The post asks if CSV files can be used instead of PDFs with LlamaIndex for the knowledge base. The comments suggest that using CSV files can be a bit tricky, and the recommended approach is to load the CSV into a SQL database or a dataframe, and then use the Pandas query engine or text-to-SQL. For highly numerical CSVs, the community members suggest using the paged CSV reader for the best results.
But using CSV is a little tricky, the recommended is to load your CSV into a SQL database (without LLMs, just with code) or dataframe and then use pandas query engine or the text-to-sql
