@mastra/voice-sarvam 0.1.2 → 0.1.3-alpha.1

package/README.md

@@ -1,6 +1,6 @@
 # @mastra/voice-sarvam
 
-Sarvam Voice integration for Mastra, providing Text-to-Speech (TTS) capabilities using Sarvam's voice technology.
+Sarvam Voice integration for Mastra, providing Text-to-Speech (TTS) and Speech-to-text (STT) capabilities using Sarvam's voice technology.
 
 ## Installation
 
@@ -21,17 +21,24 @@ SARVAM_API_KEY=your_api_key
 ```typescript
 import { SarvamVoice } from '@mastra/voice-sarvam';
 
-const voice = new CompositeVoice({
-  speakProvider: new SarvamVoice({
-    speechModel: { apiKey: 'YOUR-API-KEY' },
-    speaker: 'meera',
-  }),
+const voice = new SarvamVoice({
+  speechModel: {
+    model: 'bulbul:v1',
+    apiKey: process.env.SARVAM_API_KEY!,
+    language: 'en-IN',
+  },
+  listeningModel: {
+    apiKey: process.env.SARVAM_API_KEY!,
+    model: 'saarika:v2',
+    languageCode: 'unknown', // By default only works with saarika:v2
+  },
+  speaker: 'meera',
 });
 
 // Create an agent with voice capabilities
 export const agent = new Agent({
   name: 'Agent',
-  instructions: `You are a helpful assistant with voice capabilities.`,
+  instructions: `You are a helpful assistant with both TTS and STT capabilities.`,
   model: google('gemini-1.5-pro-latest'),
   voice: voice,
 });
@@ -41,7 +48,7 @@ const speakers = await voice.getSpeakers();
 
 // Generate speech and save to file
 const audio = await agent.speak("Hello, I'm your AI assistant!");
-const filePath = path.join(process.cwd(), 'agent.mp3');
+const filePath = path.join(process.cwd(), 'agent.wav');
 const writer = createWriteStream(filePath);
 
 audio.pipe(writer);
@@ -62,11 +69,14 @@ const streamWriter = createWriteStream(streamFilePath);
 audioStream.pipe(streamWriter);
 
 console.log(`Speech saved to ${filePath} and ${streamFilePath}`);
+
+// Generate Text from an audio stream
+const text = await voice.listen(audioStream);
 ```
 
 ## Features
 
-- High-quality Text-to-Speech synthesis
+- High-quality Text-to-Speech and Speech-to-Text synthesis
 - Support for 10+ Indian languages
 - Choice of 10+ diverse speakers
 - Advanced voice customization options

package/dist/_tsup-dts-rollup.d.cts

@@ -1,14 +1,29 @@
 import { MastraVoice } from '@mastra/core/voice';
 
-export declare const SARVAM_LANGUAGES: readonly ["hi-IN", "bn-IN", "kn-IN", "ml-IN", "mr-IN", "od-IN", "pa-IN", "ta-IN", "te-IN", "en-IN", "gu-IN"];
+export declare const SARVAM_STT_LANGUAGES: readonly ["hi-IN", "bn-IN", "kn-IN", "ml-IN", "mr-IN", "od-IN", "pa-IN", "ta-IN", "te-IN", "en-IN", "gu-IN", "unknown"];
 
-export declare const SARVAM_MODELS: readonly ["bulbul:v1"];
+export declare const SARVAM_STT_MODELS: readonly ["saarika:v1", "saarika:v2", "saarika:flash"];
+
+export declare const SARVAM_TTS_LANGUAGES: readonly ["hi-IN", "bn-IN", "kn-IN", "ml-IN", "mr-IN", "od-IN", "pa-IN", "ta-IN", "te-IN", "en-IN", "gu-IN"];
+
+export declare const SARVAM_TTS_MODELS: readonly ["bulbul:v1"];
 
 export declare const SARVAM_VOICES: readonly ["meera", "pavithra", "maitreyi", "arvind", "amol", "amartya", "diya", "neel", "misha", "vian", "arjun", "maya"];
 
-export declare type SarvamLanguage = (typeof SARVAM_LANGUAGES)[number];
+declare interface SarvamListenOptions {
+    apiKey?: string;
+    model?: SarvamSTTModel;
+    languageCode?: SarvamSTTLanguage;
+    filetype?: 'mp3' | 'wav';
+}
+
+export declare type SarvamSTTLanguage = (typeof SARVAM_STT_LANGUAGES)[number];
+
+export declare type SarvamSTTModel = (typeof SARVAM_STT_MODELS)[number];
+
+export declare type SarvamTTSLanguage = (typeof SARVAM_TTS_LANGUAGES)[number];
 
-export declare type SarvamModel = (typeof SARVAM_MODELS)[number];
+export declare type SarvamTTSModel = (typeof SARVAM_TTS_MODELS)[number];
 
 export declare class SarvamVoice extends MastraVoice {
     private apiKey?;
@@ -17,9 +32,10 @@ export declare class SarvamVoice extends MastraVoice {
     private properties;
     protected speaker: SarvamVoiceId;
     private baseUrl;
-    constructor({ speechModel, speaker, }?: {
+    constructor({ speechModel, speaker, listeningModel, }?: {
         speechModel?: SarvamVoiceConfig;
         speaker?: SarvamVoiceId;
+        listeningModel?: SarvamListenOptions;
     });
     private makeRequest;
     private streamToString;
@@ -29,13 +45,13 @@ export declare class SarvamVoice extends MastraVoice {
     getSpeakers(): Promise<{
         voiceId: "meera" | "pavithra" | "maitreyi" | "arvind" | "amol" | "amartya" | "diya" | "neel" | "misha" | "vian" | "arjun" | "maya";
     }[]>;
-    listen(_input: NodeJS.ReadableStream, _options?: Record<string, unknown>): Promise<string | NodeJS.ReadableStream>;
+    listen(input: NodeJS.ReadableStream, options?: SarvamListenOptions): Promise<string>;
 }
 
 declare interface SarvamVoiceConfig {
     apiKey?: string;
-    model?: SarvamModel;
-    language?: SarvamLanguage;
+    model?: SarvamTTSModel;
+    language?: SarvamTTSLanguage;
     properties?: {
         pitch?: number;
         pace?: number;

package/dist/_tsup-dts-rollup.d.ts

@@ -1,14 +1,29 @@
 import { MastraVoice } from '@mastra/core/voice';
 
-export declare const SARVAM_LANGUAGES: readonly ["hi-IN", "bn-IN", "kn-IN", "ml-IN", "mr-IN", "od-IN", "pa-IN", "ta-IN", "te-IN", "en-IN", "gu-IN"];
+export declare const SARVAM_STT_LANGUAGES: readonly ["hi-IN", "bn-IN", "kn-IN", "ml-IN", "mr-IN", "od-IN", "pa-IN", "ta-IN", "te-IN", "en-IN", "gu-IN", "unknown"];
 
-export declare const SARVAM_MODELS: readonly ["bulbul:v1"];
+export declare const SARVAM_STT_MODELS: readonly ["saarika:v1", "saarika:v2", "saarika:flash"];
+
+export declare const SARVAM_TTS_LANGUAGES: readonly ["hi-IN", "bn-IN", "kn-IN", "ml-IN", "mr-IN", "od-IN", "pa-IN", "ta-IN", "te-IN", "en-IN", "gu-IN"];
+
+export declare const SARVAM_TTS_MODELS: readonly ["bulbul:v1"];
 
 export declare const SARVAM_VOICES: readonly ["meera", "pavithra", "maitreyi", "arvind", "amol", "amartya", "diya", "neel", "misha", "vian", "arjun", "maya"];
 
-export declare type SarvamLanguage = (typeof SARVAM_LANGUAGES)[number];
+declare interface SarvamListenOptions {
+    apiKey?: string;
+    model?: SarvamSTTModel;
+    languageCode?: SarvamSTTLanguage;
+    filetype?: 'mp3' | 'wav';
+}
+
+export declare type SarvamSTTLanguage = (typeof SARVAM_STT_LANGUAGES)[number];
+
+export declare type SarvamSTTModel = (typeof SARVAM_STT_MODELS)[number];
+
+export declare type SarvamTTSLanguage = (typeof SARVAM_TTS_LANGUAGES)[number];
 
-export declare type SarvamModel = (typeof SARVAM_MODELS)[number];
+export declare type SarvamTTSModel = (typeof SARVAM_TTS_MODELS)[number];
 
 export declare class SarvamVoice extends MastraVoice {
     private apiKey?;
@@ -17,9 +32,10 @@ export declare class SarvamVoice extends MastraVoice {
     private properties;
     protected speaker: SarvamVoiceId;
     private baseUrl;
-    constructor({ speechModel, speaker, }?: {
+    constructor({ speechModel, speaker, listeningModel, }?: {
         speechModel?: SarvamVoiceConfig;
         speaker?: SarvamVoiceId;
+        listeningModel?: SarvamListenOptions;
     });
     private makeRequest;
     private streamToString;
@@ -29,13 +45,13 @@ export declare class SarvamVoice extends MastraVoice {
     getSpeakers(): Promise<{
         voiceId: "meera" | "pavithra" | "maitreyi" | "arvind" | "amol" | "amartya" | "diya" | "neel" | "misha" | "vian" | "arjun" | "maya";
     }[]>;
-    listen(_input: NodeJS.ReadableStream, _options?: Record<string, unknown>): Promise<string | NodeJS.ReadableStream>;
+    listen(input: NodeJS.ReadableStream, options?: SarvamListenOptions): Promise<string>;
 }
 
 declare interface SarvamVoiceConfig {
     apiKey?: string;
-    model?: SarvamModel;
-    language?: SarvamLanguage;
+    model?: SarvamTTSModel;
+    language?: SarvamTTSLanguage;
     properties?: {
         pitch?: number;
         pace?: number;

package/dist/index.cjs

@@ -22,6 +22,14 @@ var SARVAM_VOICES = [
 ];
 
 // src/index.ts
+var defaultSpeechModel = {
+  model: "bulbul:v1",
+  apiKey: process.env.SARVAM_API_KEY,
+  language: "en-IN"
+};
+var defaultListeningModel = {
+  model: "saarika:v2",
+  apiKey: process.env.SARVAM_API_KEY};
 var SarvamVoice = class extends voice.MastraVoice {
   apiKey;
   model = "bulbul:v1";
@@ -31,18 +39,18 @@ var SarvamVoice = class extends voice.MastraVoice {
   baseUrl = "https://api.sarvam.ai";
   constructor({
     speechModel,
-    speaker
+    speaker,
+    listeningModel
   } = {}) {
-    const defaultSpeechModel = {
-      model: "bulbul:v1",
-      apiKey: process.env.SARVAM_API_KEY,
-      language: "en-IN"
-    };
     super({
       speechModel: {
         name: speechModel?.model ?? defaultSpeechModel.model,
         apiKey: speechModel?.apiKey ?? defaultSpeechModel.apiKey
       },
+      listeningModel: {
+        name: listeningModel?.model ?? defaultListeningModel.model,
+        apiKey: listeningModel?.model ?? defaultListeningModel.apiKey
+      },
       speaker
     });
     this.apiKey = speechModel?.apiKey || defaultSpeechModel.apiKey;
@@ -116,9 +124,39 @@ var SarvamVoice = class extends voice.MastraVoice {
       }));
     }, "voice.deepgram.getSpeakers")();
   }
-  //Todo: Implement the listen method
-  async listen(_input, _options) {
-    throw new Error("Listening method coming soon.");
+  async listen(input, options) {
+    return this.traced(async () => {
+      const chunks = [];
+      for await (const chunk of input) {
+        if (typeof chunk === "string") {
+          chunks.push(Buffer.from(chunk));
+        } else {
+          chunks.push(chunk);
+        }
+      }
+      const audioBuffer = Buffer.concat(chunks);
+      const form = new FormData();
+      const mimeType = options?.filetype === "mp3" ? "audio/mpeg" : "audio/wav";
+      const blob = new Blob([audioBuffer], { type: mimeType });
+      form.append("file", blob);
+      form.append("model", options?.model || "saarika:v2");
+      form.append("language_code", options?.languageCode || "unknown");
+      const requestOptions = {
+        method: "POST",
+        headers: {
+          "api-subscription-key": this.apiKey
+        },
+        body: form
+      };
+      try {
+        const response = await fetch(`${this.baseUrl}/speech-to-text`, requestOptions);
+        const result = await response.json();
+        return result.transcript;
+      } catch (error) {
+        console.error("Error during speech-to-text request:", error);
+        throw error;
+      }
+    }, "voice.sarvam.listen")();
   }
 };
 

package/dist/index.js

@@ -20,6 +20,14 @@ var SARVAM_VOICES = [
 ];
 
 // src/index.ts
+var defaultSpeechModel = {
+  model: "bulbul:v1",
+  apiKey: process.env.SARVAM_API_KEY,
+  language: "en-IN"
+};
+var defaultListeningModel = {
+  model: "saarika:v2",
+  apiKey: process.env.SARVAM_API_KEY};
 var SarvamVoice = class extends MastraVoice {
   apiKey;
   model = "bulbul:v1";
@@ -29,18 +37,18 @@ var SarvamVoice = class extends MastraVoice {
   baseUrl = "https://api.sarvam.ai";
   constructor({
     speechModel,
-    speaker
+    speaker,
+    listeningModel
   } = {}) {
-    const defaultSpeechModel = {
-      model: "bulbul:v1",
-      apiKey: process.env.SARVAM_API_KEY,
-      language: "en-IN"
-    };
     super({
       speechModel: {
         name: speechModel?.model ?? defaultSpeechModel.model,
         apiKey: speechModel?.apiKey ?? defaultSpeechModel.apiKey
       },
+      listeningModel: {
+        name: listeningModel?.model ?? defaultListeningModel.model,
+        apiKey: listeningModel?.model ?? defaultListeningModel.apiKey
+      },
       speaker
     });
     this.apiKey = speechModel?.apiKey || defaultSpeechModel.apiKey;
@@ -114,9 +122,39 @@ var SarvamVoice = class extends MastraVoice {
       }));
     }, "voice.deepgram.getSpeakers")();
   }
-  //Todo: Implement the listen method
-  async listen(_input, _options) {
-    throw new Error("Listening method coming soon.");
+  async listen(input, options) {
+    return this.traced(async () => {
+      const chunks = [];
+      for await (const chunk of input) {
+        if (typeof chunk === "string") {
+          chunks.push(Buffer.from(chunk));
+        } else {
+          chunks.push(chunk);
+        }
+      }
+      const audioBuffer = Buffer.concat(chunks);
+      const form = new FormData();
+      const mimeType = options?.filetype === "mp3" ? "audio/mpeg" : "audio/wav";
+      const blob = new Blob([audioBuffer], { type: mimeType });
+      form.append("file", blob);
+      form.append("model", options?.model || "saarika:v2");
+      form.append("language_code", options?.languageCode || "unknown");
+      const requestOptions = {
+        method: "POST",
+        headers: {
+          "api-subscription-key": this.apiKey
+        },
+        body: form
+      };
+      try {
+        const response = await fetch(`${this.baseUrl}/speech-to-text`, requestOptions);
+        const result = await response.json();
+        return result.transcript;
+      } catch (error) {
+        console.error("Error during speech-to-text request:", error);
+        throw error;
+      }
+    }, "voice.sarvam.listen")();
   }
 };
 

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@mastra/voice-sarvam",
-  "version": "0.1.2",
+  "version": "0.1.3-alpha.1",
   "description": "Mastra Sarvam AI voice integration",
   "type": "module",
   "files": [
@@ -23,7 +23,7 @@
   },
   "dependencies": {
     "zod": "^3.24.2",
-    "@mastra/core": "^0.6.3"
+    "@mastra/core": "^0.6.4-alpha.1"
   },
   "devDependencies": {
     "@microsoft/api-extractor": "^7.52.1",
@@ -34,6 +34,16 @@
     "vitest": "^2.1.9",
     "@internal/lint": "0.0.1"
   },
+  "keywords": [
+    "mastra",
+    "sarvam",
+    "tts",
+    "stt",
+    "indian-languages",
+    "speech-to-text",
+    "text-to-speech",
+    "speech-recognition"
+  ],
   "scripts": {
     "build": "tsup src/index.ts --format esm,cjs --experimental-dts --clean --treeshake=smallest --splitting",
     "build:watch": "pnpm build --watch",