Local LLM module in a web system for generating textual reports via prompt

Abstract

This study addresses the need to enhance the information systems of the Laboratório de Práticas da Computação at Universidade Franciscana (UFN), with a particular emphasis on human-computer interaction. The existing systems lacked advanced capabilities for automated report generation, relying instead on complex manual SQL queries. To address this limitation, a textual interaction module was developed, integrating Natural Language Processing (NLP) techniques and Large Language Models (LLMs) to translate natural language questions into SQL queries, thereby streamlining the report generation process. The implementation employed the Llama 3.1 8B model and leveraged the LangChain framework, together with the Ollama and xmlrpc libraries, to support a Retrieval-Augmented Generation (RAG) architecture. The proposed solution produced consistent and accurate results for short, well-structured queries when compared to traditional manual SQL queries.