NATURAL LANGUAGE DATABASE HELPER: AN INTELLIGENT AGENTIC WORKFLOW FOR AUTOMATED SQL QUERY GENERATION

The Natural Language Database Helper (NLDH) is a smart web-based application that aims to provide a technical interface between human language and complicated database management systems. In the current data intensive environment, retrieving data may demand expertise in Structured Query Language (SQL) and this is a major challenge to the non-technical stakeholder. This project enhances easy interaction with the database by enabling users to query relational databases using plain English. The system is based on the Gemma2:9B model using Ollama by processing large language models (LLM) locally, which guarantees high performance and data privacy. The fundamental novelty is a Two-Step Agentic Workflow: the LLM will find the tables that are relevant to the query and schema overview provided by the user; it will then use the finer metadata of the respective tables to come up with an exact and executable SQL query. The application is a decoupled architecture with an interactive JavaScript front-end and a strong back-end based on PHP to secure communications via API. This helps to increase the efficiency of operations, minimize human syntax errors, and democratize the data access.