NerfSQL is a schema-aware text-to-SQL system that converts natural language queries into executable SQL, validates them against a live database, and iteratively corrects errors until a valid result is produced.

The system uses:

• Retrieval-Augmented Generation (RAG) over database schema

• A constrained LLM loop for SQL generation and correction

• Execution-based validation against a real database

• Stateful orchestration via LangGraph

• Fast inference using Groq models