This repository demonstrates advanced agentic patterns for real-time voice applications using OpenAI's Realtime API. It targets developers looking to prototype multi-agent voice systems, enabling sequential agent handoffs, background escalations to more capable models, and state-machine-driven data collection for tasks like authentication.

How It Works

The system orchestrates multiple agents based on a defined graph, allowing for sequential handoffs and conditional routing. It leverages a state-machine approach within agent prompts to ensure accurate, step-by-step data collection, including character-by-character confirmation for sensitive information. For complex or high-stakes decisions, it can escalate to more powerful models like o1-mini in the background.

Quick Start & Requirements

• Install dependencies: npm i
• Add OPENAI_API_KEY to your environment variables.
• Start the server: npm run dev
• Access the demo at http://localhost:3000.
• Requires Node.js and npm.

Highlighted Details

• Demonstrates sequential agent handoffs inspired by OpenAI Swarm.
• Features background escalation to more intelligent models for critical decisions.
• Enables state-machine prompting for precise data collection (e.g., authentication).
• Includes example agent configurations for authentication and customer service flows.

Maintenance & Community

• Core Contributors: Noah MacCallum, Ilan Bigio.
• No explicit community links (Discord/Slack) or roadmap are provided in the README.

Licensing & Compatibility

• The README does not specify a license.

Limitations & Caveats

The project is described as a "simple demonstration" and a "minimal demo," suggesting it may not be production-ready without further development. The lack of a specified license could pose compatibility issues for commercial or closed-source use.