In the event of indoor disasters, evacuation processes face a multitude of real-time constraints, including the limited time available for a successful evacuation and the necessity of guiding occupants clearly and safely. These constraints are not limited to office and residential buildings but also apply to airports, hospitals, cruise ships and more. Tragic incidents such as Grenfell Tower’s deadly 2017 fire and that of the shopping mall in Russia in 2018 both serve as poignant examples, where sudden and rapidly spreading fires resulted in the loss of + 70 and 62 lives respectively. These events highlight the critical need for efficient evacuation plans to ensure quick and safe evacuations. Prior research conducted on this issue, employs various computational disciplines such as heuristic programming, machine learning and network optimization, with varying degrees of success. This paper introduces a novel reactive-planning framework that incorporates real-world knowledge of building topology and planning policies into a knowledge-based planner and an event-driven replanner. This approach aims to enhance the efficiency and effectiveness of evacuation processes by dynamically adapting to changing conditions and providing real-time guidance to occupants.