Question-Answering with Astra DB and LangChain: A Vector Search Demo

Overview

This project demonstrates how to build a question-answering (QA) system using LangChain, OpenAI, and Astra DB. The system processes a PDF document, stores its content in a vector database, and allows interactive querying to retrieve relevant information.

Introduction

This project aims to create an efficient QA system by leveraging advanced NLP (Natural Language Processing) techniques. It uses:

• LangChain: A framework for building applications with large language models (LLMs).
• OpenAI: Provides language models and embeddings to process and understand the text.
• Astra DB: A scalable, cloud-native database that stores and retrieves text chunks as vectors.

Prerequisites

To run this demo, ensure you have the following:

1. Serverless Cassandra with Vector Search Database on Astra DB: https://accounts.datastax.com/session-service/v1/login
   • Obtain a DB Token with the role of Database Administrator.
   • Copy your Database ID. These connection parameters will be required shortly.
   • For detailed instructions, refer here: https://docs.datastax.com/en/astra-db-serverless/get-started/quickstart.html#_prepare_for_using_your_vector_database
2. OpenAI API Key:
   • You will need an API key from OpenAI for this demo to function. https://cassio.org/start_here/#llm-access

Technologies Used

• LangChain: Framework for building applications with LLMs.
• OpenAI: Provides the GPT models used for embeddings and language understanding.
• Astra DB: Cloud-native database service by DataStax.
• PyPDF2: Library for reading PDF files.
• Datasets: Hugging Face library for loading and processing datasets.

Abbreviations and Definitions

1. QA (Question-Answering): A type of information retrieval that involves answering questions posed by users based on a given dataset or document.
2. NLP (Natural Language Processing): A field of AI that focuses on the interaction between computers and human languages.
3. LLM (Large Language Model): A type of machine learning model trained on a large dataset of text to understand and generate human-like text.
4. Vector Store: A database that stores text data as vectors (numerical representations) to enable efficient similarity search.
5. Embedding: A representation of text data in a continuous vector space, often used to measure the similarity between different pieces of text.

Thank you for visiting my repository! Your interest and support are greatly appreciated.