Design and Implementation of a RAG-Based Conversational Assistant for Industrial Technical Support

In modern industrial and service contexts, the ability to provide rapid and highly accurate technical assistance is crucial for maintaining operational efficiency and reducing downtime. When such support is delivered through an automated conversational agent, it becomes a concrete asset that improves human operators, especially in companies managing large, heterogeneous collections of technical documents. This is the case of IMESA S.p.A., a manufacturer of professional laundry equipment that serves an international customer base and requires continuous, multilingual post-sales support. This thesis presents the design and implementation of an advanced assistive chatbot based on Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG). The system grounds its answers on the company's proprietary documentation, handling complex layouts, tables, and images, and delivers precise, verifiable responses in the user's language, significantly reducing the risk of hallucination. The development includes document ingestion, retrieval optimization, generation, and a finite-state machine (FSM) that governs the conversational flow. Experimental results show that the chatbot provides accurate assistance and meets the reliability requirements for its deployment in the industrial field.