Multilingual AI Chatbot

The rapid advancement of conversational artificial intelligence has shifted focus from massive, API-dependent Large Language Models (LLMs) to efficient, locally deployable Small Language Models (SLMs). This research presents the development of a Multilingual AI Chatbot utilizing the Microsoft Phi-3:mini model for text generation and the Moondream model for multimodal visual reasoning. The system is built on a high-performance stack comprising a React frontend, FastAPI backend, and MongoDB for persistent data storage. A key feature of the implementation is the support for real-time response streaming via Server-Sent Events (SSE) and multimodal input processing using Base64 image encoding. Experimental analysis indicates that this architecture provides a secure, low-latency, and contextually aware conversational experience suitable for data-sensitive environments.