Retrieveal Augmented Generation Model using LangChain, Weaviate, and OpenAI ChatGPT

Overview:

This project leverages a RAG, or Retrieval Augmented Generation, model and incorporates a large language model to retrieve answers to questions only from specified materials. The recommended readings, lecture notes, and syllabus from CS89B Natural Language Processing have been used to test RAG model application in a working chatbot. Some of the downsides of large language models are a tendency toward hallucination (made up answers) and a lack of specific referenced content to validate information returned by the model. A RAG model helps to mitigate these downsides by drawing answers from specific content and by creating traceback mechanisms to the sources of model answers.

The ‘Technical How to Guide’ is meant to provide reference materials in how to get the code base to work. Implementation requires coordination of multiple packages with interdependencies, API keys and use of the Linux subsystem on a Windows machine. The final presentation delves further into methodology and results of the tests for the RAG model implementation to support queries of general course materials based on the course content and a quick test of the model’s ability to answer the Quiz questions from the course.

Project Deliveables (Links to GitHub):

• Python Code
• Technical How to Guide

Featured Image: