Stream Realtime AI Interactions via WebRTC & WebSockets

This collection provides Python code samples demonstrating how to build real-time, streaming interactions using WebSocket and WebRTC protocols. It includes examples for both simple chat implementations and advanced scenarios with function calling, enabling developers to create responsive agents that communicate in real time.

Use these samples when you need to implement low-latency, bidirectional conversations where immediate response streaming is critical. Use the WebRTC examples (realtime_agent_with_function_calling_webrtc.py, simple_realtime_chat_webrtc.py) for peer-to-peer audio/video scenarios, and the WebSocket examples (realtime_agent_with_function_calling_websocket.py, simple_realtime_chat_websocket.py) for text-based or server-mediated streaming.

Do NOT use these samples for batch processing or scenarios where request-response latency is acceptable, as the real-time infrastructure adds complexity. Avoid WebRTC implementations if you only need server-client text streaming, where WebSocket patterns are simpler and sufficient.

You provide the configuration and define any functions your agent should be able to call during real-time interactions. The samples include a utils.py helper module to support the implementation. You receive working Python code examples that establish real-time connections, stream responses, and optionally handle function calling within the conversation flow. The function calling examples demonstrate how to extend agent capabilities with executable tools during live interactions.

These samples are designed for Python developers building real-time streaming applications. If you're comparing protocols, the collection provides both WebSocket and WebRTC implementations so you can choose based on your transport requirements - WebRTC for direct peer connections with media support, WebSocket for server-mediated text and data streaming.