Sharyx OS 🎙️

The Open-Source Framework for Building Production-Grade Real-Time AI Voice Agents.

Sharyx OS is an orchestration engine that connects Speech-to-Text (STT), Large Language Models (LLM), and Text-to-Speech (TTS) into a seamless, low-latency conversation loop.

🚀 Key Features

• ⚡ Low Latency: Optimized streaming architectures for near-instant responses.
• 🔄 Recursive Tool Calling: Enable your agents to search, book, and act in real-time.
• 🔌 Provider Agnostic: Swap between OpenAI, Gemini, Deepgram, ElevenLabs, and more with one line of code.
• 📞 Multi-Channel: Native support for Twilio, Plivo, and WebRTC.
• 🧠 Persistent Memory: Redis-backed session management for stateful conversations.

📦 Quick Start

1. Install

npm install sharyx-os

2. Create your first agent

import { createAgent, OpenAILLM, DeepgramSTT, ElevenLabsTTS } from 'sharyx-os';

const agent = createAgent({
  stt: new DeepgramSTT({ apiKey: process.env.DEEPGRAM_API_KEY }),
  llm: new OpenAILLM({ apiKey: process.env.OPENAI_API_KEY }),
  tts: new ElevenLabsTTS({ apiKey: process.env.ELEVEN_LABS_API_KEY }),
  systemPrompt: "You are a helpful assistant for a medical clinic."
});

agent.start({ port: 3000 });

🧩 Architecture

graph LR
    User((User)) --> Transport[Telephony/Web]
    Transport -- Audio --> STT[Deepgram]
    STT -- Text --> Orchestrator
    Orchestrator -- Context --> LLM[GPT-4/Gemini]
    LLM -- Action --> Tools[Google Cal/HubSpot]
    LLM -- Response --> TTS[ElevenLabs/Cartesia]
    TTS -- Audio --> Transport
    Transport --> User

🛠️ Included Integrations

• CRM: HubSpot Lead Capture.
• Calendar: Google Calendar Appointment Booking.
• Messaging: WhatsApp Cloud API Notifications.

🗺️ Roadmap & Community

We are building the future of voice-first interfaces. Want to help?

• Check out CONTRIBUTING.md to add new providers.
• Join our Discord (Coming Soon).

📄 License

This project is licensed under the MIT License.