GetStream/Vision-Agents: Open Vision Agents by Stream. Build Vision Agents quickly with any model or video provider. Uses Stream's edge network for ultra-low latency.

Vision Agents is an open-source Python framework for building real-time video AI agents that can watch, listen, and understand video streams. It provides low-latency (<30ms) video processing through Stream's edge network, integrates with major AI providers (OpenAI, Gemini, Claude), and supports pluggable video processors like YOLO and Roboflow. The framework includes native SDKs for multiple platforms, built-in features like turn detection and voice activity detection, and extensive integrations with 30+ services including speech-to-text, text-to-speech, and LLM providers. Example applications include sports coaching, security cameras with theft detection, and phone-based RAG systems.