AlazarG19/RAG-LLM-Chatbot-For-Ethiopia-Ecommerce-Site-Jiji

# RAG LLM for Ecommerce Site Jiji ## Project Overview This project mainly focuses on providing various information and insights to customers who want to know more about the different items that are listed in the website. The project involves scrapping various data from jiji and using that as context to create a rag model that can be used by users to answer various prompt. ### Data The data was collected using a webscraper that was made by selenium. It has some limitations like - Only provides Item title, description ,price and location of the product - Doesn't provide all data provided for that certain item because a headless browser is used to extract the resource we can be resource-intensive. The data contains the title, description, price and location information. This is saved to a csv file initally before preparing the embeddings. ## Tools used ### Selenium The main component used for the webscrapper is a browser automation. This requires the chromedriver to be setup and some limitations that had occured was some sites use scrapper prevention tool so for extra precaution we used selenium stealth. The scrapped data is saved to a csv file for later use. ### LangChain. Langchain is an open-source framework designed to simplify the creation of applications using LLMs. The llm used for this case is Gemini but it was through langchain we were able to retrieve the various libraries used to create embeddings, ask various question and create a prompt template. We created various embeddings from the csv file created from the webscrapper to store to the vector database using langchain. ### Weaviate Weaviate is a popular vector database that is used for many ai applications. We used the embedding converted by Langchain(Gemini) to create a collection and store the various vectors. This will be a huge resource to select the specific context we will send to langcain. ### Streamlit Streamlit is an open-source framework designed to create beautiful web applications for …