**Iris: An AI-Driven Virtual Tutor for Computer Science Education** This paper introduces Iris, a chat-based virtual tutor integrated into the interactive learning platform Artemis. Iris provides personalized and context-aware assistance to computer science students, guiding them through programming exercises. The system is designed to offer subtle hints and counter-questions to foster independent problem-solving skills, avoiding complete solutions. Iris uses a Chain-of-Thought approach with GPT-3.5-Turbo, incorporating role prompts and few-shot learning to enhance its effectiveness. An empirical evaluation shows that students perceive Iris as effective, understanding their queries and providing relevant support. While students find Iris valuable for programming exercises and homework, they also feel confident solving tasks without it in exams. The findings highlight students' appreciation for Iris' immediate and personalized support, though they view it as a complement rather than a replacement for human tutors. Iris creates a safe space for students to ask questions without fear of judgment. The study addresses research questions about Iris' effectiveness, students' comfort with asking questions, and their reliance on Iris. The results indicate that Iris is well-received for its ability to understand queries and provide assistance, enhancing understanding and engagement. Students feel comfortable asking questions to Iris, but they prefer human interaction for certain contexts. Iris is seen as a helpful tool for routine tasks, but students remain confident in their abilities to solve exams without it. Future research should explore Iris' context awareness, compare it with general-purpose AI tools, and investigate factors that influence student engagement. The integration of different LLMs and enhancing Iris' access to student code are also recommended to improve its effectiveness and user experience.**Iris: An AI-Driven Virtual Tutor for Computer Science Education** This paper introduces Iris, a chat-based virtual tutor integrated into the interactive learning platform Artemis. Iris provides personalized and context-aware assistance to computer science students, guiding them through programming exercises. The system is designed to offer subtle hints and counter-questions to foster independent problem-solving skills, avoiding complete solutions. Iris uses a Chain-of-Thought approach with GPT-3.5-Turbo, incorporating role prompts and few-shot learning to enhance its effectiveness. An empirical evaluation shows that students perceive Iris as effective, understanding their queries and providing relevant support. While students find Iris valuable for programming exercises and homework, they also feel confident solving tasks without it in exams. The findings highlight students' appreciation for Iris' immediate and personalized support, though they view it as a complement rather than a replacement for human tutors. Iris creates a safe space for students to ask questions without fear of judgment. The study addresses research questions about Iris' effectiveness, students' comfort with asking questions, and their reliance on Iris. The results indicate that Iris is well-received for its ability to understand queries and provide assistance, enhancing understanding and engagement. Students feel comfortable asking questions to Iris, but they prefer human interaction for certain contexts. Iris is seen as a helpful tool for routine tasks, but students remain confident in their abilities to solve exams without it. Future research should explore Iris' context awareness, compare it with general-purpose AI tools, and investigate factors that influence student engagement. The integration of different LLMs and enhancing Iris' access to student code are also recommended to improve its effectiveness and user experience.