AI-Powered Chatbot for Mental Health

Abstract

Mental health is one of the critical problems globally that is impeded by concerns of stigma, price, and accessibility. Although AI-driven chatbots are a promising solution, most current platforms are based on cloud-based architecture due to which the data privacy and latency issue are of a significant concern. In this project, I present a secure and responsive mental health chatbot, Axon Ally, focused on the privacy of the user and providing real-time assistance. The system is developed on a hybrid architecture, which is a unique mixture of a locally hosted, fine-tuned Large Language Model (LLaMA 3.2 1B Instruct) and the secure cloud capabilities of Firebase by Google. The frontend is a cross-platform mobile app created in Flutter, which offers a user-friendly and relaxing user interface. As a middleware, the backend is a FastAPI server that receives user requests and serves as the interface with the local AI model and safely performs data transactions with Firebase. Firebase Authen tication and Firestore perform user authentication and conversation storage respectively, which is encrypted and provides the security measures of any sensitive data with the industry-standard. The present project can prove that it is possible to implement a lightweight but efficient AI chatbot that does not rely on third-party AI services. The localization of the language model ensures privacy of the users and reduces the response time. The study confirms that this architecture offers a curious, reliable, and approachable platform to provide early support of mental health to critical disavowal between technological creativity and ethical client treatment. Axon Ally is an example of AI that can be used in sensitive areas as a demonstration of privacy-focused applications.