Loading Large CSV in Pinecone for Question Answering

I am building a question answering bot over a larger CSV file 70K rows. I used text to SQL approach basically dumping data into sqlite and then asking llm to generate SQL queries to answer the user queries. This approach works only for basic questions. For complex queries the LLM is not able to generate proper SQL queries. Also I have a csv file which contains metadata about the column definition for the 70k file.

How can I use pinecone for this bot.Will using Pinecone improve the QA ability . Any suggestions are welcome.