This repo contains a notebook, a simple use case of Llama assistant model for answering questions related to car handling. This is a simple Retrieval-Augmented Generation, aka RAG model, that extracts related parts from the manual according to the user's query for a clear, validated answer by the assistant.
In this use case, we use the manual document of the Toyota Verona car available in this file:
- Car manual original PDF
- This document is created by converting the .pdf to a text file to be used in the RAG model: Car manual text version
ToDo:
- More document cleaning is needed as there is some irrelevant information in the file eg. removing the indexes, and image captions and making it more structured.
- Improve the system instructions
- Create a set of test questions to assess the performance of the answers and detect the hallucination