Chatty Shoes is designed with a highly responsive, scalable and robust backend, leveraging state-of-the-art technology to facilitate user-friendly online shoe shopping via natural language interactions. The backbone of the system is a serverless architecture based on Cloud Functions. This decentralised structure comprises three primary functions: Session Creation, Message Management and Information Retrieval. Session Creation: This function is responsible for initiating conversations with customers. Every conversation represents a unique session, providing the framework for interactive and dynamic dialogues. The data from these sessions, including the conversation history, is securely stored in Firestore, ensuring a persistent and seamless user experience across different sessions. Message Management: Powering the heart of the conversation is a GPT-3.5-turbo, a language model renowned for its ability to understand and generate human-like text. This feature enables Chatty Shoes to handle Frequently Asked Questions (FAQs) within the context of the conversation, thereby giving users instant responses to their queries. Additionally, it provides the AI agent with the ability to perform information retrieval operations on products when the system deems it necessary, making it proactive and more engaging. Information Retrieval: The last key component of the backend is the Semantic Search function, which operates over Pinecone's vectorial database. This feature enhances Chatty Shoes' product recommendation abilities by retrieving product data based on the contextual understanding of user inputs rather than merely keyword matching. This means the AI can effectively understand and respond to nuanced customer preferences, thus improving their overall shopping experience. Together, these elements synergise to create a powerful backend that underpins Chatty Shoes' mission: to revolutionize online shoe shopping by facilitating better, deeper and more intuitive conversations.