LlamaIndex provides NLSQLTableQueryEngine and SQL-focused components for building text-to-SQL applications with emphasis on data retrieval quality. Open-source (MIT license) Python library. The NLSQLTableQueryEngine provides simpler, more focused text-to-SQL than LangChain’s agent approach—better for straightforward query generation scenarios. SQLTableRetrieverQueryEngine adds semantic table selection for large schemas—embedding table descriptions and retrieving relevant tables before query generation. ObjectIndex enables storing and retrieving SQL table schemas alongside other document types. Key advantage: schema retrieval optimized for context window management when databases have hundreds of tables. Supports PostgreSQL, MySQL, SQLite through SQLAlchemy. Integration with vector stores (Pinecone, Weaviate, Chroma) for schema embeddings. For production: combine with LlamaIndex evaluation framework for accuracy measurement. Page should cover: NLSQLTableQueryEngine setup, table retrieval for large schemas, ObjectIndex configuration, comparison with LangChain SQL Agent, evaluation framework usage, and scaling strategies for complex databases.