agents-plugin-sarvam

Sarvam AI plugin for LiveKit Agents.

Provides text-to-speech (TTS) using Sarvam AI's Bulbul models and speech-to-text (STT) using Saaras/Saarika models, with support for 22+ Indian languages.

Installation

pnpm add @livekit/agents-plugin-sarvam

Usage

TTS

import * as sarvam from '@livekit/agents-plugin-sarvam';

const tts = new sarvam.TTS({
  speaker: 'anushka',
  model: 'bulbul:v2',
  targetLanguageCode: 'en-IN',
});

TTS (Prefer non-streaming REST over WebSocket)

import * as sarvam from '@livekit/agents-plugin-sarvam';

const tts = new sarvam.TTS({
  model: 'bulbul:v2',
  streaming: false,
});

Setting streaming: false disables the plugin's native WebSocket streaming path. In an AgentSession, LiveKit will use sentence chunking + non-streaming synthesize() calls instead.

STT (Transcribe)

import * as sarvam from '@livekit/agents-plugin-sarvam';

const stt = new sarvam.STT({
  model: 'saaras:v3',
  languageCode: 'en-IN',
  mode: 'transcribe',
});

STT (Prefer non-streaming REST over WebSocket)

import * as sarvam from '@livekit/agents-plugin-sarvam';

const stt = new sarvam.STT({
  model: 'saaras:v3',
  streaming: false,
});

Setting streaming: false disables the plugin's native WebSocket streaming path. In an AgentSession, LiveKit will use VAD + non-streaming recognize() calls instead.

STT (Translate — Indic to English)

import * as sarvam from '@livekit/agents-plugin-sarvam';

const stt = new sarvam.STT({
  model: 'saaras:v2.5',
  prompt: 'Technical interview discussion',
});

Set the SARVAM_API_KEY environment variable or pass apiKey directly.

STT Models

| Model | Endpoint | Languages | Modes | Prompt | | --------------------- | --------------------------- | ----------- | ----- | ------ | | saaras:v3 (default) | /speech-to-text | 22 (BCP-47) | Yes | No | | saaras:v2.5 | /speech-to-text-translate | Auto-detect | No | Yes | | saarika:v2.5 | /speech-to-text | 12 (BCP-47) | No | No |

saarika:v2.5 will be deprecated soon — all its languages are available in saaras:v3.

Supported Languages

TTS (Bulbul models)

Bengali (bn-IN), English (en-IN), Gujarati (gu-IN), Hindi (hi-IN), Kannada (kn-IN), Malayalam (ml-IN), Marathi (mr-IN), Odia (od-IN), Punjabi (pa-IN), Tamil (ta-IN), Telugu (te-IN).

STT (Saaras v3)

All TTS languages plus: Assamese (as-IN), Bodo (brx-IN), Dogri (doi-IN), Kashmiri (ks-IN), Konkani (kok-IN), Maithili (mai-IN), Manipuri (mni-IN), Nepali (ne-IN), Sanskrit (sa-IN), Santali (sat-IN), Sindhi (sd-IN), Urdu (ur-IN).

Set languageCode to 'unknown' for automatic language detection.

STT Modes (Saaras v3)

| Mode | Description | | ------------ | ------------------------------------------------------------------ | | transcribe | Standard transcription with formatting and normalization (default) | | translate | Direct speech-to-English translation | | verbatim | Exact word-for-word transcription | | translit | Romanization to Latin script | | codemix | Mixed script (English words in English, Indic in native script) |