Research Paper Answer Bot with Agentic Corrective RAG

This project implements a Research Paper Answer Bot using Agentic Corrective Retrieval-Augmented Generation (RAG) to answer questions about key research papers in Generative AI. The bot leverages a vector database and a multi-retriever setup to fetch relevant context from papers like GPT-4, InstructionGPT, Mistral, and others, generating detailed answers with cited sources.

---

Project Overview

The bot is designed to:

• Retrieve relevant content from research papers on Generative AI using a vector database with embedding models.
• Use corrective RAG strategies, including web search, when relevant documents are lacking in context.
• Display source documents alongside generated answers to ensure transparency and reference accuracy.

---

Getting Started

Clone the Repository

git clone https://github.com/your-username/ResearchPaperAnswerBot.git
cd ResearchPaperAnswerBot

Running the Project

• Open Agentic_RAG.ipynb in Jupyter Notebook or JupyterLab or googleColab.
• Update file paths as needed.
• Execute the cells in the notebook to initialize and run the project.

---

Project Details

• Vector Database: Implemented with ChromaDB for fast, efficient document retrieval.
• Retrievers: Multiple retrieval techniques (cosine similarity, re-rankers, hybrid retrieval) enhance context relevance.
• Agentic Corrective RAG: Corrective actions like web search are applied when no relevant documents are found, ensuring comprehensive responses.

---

Web Search Integration

• When local document retrieval does not provide enough context, the system automatically performs a web search and integrates relevant results into the generated response.