Emotion Aware Code Tutor Using LLM ,Eye Gaze and Typing Pattern

Learning programming is not only about writing correct code.Students may become confused, lose focus or feel confident while solving a problem, but most coding platforms do not consider these changes.This paper presents an Emotion-Aware Code Tutor that combines coding practice with behavioural monitoring and AIbased assistance.The system is developed as a web application using Python and Django, with SQLite used for storing user and coding-related information.A local Qwen2.5-Coder:3Bmodel through Ollama is used to generate coding questions, provide hints, and evaluate submitted programs.During coding, the system monitors facial expressions, eye-gaze direction and typing activity to understand the learner's current behavioural state.When the learner appears to be struggling, the system provides basic hints, while a confident state can lead to a more challenging hint.The platform also records coding attempts, scores, hints used, gazeaway events, evaluation feedback, and performance history.The system provides immediate feedback after code submission and helps learners understand their programming mistakes.The stored performance information can be used to review progress and identify areas that need improvement.By combining behavioural information with AI-based coding assistance, the system reduces the limitation of conventional coding platforms that provide the same type of support to all learners.The developed approach provides a more interactive and personalized coding-learning environment and demonstrates the potential of combining LLM assistance with learner behaviour analysis.