[How to: deal with CSV files](https://python.langchain.com/docs/how_to/sql_csv): LLM should read this page when needing to build question-answering systems over CSV data, wanting to understand the tradeoffs between using SQL or Python libraries like Pandas, and requiring guidance on securely executing code from language models. This page covers two main approaches to question answering over CSV data: using SQL by loading CSVs into a database, or giving an LLM access to Python environments to interact with CSV data using libraries like Pandas. It discusses the security implications of each approach and provides code examples for implementing question-answering chains and agents with both methods.

