Multi-Agent Orchestration for Enhanced Text-to-SQL Generation: A Schema-Aware Approach with Self-Correction

Abstract: The unification of the functionality of databases and natural language processing Databases has traditionally only provided a means to be able to write SQL queries, however there is an ongoing challenge of converting NL (Natural Language) queries into Structured Query Language(SQL) [1].However, LLMs have significant challenges in handling complex schema structures and long-range dependencies, as well as a tendency toward structural hallucination [3]. This paper presents a novel multi-agent architecture based on LangGraph with the aim to close the gap between NL and SQL through specialized decomposed reasoning [7]. Our implemented system incorporates a robust AST-based Self- Correction loop [19], a query planning agent for chain-of-thought reasoning, and a semantic-aware Schema Linking Agent using Sentence Transformers. The Schema Linking Agent alone achieves (94.0%), according to experimental results on 50 examples from the Spider benchmark. Execution Accuracy: +2.0 percentage points better than the monolithic zero-shot baseline (92.0%). With an average correction rate of 0.06, the complete four-phase system achieves 88.0% EX, indicating that chain-of-thought planning significantly lowers structural hallucinations. Semantic value mapping, dynamic schema discovery, and query complexity routing are identified as crucial paths for future enterprise deployment by a thorough gap analysis. [10], [11].

Keywords: Text-to-SQL, Multi-Agent Systems, Large Language Models, Schema Evolution, Self-Correction, Vector Search, Cross-Database Reasoning.