAyB,CAwG3B"}

package/dist/index.js

@@ -1,9 +1,11 @@
 export { VoiceNavigationController } from "./navigation-controller.js";
 export { MicrophoneHandler } from "./microphone-handler.js";
 export { AzureSpeechService } from "./services/azure-speech.js";
+export { SarvamSpeechService } from "./services/sarvam-speech.js";
 export { BedrockService } from "./services/bedrock.js";
 export { PageRegistry } from "./services/page-registry.js";
 export { parseNavigationXML, fetchNavigationXML, loadNavigationPages, } from "./services/xml-parser.js";
+export { SARVAM_LANGUAGES, SARVAM_STT_LANGUAGES, SARVAM_TTS_LANGUAGES, DEFAULT_LANGUAGE, isSttSupported, isTtsSupported, resolveTtsLanguage, normalizeLanguageCode, } from "./languages.js";
 export * from "./types.js";
 export * from "./actions.js";
 export { createFloatingControl, attachToCustomButton, setState, updateStatus, updateTranscript, updateActionIndicator, showError, emitEvent, } from "./ui.js";
@@ -29,8 +31,12 @@ const buildPersistableConfig = (config) => {
     const hasCustomHandlers = Boolean(normalizedConfig.actionHandlers &&
         Object.keys(normalizedConfig.actionHandlers).length > 0);
     const persistable = {
+        speechProvider: normalizedConfig.speechProvider,
         aws: normalizedConfig.aws ? { ...normalizedConfig.aws } : undefined,
         azure: normalizedConfig.azure ? { ...normalizedConfig.azure } : undefined,
+        sarvam: normalizedConfig.sarvam
+            ? { ...normalizedConfig.sarvam }
+            : undefined,
         opensearch: normalizedConfig.opensearch
             ? { ...normalizedConfig.opensearch }
             : undefined,
@@ -87,6 +93,155 @@ const describeOpenSearchConfig = (config) => {
         apiPath,
     });
 };
+const resolveSpeechProviderConfig = (currentConfig, provider, options) => {
+    const merged = {
+        ...currentConfig,
+        ...options,
+        speechProvider: provider,
+        azure: {
+            ...(currentConfig.azure ?? {}),
+            ...(options?.azure ?? {}),
+        },
+        sarvam: {
+            ...(currentConfig.sarvam ?? {}),
+            ...(options?.sarvam ?? {}),
+        },
+    };
+    if (provider === "sarvam") {
+        const hasTtsUrl = Boolean(merged.sarvam?.ttsUrl);
+        const hasSttUrl = Boolean(merged.sarvam?.sttUrl);
+        if (!hasTtsUrl || !hasSttUrl) {
+            merged.sarvam = {
+                ...(merged.sarvam ?? {}),
+                ttsUrl: merged.sarvam?.ttsUrl ?? "/api/speech/sarvam",
+                sttUrl: merged.sarvam?.sttUrl ?? "/api/speech/sarvam-stt",
+            };
+        }
+    }
+    if (provider === "azure") {
+        const hasTokenEndpoint = Boolean(merged.azure?.tokenEndpoint);
+        const hasDirectCredentials = Boolean(merged.azure?.subscriptionKey && merged.azure?.region);
+        if (!hasTokenEndpoint && !hasDirectCredentials) {
+            merged.azure = {
+                ...(merged.azure ?? {}),
+                tokenEndpoint: "/api/speech/token",
+            };
+        }
+    }
+    return normalizeNavigationConfig(merged);
+};
+export function buildSpeechProviderConfig(options) {
+    return resolveSpeechProviderConfig({}, options.provider, {
+        azure: options.azure,
+        sarvam: options.sarvam,
+    });
+}
+export function switchVoiceProvider(provider, options = {}) {
+    if (typeof window === "undefined" || typeof document === "undefined") {
+        throw new Error("switchVoiceProvider requires a browser environment");
+    }
+    const globalWindow = window;
+    const currentRaw = globalWindow.__voiceNavigationConfig ||
+        readPersistedConfig() ||
+        {};
+    const currentConfig = normalizeNavigationConfig(currentRaw);
+    const nextConfig = resolveSpeechProviderConfig(currentConfig, provider, options);
+    return initNavigationOnMicrophone(nextConfig);
+}
+export function createVoiceProviderDropdown(config) {
+    if (typeof window === "undefined" || typeof document === "undefined") {
+        throw new Error("createVoiceProviderDropdown requires a browser environment");
+    }
+    const container = typeof config.container === "string"
+        ? document.querySelector(config.container)
+        : config.container;
+    if (!container) {
+        throw new Error("Voice provider dropdown container was not found");
+    }
+    const defaultOptions = [
+        { provider: "sarvam", label: "Sarvam" },
+        { provider: "azure", label: "Azure" },
+    ];
+    const options = config.options?.length ? config.options : defaultOptions;
+    const wrapper = document.createElement("div");
+    wrapper.className = config.className ?? "voice-provider-dropdown";
+    wrapper.setAttribute("data-voice-provider-dropdown", "true");
+    const select = document.createElement("select");
+    select.id = config.selectId ?? "voice-provider-select";
+    select.setAttribute("aria-label", "Voice provider");
+    options.forEach((option) => {
+        const opt = document.createElement("option");
+        opt.value = option.provider;
+        opt.textContent = option.label;
+        opt.disabled = Boolean(option.disabled);
+        select.appendChild(opt);
+    });
+    let currentProvider = config.initialProvider ?? "sarvam";
+    const activeOption = options.find((option) => option.provider === currentProvider);
+    if (!activeOption || activeOption.disabled) {
+        const fallback = options.find((option) => !option.disabled);
+        if (!fallback) {
+            throw new Error("No enabled voice providers are available for dropdown");
+        }
+        currentProvider = fallback.provider;
+    }
+    select.value = currentProvider;
+    let isSwitching = false;
+    const handleChange = async () => {
+        if (isSwitching) {
+            return;
+        }
+        const nextProvider = select.value;
+        if (nextProvider === currentProvider) {
+            return;
+        }
+        const option = options.find((item) => item.provider === nextProvider);
+        if (option?.disabled) {
+            select.value = currentProvider;
+            return;
+        }
+        isSwitching = true;
+        select.disabled = true;
+        try {
+            const switchOptions = config.resolveSwitchOptions
+                ? await config.resolveSwitchOptions(nextProvider)
+                : {};
+            switchVoiceProvider(nextProvider, switchOptions);
+            currentProvider = nextProvider;
+            config.onProviderChanged?.(nextProvider);
+        }
+        catch (error) {
+            select.value = currentProvider;
+            config.onError?.(error, nextProvider);
+        }
+        finally {
+            isSwitching = false;
+            select.disabled = false;
+        }
+    };
+    select.addEventListener("change", handleChange);
+    wrapper.appendChild(select);
+    const existing = container.querySelector("[data-voice-provider-dropdown='true']");
+    if (existing) {
+        existing.remove();
+    }
+    container.appendChild(wrapper);
+    return {
+        element: select,
+        destroy: () => {
+            select.removeEventListener("change", handleChange);
+            wrapper.remove();
+        },
+        setProvider: (provider) => {
+            const option = options.find((item) => item.provider === provider);
+            if (!option || option.disabled) {
+                return;
+            }
+            currentProvider = provider;
+            select.value = provider;
+        },
+    };
+}
 export function initNavigationOnMicrophone(config) {
     if (typeof window === "undefined" || typeof document === "undefined") {
         throw new Error("Voice navigation requires a browser environment");
@@ -129,11 +284,13 @@ export function initNavigationOnMicrophone(config) {
             previousUI?.customButtonSelector !== currentUI?.customButtonSelector ||
             previousUI?.buttonPlacement !== currentUI?.buttonPlacement;
         console.log("[Voice Navigation] UI config changed:", uiConfigChanged);
+        const providerChanged = previousConfig?.speechProvider !== normalizedConfig.speechProvider;
         if ((wantsVectorSearch &&
             (!hadVectorSearch ||
                 vectorConfigChanged ||
                 !existingHasVectorSearch)) ||
-            uiConfigChanged) {
+            uiConfigChanged ||
+            providerChanged) {
             console.log("[Voice Navigation] Destroying existing controller and recreating...");
             try {
                 existingController?.destroy?.();

package/dist/languages.d.ts

@@ -0,0 +1,17 @@
+export interface SarvamLanguage {
+    code: string;
+    name: string;
+    nativeName: string;
+    script: string;
+    stt: boolean;
+    tts: boolean;
+}
+export declare const SARVAM_LANGUAGES: SarvamLanguage[];
+export declare const DEFAULT_LANGUAGE = "en-IN";
+export declare const SARVAM_STT_LANGUAGES: string[];
+export declare const SARVAM_TTS_LANGUAGES: string[];
+export declare function isSttSupported(code: string): boolean;
+export declare function isTtsSupported(code: string): boolean;
+export declare function resolveTtsLanguage(code: string): string;
+export declare function normalizeLanguageCode(input?: string | null): string;
+//# sourceMappingURL=languages.d.ts.map

package/dist/languages.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"languages.d.ts","sourceRoot":"","sources":["../src/languages.ts"],"names":[],"mappings":"AAcA,MAAM,WAAW,cAAc;IAE7B,IAAI,EAAE,MAAM,CAAC;IAEb,IAAI,EAAE,MAAM,CAAC;IAEb,UAAU,EAAE,MAAM,CAAC;IAEnB,MAAM,EAAE,MAAM,CAAC;IAEf,GAAG,EAAE,OAAO,CAAC;IAEb,GAAG,EAAE,OAAO,CAAC;CACd;AAMD,eAAO,MAAM,gBAAgB,EAAE,cAAc,EAwB5C,CAAC;AAGF,eAAO,MAAM,gBAAgB,UAAU,CAAC;AAGxC,eAAO,MAAM,oBAAoB,EAAE,MAAM,EAEf,CAAC;AAG3B,eAAO,MAAM,oBAAoB,EAAE,MAAM,EAEf,CAAC;AAuC3B,wBAAgB,cAAc,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAEpD;AAGD,wBAAgB,cAAc,CAAC,IAAI,EAAE,MAAM,GAAG,OAAO,CAEpD;AAOD,wBAAgB,kBAAkB,CAAC,IAAI,EAAE,MAAM,GAAG,MAAM,CAGvD;AAOD,wBAAgB,qBAAqB,CAAC,KAAK,CAAC,EAAE,MAAM,GAAG,IAAI,GAAG,MAAM,CA0BnE"}

package/dist/languages.js

@@ -0,0 +1,86 @@
+export const SARVAM_LANGUAGES = [
+    { code: "en-IN", name: "English (India)", nativeName: "English", script: "Latin", stt: true, tts: true },
+    { code: "hi-IN", name: "Hindi", nativeName: "हिन्दी", script: "Devanagari", stt: true, tts: true },
+    { code: "bn-IN", name: "Bengali", nativeName: "বাংলা", script: "Bengali", stt: true, tts: true },
+    { code: "ta-IN", name: "Tamil", nativeName: "தமிழ்", script: "Tamil", stt: true, tts: true },
+    { code: "te-IN", name: "Telugu", nativeName: "తెలుగు", script: "Telugu", stt: true, tts: true },
+    { code: "gu-IN", name: "Gujarati", nativeName: "ગુજરાતી", script: "Gujarati", stt: true, tts: true },
+    { code: "kn-IN", name: "Kannada", nativeName: "ಕನ್ನಡ", script: "Kannada", stt: true, tts: true },
+    { code: "ml-IN", name: "Malayalam", nativeName: "മലയാളം", script: "Malayalam", stt: true, tts: true },
+    { code: "mr-IN", name: "Marathi", nativeName: "मराठी", script: "Devanagari", stt: true, tts: true },
+    { code: "pa-IN", name: "Punjabi", nativeName: "ਪੰਜਾਬੀ", script: "Gurmukhi", stt: true, tts: true },
+    { code: "od-IN", name: "Odia", nativeName: "ଓଡ଼ିଆ", script: "Odia", stt: true, tts: true },
+    { code: "as-IN", name: "Assamese", nativeName: "অসমীয়া", script: "Bengali", stt: true, tts: false },
+    { code: "ur-IN", name: "Urdu", nativeName: "اردو", script: "Perso-Arabic", stt: true, tts: false },
+    { code: "ne-IN", name: "Nepali", nativeName: "नेपाली", script: "Devanagari", stt: true, tts: false },
+    { code: "kok-IN", name: "Konkani", nativeName: "कोंकणी", script: "Devanagari", stt: true, tts: false },
+    { code: "ks-IN", name: "Kashmiri", nativeName: "کٲشُر", script: "Perso-Arabic", stt: true, tts: false },
+    { code: "sd-IN", name: "Sindhi", nativeName: "سنڌي", script: "Perso-Arabic", stt: true, tts: false },
+    { code: "sa-IN", name: "Sanskrit", nativeName: "संस्कृतम्", script: "Devanagari", stt: true, tts: false },
+    { code: "sat-IN", name: "Santali", nativeName: "ᱥᱟᱱᱛᱟᱲᱤ", script: "Ol Chiki", stt: true, tts: false },
+    { code: "mni-IN", name: "Manipuri", nativeName: "ꯃꯤꯇꯩ ꯂꯣꯟ", script: "Meitei", stt: true, tts: false },
+    { code: "brx-IN", name: "Bodo", nativeName: "बड़ो", script: "Devanagari", stt: true, tts: false },
+    { code: "mai-IN", name: "Maithili", nativeName: "मैथिली", script: "Devanagari", stt: true, tts: false },
+    { code: "doi-IN", name: "Dogri", nativeName: "डोगरी", script: "Devanagari", stt: true, tts: false },
+];
+export const DEFAULT_LANGUAGE = "en-IN";
+export const SARVAM_STT_LANGUAGES = SARVAM_LANGUAGES.filter((lang) => lang.stt).map((lang) => lang.code);
+export const SARVAM_TTS_LANGUAGES = SARVAM_LANGUAGES.filter((lang) => lang.tts).map((lang) => lang.code);
+const LANGUAGE_BY_CODE = new Map(SARVAM_LANGUAGES.map((lang) => [lang.code, lang]));
+const APP_LOCALE_ALIASES = {
+    en: "en-IN",
+    hi: "hi-IN",
+    bn: "bn-IN",
+    ta: "ta-IN",
+    te: "te-IN",
+    gu: "gu-IN",
+    kn: "kn-IN",
+    ml: "ml-IN",
+    mr: "mr-IN",
+    pa: "pa-IN",
+    or: "od-IN",
+    od: "od-IN",
+    as: "as-IN",
+    ur: "ur-IN",
+    ne: "ne-IN",
+    kok: "kok-IN",
+    ks: "ks-IN",
+    sd: "sd-IN",
+    sa: "sa-IN",
+    sat: "sat-IN",
+    mni: "mni-IN",
+    brx: "brx-IN",
+    mai: "mai-IN",
+    doi: "doi-IN",
+};
+export function isSttSupported(code) {
+    return LANGUAGE_BY_CODE.get(normalizeLanguageCode(code))?.stt ?? false;
+}
+export function isTtsSupported(code) {
+    return LANGUAGE_BY_CODE.get(normalizeLanguageCode(code))?.tts ?? false;
+}
+export function resolveTtsLanguage(code) {
+    const normalized = normalizeLanguageCode(code);
+    return LANGUAGE_BY_CODE.get(normalized)?.tts ? normalized : DEFAULT_LANGUAGE;
+}
+export function normalizeLanguageCode(input) {
+    if (!input) {
+        return DEFAULT_LANGUAGE;
+    }
+    const trimmed = input.trim();
+    if (!trimmed) {
+        return DEFAULT_LANGUAGE;
+    }
+    const lowered = trimmed.toLowerCase();
+    for (const lang of SARVAM_LANGUAGES) {
+        if (lang.code.toLowerCase() === lowered) {
+            return lang.code;
+        }
+    }
+    const baseLocale = lowered.split("-")[0];
+    const aliased = APP_LOCALE_ALIASES[lowered] ?? APP_LOCALE_ALIASES[baseLocale];
+    if (aliased) {
+        return aliased;
+    }
+    return DEFAULT_LANGUAGE;
+}

package/dist/microphone-handler.d.ts

@@ -1,15 +1,20 @@
 import * as sdk from "microsoft-cognitiveservices-speech-sdk";
-import type { NavigationCallbacks, AgentActionResponse } from "./types.js";
+import type { NavigationCallbacks, AgentActionResponse, SpeechProvider } from "./types.js";
 import { AzureSpeechService } from "./services/azure-speech.js";
 import { BedrockService } from "./services/bedrock.js";
+import { SarvamSpeechService } from "./services/sarvam-speech.js";
 interface MicrophoneConfig {
-    azureSpeechService: AzureSpeechService;
+    speechProvider?: SpeechProvider;
+    azureSpeechService?: AzureSpeechService;
+    sarvamSpeechService?: SarvamSpeechService;
     bedrockService: BedrockService;
     language?: string;
     silenceTimeout?: number;
 }
 export declare class MicrophoneHandler {
+    private speechProvider;
     private azureSpeechService;
+    private sarvamSpeechService;
     private bedrockService;
     private language;
     private silenceTimeout;
@@ -50,6 +55,7 @@ export declare class MicrophoneHandler {
     disposeRecognizer(): void;
     createRecognizer(): Promise<sdk.SpeechRecognizer>;
     startRecording(callbacks?: NavigationCallbacks): Promise<boolean>;
+    private startSarvamRecording;
     stopRecording(): Promise<string>;
     static isMicrophoneAvailable(): Promise<boolean>;
 }