Sarvam AI

Overview

Sarvam AI provides a modular speech and language platform with deep support for Indian languages and code-mixed speech. Unlike bundled voice-agent APIs, the Teler bridge orchestrates three independent Sarvam services — speech-to-text, a conversational LLM, and text-to-speech — giving you full control over each stage rather than a single opaque pipeline.

Each stage is independently configurable: choose the STT model, LLM, TTS model, language, and speaker separately. If the primary LLM becomes unavailable, the bridge automatically falls back to a secondary model, so a single provider hiccup doesn’t take the whole call down.

Key Features

• Independent pipeline stages — STT, LLM, and TTS are three separate, independently configurable services, not one bundled black box — swap any one stage without touching the others.
• Automatic LLM fallback — If the primary conversational model becomes unavailable mid-call, the bridge automatically retries against a secondary model, rather than dropping the response entirely.
• Built for Indian languages and code-mixed speech — Language and speaker are both configurable per deployment, with native support for code-mixed input (e.g., Hindi-English) rather than treating it as an edge case.
• Barge-in support out of the box — When the caller speaks over the agent, the bridge clears buffered audio so the caller can interrupt naturally.

How to Integrate

Prerequisites

• A Sarvam account with an API key
• A Teler account with an active API key
• One Teler phone number on your account
• An ngrok auth token (for local development)
• Docker and Docker Compose

Configuration

The STT model, LLM (plus fallback), TTS model, language, speaker, and greeting are all configurable too, each with a sensible default — see the repo’s .env.example for the full list.

Step 1: 5-minute setup

git clone https://github.com/frejun-tech/teler-sarvam-node-bridge
cd teler-sarvam-node-bridge
cp .env.example .env
docker compose up -d --build

Fill in your .env with the required variables above before starting the containers.

Step 2: Initiate a call

curl -X POST https://your-domain/api/v1/calls/initiate-call \
  -H "Content-Type: application/json" \
  -d '{
    "fromNumber": "+91XXXXXXXXXX",
    "toNumber":   "+91XXXXXXXXXX"
  }'

How the pipeline works, end to end

• Caller audio in — Raw audio frames from the Teler call are forwarded to Sarvam’s realtime STT endpoint, transcribed according to SARVAM_STT_MODEL.
• Transcription → LLM — The transcribed text is sent to Sarvam’s chat-completions API, which generates a response using the primary LLM (SARVAM_LLM_MODEL). If that model is unavailable, the bridge falls back to SARVAM_LLM_MODEL_FALLBACK.
• LLM response → TTS — The generated text is streamed to Sarvam’s realtime TTS endpoint, synthesized to speech using the configured SARVAM_TTS_MODEL and SARVAM_SPEAKER.
• Agent audio out — Synthesized audio is forwarded back to the Teler call as binary frames.
• Barge-in — When the caller speaks over the agent, the bridge clears buffered audio to allow interruption.
• Greeting — A STATIC_RESPONSE_TEXT greeting is played when the call first connects.

Troubleshooting

Integration at a Glance