Sayna SDK

Client and server SDKs for real-time voice streaming in AI agents

This repository contains the official Sayna SDKs for adding voice capabilities to existing AI agents. Sayna provides a unified voice streaming layer that handles speech-to-text (STT) and text-to-speech (TTS) interactions, abstracting the complexity of managing real-time voice pipelines.

What is Sayna?

Sayna enables AI agents to support natural voice conversations by providing:

• Real-time audio streaming over WebSockets
• Multi-provider support for STT (Deepgram, Google) and TTS (ElevenLabs, Google, Deepgram)
• Optional LiveKit integration for multi-participant voice rooms
• Low-latency voice pipeline management
• Type-safe client libraries

Available SDKs

This monorepo contains three SDKs:

• JavaScript SDK - Browser-based client for connecting to Sayna voice rooms
• Node.js SDK - Server-side SDK for Node.js applications
• Python SDK - Async Python SDK for server-side voice streaming

Each SDK has its own README with detailed installation instructions and API documentation.

Repository Structure

saysdk/
├── js-sdk/              # Browser JavaScript SDK
├── node-sdk/            # Node.js server SDK
├── python-sdk/          # Python server SDK
├── api-reference.md     # Complete API documentation
└── README.md            # This file

Documentation

• API Reference - Complete REST and WebSocket API documentation
• JavaScript SDK - Browser client documentation
• Node.js SDK - Node.js SDK documentation
• Python SDK - Python SDK documentation

Use Cases

• Voice-enabled chatbots and conversational AI
• AI-powered phone systems and telephony applications
• Multi-participant voice rooms with AI agents
• Real-time transcription services
• Voice synthesis and speech generation

License

Apache License 2.0 - see LICENSE for details