Conventional surveillance systems that depend on manual monitoring often struggle to promptly detect emergencies such as fires or traffic accidents, leading to delayed responses and increased risk of loss. To address these challenges, this project introduces an Application Programming Interface (API) designed for emergency event detection, leveraging the advanced YOLOv8 (You Only Look Once Version 8) deep learning model in conjunction with security camera networks. The proposed solution enables real-time identification of fires and accidents from both uploaded video files and live camera feeds. The system is developed using Python, FastAPI, and OpenCV, ensuring rapid and accurate video analysis even on standard hardware configurations. Experimental evaluations indicate that the API achieves an accuracy exceeding 80% and maintains a processing speed of at least 10 frames per second (FPS), underscoring its effectiveness and suitability for deployment in contemporary security environments.