AI-Driven Mental Health Chatbot: Empowering Well-Being with Conversational AI and Retrieval-Augmented Generation

Mental health issues, including stress, anxiety, and depression, have seen a significant rise globally, impacting millions of lives and creating an urgent need for innovative solutions. While traditional mental health services are beneficial, barriers such as social stigma, high costs, and limited availability hinder widespread access. AI-driven chatbots offer a promising avenue for scalable, accessible mental health support. However, a notable research gap exists in developing chatbots that can generate empathetic, contextually accurate responses grounded in reliable mental health literature. In this study, two Retrieval-Augmented Generation (RAG) models are proposed: one employs all-MiniLM-L6-v2 embeddings with FAISS for efficient retrieval and OpenAI GPT for response generation, while the other leverages GoogleGenerativeAIEmbeddings, Pinecone for vector storage, and BART for response generation. Both models were implemented in Google Colab, designed to deliver personalized mental health assistance based on trusted resources. Experimental evaluation reveals that the chatbot produces highly empathetic and context-aware responses, effectively addressing user needs. The results show that our model achieved a 0.85 cosine similarity score.