ObserveAI - Observe. Understand. Act.

ObserveAI — Project Description

ObserveAI is a smart system that turns any video feed into an active, decision-making assistant.
Instead of relying on people to watch cameras 24/7 and take manual actions, ObserveAI connects vision, language, and real-world tools to help teams see, understand, and act instantly.

We built a full working demo where users can give natural language prompts like “Let me know if there’s a long queue at the supermarket and lower the AC if it gets crowded.”
ObserveAI understands the video using Gemini 2.5 Flash, reasons using OpenAI, and runs actions through a LangGraph agent that connects to real-world tools — like changing music, adjusting AC, calling security, or notifying staff.

The system runs on Streamlit for the user interface, and we integrated multiple MCP (Model Context Protocol) tools to demonstrate how you can control external services based on what’s happening on camera.

We believe this idea has real impact in places like gyms, coffee shops, supermarkets, concerts, and more — where camera data already exists but isn’t being used to its full potential.

What makes it special is the combination of live vision analysis, natural language control, and real-world automation — all in one connected system that’s easy to use and actually works.