I am trying to build a knime workflow where we can plug a csv or database reader to GPT4all connector in a way that it uses this file as the local database for answering questions. Is there a way we can do this?
@Pranjal_m here is one example using knime 4.0 and csv. You can also check the linked articles there
This example uses gpt4all and the newer nodes. Here it parses pdfs but you could feed the lines of CSV files.
Depending on the number of documents you provide this will be the information that will be drawn from the vector store so matching your question so there LLM will have this as additional information not the whole CSV.
This approach tries to employ agents to scan a database and come up with SQL code that can then answer the question based on the whole DB. Challenge is to get the SQL dialect right.
2 Likes